Geophysical Research Letters

Published by Wiley
Online ISSN: 0094-8276
Learn more about this page
Recent publications
A schematic of overlapping cold wakes (upper panel) and the evolving vertical temperature profiles (lower panel) at the overlapping location (the gray dot in the upper panel). Cold wake 1 and cold wake 2 are two successive cold wakes induced by the same tropical cyclone. ΔSST1 (ΔSST2) is the SST cooling for the cold wake 1 (2).
Annually accumulated Tropical cyclone (TC)‐induced sea surface temperature (SST) cooling (unit: °C) averaged over 1982–2016 with double counting (DC, a) and without DC (NDC, b). The contour lines of climatological seasonal amplitude of SST are added in panel (b) for comparison. The magnitude of SST seasonal cycle is calculated as the difference between maximum and minimum climatological monthly SST. Annually accumulated TC‐induced diffusivity (unit: cm²/s) averaged over 1982–2016 with DC (c) and without DC (d). The area‐weighted average SST cooling (blue) and diffusivity (red) within 40°S–40°N over 1982–2016 with DC (e) and without DC (f). Dots: AVHRR‐only, diamonds: AVHRR + AMSR for inter‐validation. RT stands for relative trend (linear trend divided by the mean value). Linear trends are obtained from a least square method and P‐values using Student's t test are shown. Note the different scales in (a–d).
Area‐weighted averages of the annually accumulated Tropical cyclone (TC)‐induced sea surface temperature (SST) cooling (ΔSST, unit: °C, blue) and diffusivity (unit: cm²/s, red) in major ocean regions from 1982 to 2016. All the results are without DC. Absolute trends (unit: °C/decade for SST cooling and cm²/s/decade for diffusivity) and relative trends (unit:/decade, in the parentheses) are labeled. (a) GL: global; (b) NH: Northern Hemisphere; (c) SH: Southern Hemisphere; (d) NWP: Northwest Pacific,100°E−180°, 0°–40°N; (e) NEP: Northeast Pacific, 180°–90°W, 0°–40°N; (f) NA: North Atlantic, 90°W–30°E, 0°–40°N; (g) NI: North Indian Ocean, 30°E−100°E, 0°–40°N; (h) SP: South Pacific, 135°E−70°W, 40°S–0°; (i) SI: South Indian Ocean, 10°E−135°E, 40°S–0°. Linear trends are obtained from a least square method. Trends with P‐values less than 0.05 are marked with asterisks.
Time series of standardized factors influencing the Tropical cyclone (TC)‐induced sea surface temperature (SST) cooling and diffusivity, including upper ocean stratification (red), annual average TC translation speed (yellow), TC intensity (green), TC number (blue), TC lifetime (pink) and TC size (violet) from 1982 to 2016 in ocean basins defined the same as in Figure 3. The original time series are smoothed using a 5‐year moving average. Dashed lines are linear fits using a least square method. Linear trends with P‐values less than 0.05 are marked with asterisks.
Tropical cyclones (TCs) are an important source of turbulent mixing for the upper ocean at low latitudes, causing sea surface cooling and subsurface warming. A new estimate of annually accumulated sea surface cooling and upper ocean diapycnal diffusivity induced by TCs is obtained by using quantified cold wake sizes, which were largely ignored by previous studies. Both the annually accumulated tropical cyclone‐induced sea surface cooling and upper ocean diffusivity on a global scale show a significant decreasing trend over the period of 1982–2016, at a rate of −0.09 ± 0.03 °C/decade and −0.03 ± 0.01 cm²/s/decade respectively. The strengthening of ocean stratification with global warming contributes to the decrease of sea surface cooling and mixing, while the changes of tropical cyclone characteristics (such as translation speed, intensity, number, lifetime and size) contribute differently in various ocean basins.
Deployment of the AQG‐B in the summit craters area of Mt. Etna volcano. (a) Sketch map of Mt. Etna showing the position of PDN, SLN, and MNT stations. The shaded yellow area encloses all the possible horizontal positions of the mass source (Figure 4 and text). Inset: aerial photo of the PDN observatory (Google Maps). (b) Picture of the AQG‐B. The height of the sensor head is 100 cm. (c) Panorama looking NW, from the slope to the S of the observatory, showing the proximity of PDN to the summit craters. (d) Section (not in scale) crossing the observatory and showing the installation configuration of the AQG‐B.
Sensitivity analysis on data from the AQG‐B. (a) Spectrogram of the vertical‐component seismic signal from the Pizzi Deneri station (EPDN seismic station; 1 August–3 December 2020). The green and orange boxes mark periods of low and high volcanic tremor amplitude, respectively. (b) Power spectral density of the seismic velocity recorded at EPDN, during the phases of low (green) and high (orange) tremor amplitude marked in the spectrogram. Black curves indicate the new high and low noise models from Peterson (1993). (c) Allan deviation of the corrected gravity data from the AQG‐B at PDN, during the low (green) and high (orange) tremor phases. The green and orange dashed lines in the Allan plot indicate a sensitivity of 1,200 and 1,600 nm/s²/τ $\tau $1/2, respectively.
Gravity time series and wavelet coherence analysis. (a) Gravity time series from PDN (1 August–3 December 2020), after removal of the perturbing effects shown in panels (b–f) of Figure S1 in Supporting Information S1. Data are averaged over 10‐min (gray curve), 1‐hr (black curve) and 5‐hr (red curve) intervals. (b) Gravity time series from SLN, after removal of Earth tide, atmospheric pressure, ground tilt and polar motion. Data are averaged over 10‐s (gray curve) and 10‐min (black curve) intervals. (c) Wavelet coherence between the signals in panels (a and b) Black arrows indicate the relative phase relationship (in‐phase pointing right, anti‐phase pointing left, and one signal leading the other by 90° pointing straight up or down). The cone of influence, where edge effects may distort the results, is shown as a lighter shade.
Gravity time series during high‐coherence phases and derivation of mass source depths. (a and b) Gravity time series from PDN (black curves) and SLN (red curves), during the first (h‐c 1) and second (h‐c 2) subintervals of high coherence. As for h‐c 2, the available signal from MNT is also shown (blue curve in panel (b). (c and d) Scatterplots of SLN against PDN, during h‐c 1, and SLN against PDN and SLN against MNT, during h‐c 2. For each scatterplot, the slope of the best‐fitting line (s.) and the correlation coefficient (c. f.) are reported. (e) PDN/SLN (red area) and MNT/SLN (blue area) gravity amplitude ratio versus depth of the mass source. The calculations are performed for all the possible positions of the mass source enclosed in the yellow area of Figure 1a. Dashed segments mark the values of the observed amplitude ratios during h‐c 1 and h‐c 2.
Plain Language Summary Mass redistributions occurring in the Earth's interior, for example, when a magma batch is displaced through the feeding system of an active volcano, may induce tiny changes in gravity over time, measurable on the ground. Measurement of such changes requires high‐precision devices, namely, the gravimeters, which can detect variations as small as one part in 10⁹ of the gravity acceleration on Earth. However, standard gravimeters are not ideally suited for use in harsh field conditions, especially when continuous measurements are the target. Recent advances in quantum technology have allowed the development of a portable gravimeter which can successfully operate under field conditions. Here we present the world's first application of this quantum gravimeter for monitoring and studying an active volcano. The device was deployed only 2.5 km away from the summit active craters of Mt. Etna volcano (Italy) and has provided a high‐quality gravity time series. Inspection of this time series highlighted gravity changes which are reflective of bulk volcanic processes, involving magma and exsolved gas in the upper part of Mt. Etna's plumbing system.
2008–2010 June, July, and August relationship between idealized tracer χ40−50 and daily maximum 2‐m temperature T from GEOS‐Chem simulations. (a) Daily regression slope dχ40−50/dT. Regressions that are not statistically significant are hatched (p > 0.05). (b) Mean meridional gradient of tracer ∂ϕχ40−50. (c) Mean meridional gradient of temperature ∂ϕT. (d) Gradient ratio ∂ϕχ40−50/∂ϕT. Regions with |∂ϕT| < 0.2 K/° are hatched (backslash), superimposed on forward slash from (a). (e) Scatter plot of ∂ϕχ40−50/∂ϕT versus dχ40−50/dT averaged over 10° latitude × 20° longitude domains, colored by ∂ϕT as in (c). Symbols show regions binned by |∂ϕT|. Dashed line shows the 1:1 slope. RMSE between gradient ratio and regression for each |∂ϕT| bin including outliers (Figure S3a in Supporting Information S1) is indicated.
Idealized tracer scatter plots averaged over 10° latitude × 20° longitude regions. (a) June, July, and August (JJA) gradient ratio ∂ϕχ20−30/∂ϕT versus regression dχ20−30/dT, colored by ∂ϕT with three |∂ϕT| bins (in K/°) shown by symbols. (b) Same as (a) but for DJF ∂ϕχ40−50/∂ϕT versus dχ40−50/dT. (c) JJA ∂ϕχ40−50/∂ϕQ versus dχ40−50/dQ, colored by specific humidity gradient ∂ϕQ with three |∂ϕQ| bins (in g/kg/°) shown by symbols. (d) JJA ∂ϕχ20−30/∂ϕχ40−50 versus dχ20−30/dχ40−50, colored by ∂ϕχ40−50 with three |∂ϕχ40−50| bins (in ppm/°) shown by symbols. Dashed lines show the 1:1 slope. RMSE between gradient ratio and regression for each bin including outliers (not shown) is shown in brackets.
Same as Figure 1, but for 2008–2010 June, July, and August O3‐T relationship from GMI simulations. (a) Daily regression slope dO3/dT. (b) Mean meridional gradients of ozone ∂ϕO3. (c) Mean meridional gradients of temperature ∂ϕT. (d) Gradient ratio ∂ϕO3/∂ϕT. (e) ∂ϕO3/∂ϕT versus dO3/dT. Open symbols and RMSE in brackets are from the transport‐only simulation. See Figure S3b in Supporting Information S1 for outliers.
The daily variation of ground‐level ozone (O3), a harmful pollutant, is positively correlated with air temperature (T) in many midlatitude land regions in the summer. The observed temporal regression slope between O3 and T (dO3/dT) is referred to as the “ozone‐climate change penalty” and has been proposed as a way to predict the impact of future climate warming on O3 from observations. Here, we use two chemical transport models to show that the spatial variation of dO3/dT is primarily determined by simultaneous meridional advection of O3 and T. Furthermore, the sign and magnitude of dO3/dT can be approximated by their climatological meridional gradient ratio (O3 gradient divided by T gradient). Consideration of expected changes in the meridional gradients of T and O3 due to climate change indicates that dO3/dT will likely change. Caution is needed when using the observed climate penalty to predict O3 changes.
Experiments in a water‐recirculating flume to visualize the release of fluorescent dye from the sediment into the surface water. Refractive‐index‐matched sediment and translucent vegetation dowels were used. A green, fluorescent dye was injected into the sediment and a blue lamp was used to excite the dye.
The concentration of the fluorescent dye in the sediment, represented by the fluorescence intensity of the emitted green light, decays over time. The flow was started at time = 0 hr. The black and red symbols represent the fluorescence intensities relative to the background intensities in channels without vegetation and with vegetation of volume fraction ϕv=0.05 ${\phi }_{v}=0.05$, at flow velocities of 4.0 and 3.6 cm/s, respectively. The black and red solid curves represent the fits of the measurements to the mass transfer model (Equations 1 and 2) with R2=0.99 ${R}^{2}=0.99$ for both cases. The model fits were conducted when the streamwise fluorescence intensity decreased uniformly (see Text S3 in Supporting Information S1 for details). For these two experiments, the horizontal area of the sediment‐water interface ASWI=0.19m2 ${A}_{SWI}=0.19\,{\mathrm{m}}^{2}$; the sediment porosity ϕs=0.3 ${\phi }_{s}=0.3$; the volume of pore space in the sediment Vol,s=1200±9mL ${V}_{ol,s}=1200\pm 9\,\text{mL}$; and the volume of surface water Vol,w=2830L ${V}_{ol,w}=2830\,\,\mathrm{L}$. The fitted parameters are VH ${V}_{H}$ and background image intensity. Dashed lines show the model predictions.
(a) The effective hyporheic exchange velocity VH ${V}_{H}$ versus mean flow velocity U $U$ for cases without vegetation (black) and with vegetation of a volume fraction ϕv=0.05 ${\phi }_{v}=0.05$ (red). The black solid line (y=(0.2x+0.4)×10−4 $y=(0.2x+0.4)\times {10}^{-4}$) and the red‐dashed line (y=(1.1x−0.3)×10−4 $y=(1.1x-0.3)\times {10}^{-4}$) represent linear fits to measurements without and with vegetation with R2=0.89 ${R}^{2}=0.89$ and R2=0.95 ${R}^{2}=0.95$, respectively. (b) VH ${V}_{H}$ versus the total near‐bed turbulent kinetic energy kt ${k}_{t}$ for cases without vegetation (black) and with vegetation (red). The black solid line (y=(7.2x+8.2)×10−5 $y=(7.2x+8.2)\times {10}^{-5}$) and the red‐dashed line (y=(7.2x+4.7)×10−5 $y=(7.2x+4.7)\times {10}^{-5}$) represent linear fits of the measurements without and with vegetation with R2=0.92 ${R}^{2}=0.92$ for both cases.
Plain Language Summary The exchange of contaminants and nutrients between surface and subsurface water in the hyporheic zone of rivers and wetlands controls water quality as well as the metabolism of benthic microbes and the associated biogeochemical cycle. Vegetation, which is ubiquitous in aquatic ecosystems, has been found to affect the surface‐subsurface exchange and as such impact water quality and stream biogeochemical cycle. However, how vegetation impacts this exchange remains unclear, making it difficult to predict the contaminant transport and biogeochemical cycle in streams, lakes, and coastal areas with vegetation. In this study, we directly visualized the release of a fluorescent dye from the transparent sediment into the surface water in a water‐recirculating tank filled with translucent vegetation. We discovered that vegetation can significantly increase the exchange in the hyporheic zone. Furthermore, we proposed a model to predict the impacts of vegetation on hyporheic exchange. We believe that this finding will help improve predictions of contaminant transport and biogeochemical cycle in streams and other aquatic ecosystems. The results of this study will also help ecologists design stream restoration projects that use vegetation to increase the retention and degradation of contaminants in sediment.
(a‐c) Step‐by‐step explanation of quantile line plots (see subplot titles) (d‐f) Magnitude estimate development for three different predictability models: Gutenberg‐Richter (GR) (not predictable during growth phase), skewed GR (not point‐predictable, but deviation from prior already during growth phase) and Gaussian (predictable). For each model, we use the same hypothetical event with prototypical triangular moment rate functions representing the first order moment release history; for predictable models to be viable, second order features of the moment rate function would differ between smaller and larger events. The dashed black line indicates the cumulative moment release. An additional visualization of the options at a fixed time is shown in Figure S3a in Supporting Information S1.
Distribution of events and histograms for magnitude distribution for the three STF datasets. The events are color coded by their data set. Ye et al. is plotted on top of USGS, on top of SCARDEC. This might lead to few events not being visible due to overlaps.
(a) Probability density functions (PDFs) calculated from the source time function (STF) model just before onset, and at 2, 4, 6 and 8 s after onset. Colored ticks on the PDFs indicate 0.05, 0.2, 0.5, 0.8, 0.95 quantiles. (b‐f) Example predictions from the STF model visualized by the 0.05, 0.2, 0.5, 0.8, 0.95 quantiles over time. (b) shows the same event as (a) The lower right gives information on the event. The black dashed line shows the moment released so far, that is, the trivial lower bound. The bottom plots show the STFs. The upper right indicates the STF database used. For a step‐by‐step explanation on the quantile plots, see Figures 1a–1c.
Average predicted Probability density functions (PDFs) based on source time function (STFs) (a)‐(f) and teleseismic waveforms (g)‐(h) grouped by magnitude bin. Left column shows results at time t after onset PM|Ot $\left(\mathbb{P}\left(M\vert {O}_{t}\right)\right)$, right column after cumulative moment equals magnitude M¯ $\bar{M}$ PM|OM¯ $\left(\mathbb{P}\left(M\vert {O}_{\bar{M}}\right)\right)$. The STF model has been trained on the SCARDEC data set and evaluated on each STF data set. See Figure S8 in Supporting Information S1 for STF results from a neural network trained with the USGS data set. PDFs were truncated in visualization to avoid overlap between different times/base magnitudes. Black dotted lines in (b, d, f, g) indicate the current base magnitude. The apparent skew between buckets in panel (b) and prediction difference in (g) for M¯=6.0 $\bar{M}=6.0$ likely results from SCARDEC processing artifacts. For determining tM¯ ${t}_{\bar{M}}$ in (h) we used the SCARDEC data set. See Figure S12 in Supporting Information S1 for plots with the other STF datasets. Events differ between panels (g) and (h): (h) only includes those events present in both the teleseismic data set and SCARDEC (∼3,500 events) and (g) all of the former (∼38,000 events).
Plain Language Summary Earthquakes are among the most destructive natural hazards known to humankind. While earthquakes can not be predicted, it is possible to record them in real‐time and provide warnings to locations that the shaking has not reached yet. Warning times usually range from few seconds to tens of seconds. For very large earthquakes, the rupture itself, which is the process sending out the seismic waves, can have a similar duration. Whether the final size of the earthquake, its magnitude, can be determined while the rupture is still ongoing is an open question. Here we show that this question is inherently probabilistic ‐ how likely is an event to become large? We develop a formulation of rupture predictability in terms of conditional probabilities and a framework for estimating these from data. We apply our approach to two observables: moment rate functions, describing the energy release over time during a rupture, and seismic waveforms at distances of several thousand kilometers. The final earthquake magnitude can only be predicted after the moment rate peak, at approximately half the event duration. Even then, it is impossible to foresee future subevents. Our results suggests that ruptures exhibit a universal initiation behavior, independent of their size.
(a) Photoproduction rate of Carbonyl sulfide (COS) in 0.2‐μm‐filtered seawater collected from the surface of the Indian Ocean. Production rates were calculated by dividing the net (light‐dark) COS concentration produced in the photochemical experiments by the incubation time. The value above each error bar is the mean production rate of COS. Error bars denote the standard deviation of measurements from multiple quartz bottles (n = 3). (b) Photochemical production rate of COS in the samples plotted against the initial absorption coefficient at 350 nm (a350). The vertical error bars denote the standard deviation of replicate samples (n = 3).
(a) Carbonyl sulfide (COS) concentration measured in 0.2‐μm‐filtered Indian seawater added 50 μM cysteine exposed to simulated sunlight with varying cutoff filters for 1 hr in quartz bottles. Full indicates no cutoff filter and dark indicates samples wrapped in aluminum foil. The value above each error bar is the concentration of COS. Error bars denote the standard deviation of multiple bottles for each spectral treatment (n = 3). (b) Relative contribution of ≤320 nm solar radiation (UVB) and 400 > λ > 320 nm solar radiation (UVA) to the total photochemical production of COS.
(a) Carbonyl sulfide (COS) concentration measured in 0.2‐μm‐filtered Indian seawater with and without 50 μM cysteine in quartz bottles. (b) COS concentration profiles in 0.2‐μm‐filtered Indian seawater added 50 μM cysteine in quartz bottles with different nitrate concentrations under sunlight conditions. (c) Comparison of the COS production rate in 0.2‐μm‐filtered Indian seawater added 50 μM cysteine in quartz bottles with different nitrate concentrations; (d) COS concentration measured in 0.2‐μm‐filtered Indian seawater added 50 μM cysteine in quartz bottles as well as added 50 μM cysteine and 500 μM nitrate with and without isopropyl alcohol as the •OH scavenger exposed to sunlight for 5 hr in quartz bottles; (e) the photochemical mechanism of COS in surface seawater. Error bars denote the standard deviation of measurements from multiple quartz bottles (n = 3).
Carbonyl sulfide (COS) plays an important role in the sulfur cycle and climate change. Yet, much remains unknown about the photochemical mechanisms of COS in nutrient‐rich seawater. We measured the photochemical production rates of COS in the surface seawater of the Indian Ocean under sunlight irradiation. The photochemical production of COS was mainly initiated by ultraviolet (UV) radiation with UVA contributing approximately 68% to the total COS production. Using cysteine, a typical proxy of dissolved organic sulfur, the effect of enhanced nitrate concentration on COS formation was conducted in authentic seawater during simulated sunlight irradiation, indicating the enhancement of the COS formation with increasing nitrate concentrations. This result revealed that the generation of hydroxyl radical with nitrate photolysis plays a key role in the COS formation process. These findings improve our understanding of the marine COS photoproduction cycle and the impact of nitrate on the COS photochemical production in surface seawater.
Distribution of intensity change (units: kt/24 hr) and ΔR34 (units: km/24 hr). Green and purple dots denote rapid growth (RG) and rapid shrinkage (RS) events identified by the isolation forest algorithm, respectively. Gray circles denote the events not identified as RG or RS. Dashed and solid lines denote the thresholds of the 95th (5th) and 90th (10th) percentiles. The histogram on the right shows the probability distribution of ΔR34 (units: %).
Distributions of (a) lifetime maximum size (LMS) and (b) lifetime maximum integral kinetic energy (LMIKE) for Atlantic hurricanes. IKE was computed using the wind profile derived using Holland et al. (2010). Gray histograms and black lines depict the distribution of all tropical cyclones (TCs). Orange, green, and purple lines show the probability distribution functions (PDFs) for the TCs that undergo rapid growth (RG) defined as the 90th (≥75 km/24 hr) and 95th (≥103 km/24 hr) percentiles, and TCs undergoing rapid intensification (RI; ≥30 k/24 hr), respectively. The PDFs were smoothed using kernel density estimation with a bandwidth of 0.2. Vertically dashed line denotes the 90th percentile of LMIKE (≥82 TJ).
Life cycle of tropical cyclone (TC) outer size and intensity. Composite of the evolution of (a) R34 and (b) Vmax for rapid growth of outer size (RG) and non‐RG TCs (c and d) are the same with (a and b), but for the RI and non‐RI TCs. The solid lines represent average values, while the shadows represent one standard deviation.
Composite maps of environmental factors. Composite fields of (a–c) 600‐hPa relative humidity (RH, units: %); (d–f) 850‐hPa radial inflow (units: m/s); and (g–i) 700‐hPa Eady growth rate (σ, units: day–1) for (a, d, and g) RG (b, e, and h) non‐RG and (c, f, and i) the difference between RG and non‐RG events. The data was obtained from the ERA5 reanalysis. The units for x‐ and y‐axis are degrees. Stippled areas in the right panel indicate that the differences are statistically significant at the 99% confidence level based on Student's t‐test. The two circles are placed at radii of 5° and 10° from the tropical cyclone center.
Plain Language Summary The destructiveness of tropical cyclones (TCs) is mainly caused by their strong winds. The maximum wind speed, namely TC intensity, and its evolution has been intensively studied, including rapid intensification (RI). However, the extent of strong winds and its growth remains little‐explored. We investigate, for the first time, the rapid growth (RG) of the gale‐force wind radius (R34) for Atlantic hurricanes. We define RG by the 90th percentile of R34 changes, which is equivalently at least 75 km/24 hr. This threshold is supported by an objective anomaly detection algorithm, the isolation forest. There are 88% of the TCs (15/17) with large lifetime maximum size (larger than 400 km) undergoing RG. Among the 11 TCs with high destructive potential (kinetic energy higher than 82TJ), nine show RG, while only five undergo RI. The TCs with RG also show a more discernible size life cycle than those without RG. Our analysis highlights the crucial role of the rapid growth of TC outer size in changing the TC overall destructive potential, which is found to be at least as important as the widely recognized TC rapid intensification.
The first panel corresponds to the Ap index (geomagnetic activity). The second and third panel correspond to the NO concentration between 10⁻³ hPa and 10⁻² hPa, averaged over the latitude band 60–70°N for both the Medium energy electrons (MEE) (blue line) and no‐MEE (green line) run, as well as their difference (black line). The fourth and fifth panel correspond to the zonal wind velocity between 10⁻¹ hPa and 10⁻² hPa, averaged over the latitude band 60–70°N for both the MEE (blue line) and no‐MEE (green line) run, as well as their difference (black line).
The upper plot corresponds to the zonal wind velocity for the noMEE run. The lower plot shows the zonal wind velocity for the Medium energy electrons (MEE) run. The red color represents positive velocities, associated with eastward winds, while blue represents negative velocities, associated with westward winds. The data is averaged over the latitude band 60–70°N with a daily resolution. The two colored lines represent the mesopause altitude (coldest altitude) for both runs, the blue line corresponds to the MEE run, and the green line to the noMEE run.
Upper plots (from left to right): Zonal wind velocity difference between the MEE and the noMEE run, and temperature difference between both runs. Bottom plots (from left to right): NO VMR relative difference, and O3 absolute difference. For all plots, the data is averaged over the latitude band 60–70°N with a daily resolution. The two colored lines represent the mesopause altitude (coldest altitude) for both runs, the blue line corresponds to the MEE run, and the green line to the noMEE run.
OH absolute difference, the data is averaged over the latitude band 60–70°N with a daily resolution. The two colored lines represent the mesopause altitude (coldest altitude) for both runs. The blue line corresponds to the Medium energy electrons (MEE) run, and the green line to the noMEE run.
Medium energy electron (MEE) (30–1,000 keV) precipitation enhances the production of nitric (NOx) and hydrogen oxides (HOx) throughout the mesosphere, which can destroy ozone (O3) in catalytic reactions. The dynamical effect of the direct mesospheric O3 reduction has long been an outstanding question, partly due to the concurrent feedback from the stratospheric O3 reduction. To overcome this challenge, the Whole Atmosphere Community Climate Model version 6 is applied in the specified dynamics mode for the year 2010, with and without MEE ionization rates. The results demonstrate that MEE ionization rates can modulate temperature, zonal wind and the residual circulation affecting NOx transport. The required fluxes of MEE to impose dynamical changes depend on the dynamical preconditions. During the Northern Hemispheric winter, even weak ionization rates can modulate the mesospheric signal of a sudden stratospheric warming event. The result provides a first step in a paradigm shift for the understanding of the MEE direct effect.
Plasma wave observations of Van Allen Probe A during 21–22 UT on 16 October 2017. (a) SYM‐H (blue) and AE (red) indices from 12 UT on 16 October 2017 to 12 UT on 17 October 2017 (b–d) Frequency‐time spectrograms of magnetic field spectral intensity, ellipticity, and wave normal angle obtained by the EMFISIS WFR. (e–g) Frequency‐time spectrograms of magnetic field spectral intensity, ellipticity, and wave normal angle obtained by the EMFISIS MAG. (h) Integral wave amplitude of MS waves (blue) and EMIC waves (red) (i and j) Scatterplots of the magnetic field spectral intensities of EMIC waves and MS waves during 21:15 UT–21:30 UT and the mean profile (thick black curves). In Figures 1b–1d, the black lines represent lower hybrid resonance frequency fLHR, 0.5 fLHR, and the equatorial proton gyrofrequency fcP, respectively. The white lines denote fcP in Figures 1e–1g.
Energetic electron differential flux observations by Van Allen Probes during the periods of 12:00–12:40 UT, 18:20–19:00 UT, and 21:00–21:50 UT on 16 October 2017. Differential fluxes of radiation belt energetic electrons measured by the MagEIS instrument ((a) 108 keV, (b) 235 keV, (c) 470 keV, (d) 597 keV, and (e) 749 keV) and by the REPT instrument ((f) 1.8 MeV and (g) 3.4 MeV) as a function of pitch angle and time. The black shading in Figure 2c represents the Interval C of 21:15–21:30 UT on 16 October 2017 when the concurrent EMIC waves and MS waves are observed. The black shading in Figures 2a and 2b represents the Interval A and B when the Van Allen Probe A passes the exact same L‐shell range as Interval C.
Two‐dimensional plots of bounce‐averaged diffusion coefficients (from left to right: <Dαα>, <DEE>, and |<DαE>|) as a function of equatorial pitch angle αeq and electron kinetic energy Ek for (a–c) Magnetosonic (MS) waves only, (d–f) Electromagnetic ion cyclotron (EMIC) waves only, (g–i) the sum of MS and EMIC waves only, and (j–l) MS and EMIC waves simultaneously.
The temporal evolution of simulated electron PSDs at selected energies under the impact of Electromagnetic ion cyclotron (EMIC) waves only, Magnetosonic (MS) waves only, and EMIC and MS waves simultaneously in 160 min. In Figures 4e–4h, the solid lines indicate the simulated results and the dashed lines display the initial electron PSDs. The dots in Figures 4e and 4f–4h represent the electron PSDs observed during periods of ∼18:32–18:43 UT (Interval B in Figure 2) and 21:15–21:30 UT on 16 October (Interval C in Figure 2) correspondingly.
Plain Language Summary Electromagnetic ion cyclotron (EMIC) waves and Magnetosonic (MS) waves are commonly observed in the Earth's magnetosphere and play important roles in energetic electron dynamics. Usually, EMIC waves and MS waves scatter electrons at different energies and pitch angles. The peak frequency of the H⁺ band EMIC wave is usually much lower than the equatorial proton gyrofrequency. Recently, a different type of H⁺ band EMIC waves, named unusual high‐frequency EMIC waves since their peak frequency is close to the equatorial proton gyrofrequency, has been reported. Studies on the unusual high‐frequency EMIC waves found that these waves are capable of scattering sub‐MeV and MeV electrons that can also be influenced by MS waves. Moreover, observations confirm that unusual high‐frequency EMIC waves can be well connected with MS waves. In the present study, we quantitatively investigate the combined scattering effect of both wave modes on radiation belt electrons and simulate the evolution of the electron phase space densities under the impact of both waves. The simulation results show similar evolution trends as the observations, indicating the importance of incorporating these two waves and evaluating their combined effects on radiation belt particle dynamics.
Maps of stations (a) and events (b) used in this study. (a) Red triangles denote SEISConn stations. The black dashed line A‐A’ is the projected profile line. Colored lines show the surface boundaries of the eastern edge of Laurentia (blue), the Mesozoic Hartford rift basin (yellow), and the western edge of Avalonia (red). U.S. state names are shown with abbreviations (CT, Connecticut; MA, Massachusetts; NY, New York; RI, Rhode Island). (b) Red triangle represents the center of the SEISConn array. Red stars denote earthquakes used to construct migration images, within the epicentral distance limits of 30° and 90° (black dashed circles). Blue filled stars mark the three events used to construct Figure S5a in Supporting Information S1.
Scattered wavefield migration images of S‐wave velocity perturbations δββ $\left(\frac{\boldsymbol{\delta }\boldsymbol{\beta }}{\boldsymbol{\beta }}\right)$ constructed using different phases. Lighter regions have positive velocity perturbations and darker regions have negative velocity perturbations. The horizontal line visible in each image at 35 km depth is an artifact from the 1‐D background velocity model (see Text S1 in Supporting Information S1). (a) Migration image generated using forward‐scattered Ps phase, band‐pass filtered between 0.2 and 2.0 Hz. (b) Migration image generated using backscattered phases (PPs, PSp, PSsv, PSsh), band‐pass filtered between 0.03 and 1.0 Hz. (c) Composite migration image generated using both forward‐scattered and backscattered phases, band‐pass filtered between 0.03 and 1.0 Hz. (d) Uncertainty in the composite migration image derived from the bootstrap resampling. Lighter regions have smaller uncertainty and darker regions have larger uncertainty.
(a) Annotated version of Figure 2a, the migration image made with only Ps phase band‐passed filtered between 0.2 and 2.0 Hz. Markers denote corresponding features observed in single‐station stacked receiver functions band‐passed between 0.2 and 2.0 Hz (Luo et al., 2021). (b) Annotated version of Figure 2c, the composite migration image made with all available phases, band‐passed filtered between 0.03 and 1.0 Hz. (c) Schematic diagram showing one possible model to explain observed features. Solid black lines are hypothesized crustal boundaries, not constrained by the migration. NAA—Northern Appalachian Anomaly. Plotting conventions for migration results are as in Figure 2. Colored dashed lines highlight the Moho discontinuity (yellow), a west‐dipping elongate feature in the mantle lithosphere (blue), and a strong low‐velocity anomaly beneath the eastern portion of the array (red).
Estimates of the relative S‐wave velocity contrast (unitless) across the Moho discontinuity. The black solid line and shaded region show the result and uncertainty estimate from bootstrap resampling. Colored dashed lines show results from individual migration images based on the same set of 56 events but with different frequency contents, as shown by the legend.
Plain Language Summary Tectonic processes in the geologic past, such as the formation and breakup of supercontinents, modified the deep structures of the crust and upper mantle beneath eastern North America. In this study, we use a seismic imaging technique based on scattered wavefield back‐projection to investigate deep structures beneath southern New England. This imaging technique, which relies on seismic wave energy from distant earthquakes, is capable of resolving km‐scale structures when applied to data from closely spaced seismometers (∼10 km station spacing). We image an abrupt, step‐like change of the crustal thickness beneath southern New England; the details of this feature suggest a complicated tectonic history during the formation of the Appalachian Mountains. A west‐dipping interface in the upper mantle suggests the presence of a relict slab beneath southern New England, associated with a past subduction event. A region of low seismic velocity in the upper mantle beneath southeastern New England may reflect past impingement of a mantle plume or modern upwelling of asthenospheric mantle.
Probability (a) of occurrence of compound snow drought and heatwave (CSDHW) events and Arid Index map (b) showing arid, transitional, and humid regions across the world. Statistically significant grid boxes at 95% confidence level are dotted on maps.
The fractions (a) and spatial coverage trends (b–q) of compound snow drought and heatwave (CSDHW) events across the world during 1981–2020. The red and blue stack areas represent the dry‐ and warm‐type CSDHW events, respectively, the sum of which represents all CSDHW events. The linear annual trends in dry‐type, warm‐type, and all CSDHW events are represented in dashed, dotted and solid lines, respectively. The number on the left indicates the estimated linear slope based on the least‐squares method, and the asterisk denotes the statistically significant trend (p < 0.05) based on the Mann‐Kendall test. GLB, Globe; NEU, North Europe; CEU, Central Europe; MED, South Europe/Mediterranean; NAS, North Asia; EAS, East Asia; SAU, South Australia/New Zealand; TIB, Tibetan Plateau; CAS, Central Asia; WAS, West Asia; WSA, West Coast South America; ENA, East North America; CNA, Central North America; WNA, West North America; ALA, Alaska/N.W. Canada; CGI, Canada/Greenland/Iceland.
Heatwave probability (HWP, left column) and heatwave severity (HWS, right column) in dry‐ and warm‐type compound snow drought and heatwave: HWP distribution under the type of dry (a) and warm (c) snow droughts; HWS distribution under the type of dry (b) and warm (d) snow droughts; HWP (e) and HWS (f) statistics under dry (red) and warm (blue) snow droughts. Statistically significant grid boxes at 95% confidence level are dotted on maps (a and c). The limits of statistics (e and f) represent the upper and lower quartiles and the circles represent the mean values. Statistics labeled with different letters (“A” and “B”) indicate a significant difference, and letter “B” represents a greater mean value than letter “A.” Two statistics that are both labeled with “A” have no obvious difference.
Soil moisture (SM) anomaly (a and b) and vapor pressure deficit (VPD) anomaly (c and d) as well as their correlations with heatwave severity (e and f) in the first snow‐free month under dry (left column) and warm (right column) snow drought conditions. Asterisks in a–d denote that the SM (VPD) of snow‐drought grid boxes is significant (p < 0.05) lower (higher) than those of all grid boxes based on the two‐sample t‐test. Asterisks in e and f represent that the correlation is statistically significant (p < 0.05). Linear trends have been removed from all the variables.
The compound of late winter snow droughts and early spring heatwaves (CSDHW) could dramatically affect ecosystems and water availability, but has not been systematically investigated. Here we present a comprehensive assessment of CSDHW events and possible driving mechanisms. We find that 7% of the snow-covered area experiences significant (p < 0.05) CSDHW events, and an average of 35% of snow droughts are followed by heatwaves during 1981–2020. The spatial extent of CSDHW is asymmetrically enlarging, with a significant increase in Eurasia and a relatively high fluctuation in North America. Specifically, the warm-type CSDHW (i.e., snow drought with normal or above-average precipitation followed by heatwave) occurs more frequently, with spatial coverage increasing faster than the dry-type CSDHW (i.e., snow drought with below-average precipitation followed by heatwave). In comparison, dry snow drought is more likely to be followed by heatwave due to intensified soil drought and atmospheric aridity.
Time series of P and T measured at specific depths in Tenmile geyser in (a) 2011, (b) 2013, (c) 2014, and (d) 2015; measurement data in 2011 has a lower resolution than others. In 2013, 2014, and 2015, time series of Pair and Tair were also measured. The gray box, including one natural eruption and four artificially induced eruptions, is enlarged in Figure 2b.
(a) Conceptual diagram delineating one cycle of eruptive processes. The length of each arrow represents the magnitude of the corresponding flow qualitatively. (b) Measurement data from the gray box in Figure 1d, including one naturally driven and four artificially induced geyser eruptions. (c) Changes in P, T, and v (upward flow as positive value) during the naturally driven geyser. (d) Changes in P and T during four artificially induced eruptions; the gray boxes represent durations of pumping. During the last two eruptions (16:30–17:30), an action camera (GoPro) was set at a depth of approximately 30 m to observe the evolution of CO2 bubbles in the well (Figure S5 in Supporting Information S1 and Movie S1).
Relationship between atmospheric conditions (Pair and Tair) and lengths of IBEs (LIBE). (a) Two quantile density contours based on Pair and Tair from the whole observational data set are plotted in a 1‐day cycle with LIBE (109 points). The polynomial fitting curve based on middle time in intervals between two consecutive eruptions (IBEs) is delineated by a black line with a 95% confidence interval (two dashed lines; the statistics are summarized in Table S1 of Supporting Information S1). (b) Two transducers were installed in different depths with Δh. (c) Double box based on Pair and Tair measured during each IBE; the colors of boxes indicate the φIBE/Δt values. The vertical and lateral boundaries for boxes represent the 25th and 75th percentiles of Pair and Tair during each IBE, respectively.
CO2 generation from time series of ΔP' and electrical conductivity (EC) and T profiles. (a) Time series of P measured at depths of 6.6 and 17.8 m and their differences (ΔP'). (b–d) EC and T profiles were measured at times indicated in (a) Profiles before (13:34 and 17:35) and after (14:07 and 18:11) eruptions are plotted in b and d; four profiles (14:07, 15:33, 16:33, and 17:35) measured within the intervals between two consecutive eruption are plotted in (c) The grayness of boxes in a and lines in b–d are intended to convey proximity to the next eruption. Other ΔP' and temperature, level, and conductivity data are shown in Figures S7 and S8 of Supporting Information S1, respectively.
CO2‐driven cold‐water geysers periodically ejecting cold water are rare. Although coalescence and expansion of ascending CO2 bubbles can explain the eruption process, the triggering conditions and eruption cycle remain unclear. To clarify the triggering conditions, hydrostatic pressure in the well was decreased by pumping to induce eruptions. All four pumping tests successfully induced eruptions by decreasing the pressure of ∼10⁴ Pa. In the absence of artificial perturbations, similar reductions in pressure were observed during the intervals between two consecutive eruptions (IBEs). During IBE, the atmospheric pressure (Pair) and temperature (Tair) controlled the generation of the CO2 bubbles which directly induced the pressure reduction in the well. Especially under the persistent low Pair and high Tair, the length of IBE showed a minimum value of 3.90 hr during field observations. We suggest that the atmospheric perturbations are the causes of the changes in geyser periodicity, given consistent geological and hydraulic conditions.
Observation of the DL structure and E|| turbulence by the MMS3 satellite on May 5, 2018. (a, b) The MMS locations in the X‐Z and X‐Y planes in GSE coordinates. (c) Magnetic field in GSE coordinates. (d) Parallel electric field in burst mode (black) and electron density (blue). (e) Current density in field‐aligned coordinates. (f) Differential energy fluxes of electrons and parallel electron temperature (white). (g)–(i) Electron 2D VDF for three different time durations, with the black lines denoting the contours of the phase space density. Here, the velocity plane is defined as (VB,VE×B ${\mathbf{V}}_{\mathbf{B}},\,{\mathbf{V}}_{\mathbf{E}\times \mathbf{B}}$).
Acceleration of beam electrons. (a) Electron 1D VDF in the parallel direction, derived from the electrons with α < 45°. The white dotted line denotes the drift velocity of the electron beam. (b) E|| measured by EDP, the highest resolution HMFE data (65 kHz sampling, red) are plotted on top of the burst data (8 kHz sampling, black). (c) E⊥ in burst mode (8 kHz sampling). (d) Potential (black) and changes in the averaged parallel energy (blue). (e)–(h) Four snapshots of the electron 1D VDF in S1–S4 for the time durations denoted in (a).
Thermalization of beam electrons. (a) The power spectrum density of HMFE E||; the two black dotted lines denote 0.25 kHz, and 4 kHz. (b) Filtered HMFE E|| waveform with a bandwidth of 0.25–4 kHz. (c) RMS value of the electric fields in the frequency ranges of 0.25–4 kHz (black) and 4–32 kHz (blue). (d) The equivalent parallel temperature of beam electrons (black) and the Omni parallel electron temperature (blue).
Plain Language Summary Electron acceleration and thermalization in the plasma sheet (PS) of the Earth's magnetotail are fundamental research topics of magnetospheric physics. Theoretical analyses and numerical simulations have revealed that beam electrons can be accelerated and thermalized by the structure of double layer (DL). Direct observation of PS electron acceleration and thermalization is essential to demonstrate the theoretical prediction. Due to the low time and energy resolutions of observing electrons on previous satellites, it was very difficult to display the detailed evolutionary processes of electron acceleration and thermalization by DL structures. Using 3D electron phase‐space distributions with a time resolution of 30 ms and electric field data in burst mode by the magnetospheric multiscale satellites, for the first time, we provide a complete and direct observation of the detailed evolution of the acceleration and thermalization of magnetotail beam electrons by a DL structure, which indicates the energy exchange process between nonlinear electric field structures and electrons in the Earth's PS.
Plain Language Summary Eddy kinetic energy has been increasing in the Southern Ocean over the past few decades. These changes in the eddy field are of great importance because they play a crucial role in modulating the ocean circulation response to surface forcing. However, the EKE changes are inhomogeneous in the Southern Ocean. To understand the pattern of these changes, we analyze the satellite altimeter record over a period of 28 years since 1993 and carry out a set of idealized simulations. We find that the change of EKE is more related to the mean-flow than to localized wind changes. The increasing wind stress contributes to increasing EKE by intensifying the circumpolar mean flow, with local wind stress playing a minor role in the pattern of EKE changes. Strong EKE variations are generally confined downstream of major topographic features, suggesting strong modulation by topography. This study indicates the change of EKE depends on the combination of wind stress, mean-flow and topography in the Southern Ocean.
Phase‐averaged intra‐tidal variabilities of (a) current speed profile, (b) dissipation rate and buoyancy flux, (c) flux Richardson number RfII ${R}_{f}^{\mathit{II}}$ and gradient Richardson number, (d) buoyancy Reynolds number, and (e) turbulent Froude number. Error bars indicate the standard deviation of the data in each bin. Note that data for late ebb (after 7 Hour Relative to High Water) was omitted given low water level and flow velocity.
RfII ${R}_{f}^{\mathit{II}}$ as a function of Reb (after Monismith et al., 2018, incorporating additional datasets). The closed dots represent mean values of RfII ${R}_{f}^{\mathit{II}}$ binned by Reb with error bars indicating standard deviation of the data in each bin. Data from other estuaries are also shown (Holleman et al., 2016; MacDonald & Geyer, 2004). Other observational data are covariance and microstructure profile data (Monismith et al., 2018; blue lines), as well as data from quiescent sites and energetic sites in Mediterranean Sea (Vladoiu et al., 2021; blue symbols). Black lines are parameterizations based on data from shelf seas (Bluteau et al., 2013), lakes (Bouffard & Boegman, 2013) and atmospheric boundary layer (Lozovatsky & Fernando, 2013). Black symbols are DNS data from Shih et al. (2005), Chung and Matheou (2012), Zhou et al. (2017), Salehipour and Peltier (2015), Brethouwer et al. (2007), and Arthur et al. (2017). Results from laboratory experiments in green color represent homogeneous, shear (Rohr & Van Atta, 1987) and grid stirring, unsheared turbulence.
RfII ${R}_{f}^{\mathit{II}}$as a function of Ri for covariance data of Monismith et al. (2018) and estuarine data of Holleman et al. (2016) and present study. The closed dots represent mean values of RfII ${R}_{f}^{\mathit{II}}$ binned by Ri with error bars indicating standard deviation of the data in each bin. Curves outside the main graph are PDFs of RfII ${R}_{f}^{\mathit{II}}$ and Ri for these three datasets.
RfII ${R}_{f}^{\mathit{II}}$ as a function of Frt. The closed red dots represent mean values of RfII ${R}_{f}^{\mathit{II}}$ binned by Frt with error bars indicating standard deviation of the data in each bin. Observational data from deep ocean by Ijichi et al. (2020) are shown in the form of binned averages and standard deviations in blue colored symbols. Green triangles represent observations from Fraser river estuary by MacDonald and Geyer (2004). Different DNS datasets of Shih et al. (2005), Maffioli et al. (2016), and GV19 are presented as asterisks, diamonds, and triangles, respectively. Black solid lines represent (6). The best fit line based on laboratory data by G. Ivey and Imberger (1991) is indicated by black dashed line. Curves outside the main graph are PDFs of Rf and Frt for Ijichi et al. (2020, blue) and the present results (red).
The flux Richardson number Rf, also called the mixing efficiency of stratified turbulence, is important in determining geophysical flow phenomena such as ocean circulation and air-sea transports. Measuring Rf in the field is usually difficult, thus parameterization of Rf based on readily observed properties is essential. Here, estimates of Rf in a strongly turbulent, sediment-stratified estuarine flow are obtained from measurements of covariance-derived turbulent buoyancy fluxes (B) and spectrally fitted values of the dissipation rate of turbulent kinetic energy (ε). We test scalings for Rf in terms of the buoyancy Reynolds number (Reb), the gradient Richardson number (Ri), and turbulent Froude number (Frt). Neither the Reb-based nor the Ri-based scheme is able to describe the observed variations in Rf, but the Frt-based parameterization works well. These findings support further use of the Frt-based parameterization in turbulent oceanic and estuarine environments.
Large-scale CO2 sequestration into geological formations has been suggested to reduce CO2 emissions from industrial activities. However, much like enhanced geothermal stimulation and wastewater injection, CO2 sequestration has a potential to induce earthquake along weak faults, which can be considered a negative impact on safety and public opinion. This study shows the physical mechanisms of potential seismic hazards along basement faults driven by CO2 sequestration under variation in geological and operational constraints. Specifically we compare the poroelastic behaviors between multiphase flow and single-phase flow cases, highlighting specific needs of evaluating induced seismicity associated with CO2 sequestration. In contrast to single-phase injection scenario, slower migration of the CO2 plume than pressure pulse may delay accumulation of pressure and stress along basement faults that may not be mitigated immediately by shut-in of injection. The impact of multiphase flow system, therefore, needs to be considered for proper monitoring and mitigation strategies.
Location of stream water and precipitation sampling as well as the SNOTEL stations in
the study area. Name, boundaries and elevation distribution of all nine catchments shown. Upper
left insert shows the Upper Colorado Basin (black line) and the East River (star) within the western
US. Longitude and latitude are provided at the ticks of the map frame.
(a) Relative importance of a multiple linear regression (MLR) model to explain
variability of endmember mixing and splitting results. Bold grey numbers indicate regression
parameters with p-values < 0.1. The coefficient of determination (r2) of the regression models
explaining the variation of the endmember mixing and splitting results across the catchments are
shown in grey above each column. The individual relationships for significant parameters of high
relative importance are shown in inserts (b to g) and Pearson’s correlation coefficient and pvalues
are provided in the legends for each year and the long-term analyses 2015-2020. Error
bars represent standard errors based on Gaussian error propagation.
Endmember mixing (a – c) and endmember splitting (d – h) results as a function of
hydrometric data: annual sums of evapotranspiration (ET), snowfall (PS), rainfall (PR), and nonsummer
streamflow (QnS). Shown are results from individual catchments and years (circles and
color coded, see legend in subplot d) as well as median values across all catchments for individual
years (diamonds). Correlation coefficients (r) and p-values are shown in brackets in the legends,
while bold font indicates significant correlations (p<0.05). Circles represent variability across
catchments for individual years and diamonds represent variability between different years
averaged over the catchments. Error bars represent standard errors based on Gaussian error propagation. Please see Suppl. Fig. 15 for the same plots with flux volumes on the y-axis.
Understanding the partitioning of snow and rain contributing to either catchment streamflow or evapotranspiration (ET) is of critical relevance for water management in response to climate change. To investigate this partitioning, we use endmember splitting and mixing analyses based on stable isotope (18 O) data from nine headwater catchments in the East River, Colorado. Our results show that one third of the snow partitions to ET and 13% of the snowmelt sustains summer streamflow. Only 8% of the rainfall contributes to the summer streamflow, because most of the rain (67%) partitions to ET. The spatial variability of precipitation partitioning is mainly driven by aspect and tree cover across the sub-catchments. Catchments with higher tree cover have a higher share of snow becoming ET, resulting in less snow in summer streamflow. Summer streamflow did not contain more rain with higher rainfall sums, but more rain was taken up in ET.
(a) Location of the study site in the northwestern U.S. (b) Floodplain topography and channel bathymetry mapped by the Experimental Advanced Airborne Research Lidar instrument. Black dots are individual LiDAR elevation measurement points. (c) Spatial pattern of flow velocity in the mainstem channel during a discharge of 1 m³/s. (d) Distribution of Chinook salmon spawning habitat quality during a discharge of 1 m³/s. See equation 3 in Supporting Information S1 for calculation of habitat quality. (e) Off‐channel rearing habitat (OCH) as a function of water stages in 1957 and 2090. The two circled areas illustrate examples where water flows freely through the OCH at the higher stage (1957 conditions), but the habitat becomes fragmented and characterized by standing water at the lower stage (2090 conditions).
Trends estimated from linear regressions for the historic period (1957–2021, blue lines with pink fill for standard error) and over the full study extent (1957–2090, black lines with gray fill for standard error) in yearly values of average summer reach conditions for Bear Valley Creek under the ensemble mean warming scenario. Reach characteristics are: (a) Discharge and flow velocity, (b) wetted area and depth, (c) off‐channel habitat area, and (d) weighted useable area for adult spawning and juvenile rearing. Trends over the full study extent are statistically significant (p < 0.001), Q = −0.01 year + 21.74, V = −0.001 year + 2.37, A = −88.91 year + 284,539, D = −0.00055 year + 1.58, OCH = 71.5 year + 172,771.7, S = −130.78 year + 302,129, and R = −13.45 year + 99122.3 (year expressed as calendar year, e.g., 1957), whereas those for the historic period are not statistically significant. Shaded areas around the linear trends identify the 95% confidence level intervals.
Change in summer off‐channel rearing habitat, off‐channel habitats (OCH), size frequency distributions between historical, 1957, and future, 2040 and 2090, periods when the mean water stage height is predicted to decline by 10 cm. Whiskers represents OCH variability due to 24% discharge uncertainty quantified for future predictions. The symmetrical uncertainty of the discharge results in unsymmetric whiskers around the expected OCH value because OCH extension is due to the interaction between discharge stage and off‐channel topography.
Climate change threatens biodiversity through global alteration of habitats, but efficient conservation responses are often hindered by imprecise downscaling of impacts. Besides thermal effects, warming also drives important ancillary environmental changes, such as when river hydrology evolves in response to climate forcing. Earlier snowmelt runoff and summer flow declines are broadly manifested in snow‐dependent regions and relevant to socioeconomically important cold‐water fishes. Here, we mechanistically quantify how climate‐induced summer flow declines during historical and future periods cause complex local changes in Chinook salmon (Oncorhynchus tshawytscha) habitats for juveniles and spawning adults. Changes consisted of large reductions in useable habitat area and connectivity between the main channel and adjacent off‐channel habitats. These reductions decrease the capacity of freshwater habitats to support historical salmon abundances and could pose risks to population persistence in some areas.
Schematic diagram of the instrumentation setup. All results reported in this research were obtained at the Ultra‐High Voltage Laboratory affiliated with the State Grid Corporation of China in Hefei, China.
(a) High‐speed camera frames recording the typical positive leader discharge. These fames were color‐inversed and enhanced to improve visualization. These fames were captured by a Photron FASTCAM SA‐Z camera. (b) The current waveform of the typical positive leader discharge.
Details of the current pulse corresponding to the step 1st, 2nd, and 3rd. The actual peak of the three current pulses are 22.85, 1.9, and 12.55 A respectively.
High‐speed camera frames recording the step 1st, 2nd, and 3rd. These frames (without enhancement) are pseudo‐color processed. The original grayscale frames are with 12‐bit pixel depth. The size of these frames is 128 pixels (in height) × 56 pixels (in width). The color of pixels shown in these frames represents the grayscale value of these pixels. The deepest purple corresponds to the minimum grayscale value and the deepest red corresponds to the maximum grayscale value (about 526).
(a) Statistics of the number of bell type steps in a single discharge event. (b) Statistics of the number of steep‐rise type steps in a single discharge event. When the charge voltage per stage of the Marx generator was set as 70, 80, 90, 105, 120, and 150 kV, the number of 22, 24, 20, 24, 22, and 22 positive leader discharges were produces and observed, respectively. The statistics are for all these events.
Plain Language Summary Lightning is one of the most impressive, commonly experienced geophysical phenomena, driven by the propagation of positively and negatively charged, thermally‐ionized plasma channels, known as positive and negative leaders. Leaders propagate in a continuous or stepwise manner. Knowledge of leader steps is critical to understanding the radio‐frequency electromagnetic radiation and even the high energy physical phenomenon. Based on natural or triggering lightning observations, laboratory‐scale discharge experiments, and numerical simulations, some important insights (like the spontaneous emergence of space stems and the bi‐directional development of the space leaders) about negative leader steps have been given by researchers. Positive leader steps are more mysterious. We investigate positive leader steps by performing laboratory spark discharge experiments. Using a high‐speed video camera and a synchronized electrical parameter measurement system, two types of steps were distinguished in the positive leader continuous development process based on the current pulse features. One type of step exhibits a bell‐shaped current, while the other type of step exhibits a steep current rise. The former type vanishes when the rising rate of the ambient electric field increases, whereas the latter type, which could be led by a floating luminous formation, does not.
(a) Eastern Himalayan Syntaxis and eastern margin of the Indian plate (modified from Robinson et al., 2014). The Burma microplate—colored green—is rimmed by the Incertus‐arc and Indus‐Yarlung sutures; it is partly covered by the Central Myanmar basin. The Incertus‐arc suture was south (in the Cretaceous) of the Incertus arc, defined by Westerweel et al. (2019); this arc was part of the Trans‐Tethyan subduction system of Hall (2012). Insert locates (a) and shows Eurasia‐fixed GNSS‐derived displacement field. (b) Geological map centered on the Katha Range modified from Geological Map of Myanmar (2014) and Wang and Burchfiel (1997). Sagaing transform‐fault system modified from Morley and Arboit (2019) and Maurin et al. (2010). Yellow bars: studied traverses and samples (see Text S1 in Supporting Information S1 and Table S1 for detailed sample location). EHS, Eastern Himalayan Syntaxis; IYS, Indus‐Yarlung Suture; IS, Incertus‐arc suture (Jurassic‐Cretaceous ophiolite belt); JB, Jadeite Belt; KAR, Katha Range; KR, Kumon Range; MMB, Mogok Metamorphic Belt; TMB, Tagaung‐Myitkyina Belt.
Cumulative probability plots of U‐Pb zircon and rutile ages of (a) samples from this study and sample 14M76 of Zhang et al. (2018), and (b) their comparison with rocks from central and eastern S‐Tibet and the central and eastern Himalaya. Ages used include 2s uncertainties and have 90–110% ²⁰⁶Pb/²³⁸U–²⁰⁷Pb/²⁰⁶Pb age concordance. THS, Tethyan Himalaya Series; GHS, Greater Himalaya Series; LHS, Lesser Himalaya Series; IYS, Indus‐Yarlung Suture.
Pressure‐temperature‐time‐deformation (P‐T‐t‐d) data. (a) P‐T of the Katha rocks and comparison with data from central and eastern S‐Tibet and the eastern Himalaya. Our new prograde and peak data are 470–510°C, 1.0–1.5 GPa, reached at >65‒45 Ma, and 490–551°C, 0.8–1.0 GPa, reached at ∼45 Ma, respectively. The data from the Greater Himalaya Series (GHS) and Lesser Himalaya Series (LHS) are from Bhutan (Daniel et al., 2003). (b) T‐t paths, and (c) structural data of the Katha rocks; see Figure 1b for traverses studied and Text S1 in Supporting Information S1 for detailed location of samples. THS, Tethyan Himalaya Series.
The Katha Range in the evolution of the Eastern Himalayan Syntaxis (EHS) and the Sagaing transform‐fault system (SF). The nomenclature “Incertus arc” follows Westerweel et al. (2019), and describes the island‐arc system of the Trans‐Tethyan subduction system of Hall (2012). (a) Incipient Himalaya‐Tibet formation following Incertus‐arc subduction with the Burma microplate at the arc's eastern end. (b) Development of the SF system along the Indus‐Yarlung suture (IYS) and its connection with the Tethyan Himalaya Series (THS) fold‐thrust belt. (c) Major fault systems of the EHS. (d) Restoration of the imbrication of the Incertus‐arc subduction system at the western margin of the Burma microplate. Growth of the SF system imbricated the northern part of the Burma microplate, isolated the Jadeite belt and the northernmost part of the Jurassic ophiolite belt, and imbricated the Indian rocks of the Katha and Kumon Ranges. A to D demark major strands of the SF system and do not imply a time sequence of formation. Abbreviations: IYS, Indus‐Yarlung Suture; THS, Tethyan Himalaya Series; TMB, Tagaung‐Myitkyina Belt.
In the Katha Range of central Myanmar, lithologic tracers and pressure‐temperature‐deformation‐time data identify Cambro‐Ordovician, Indian‐affinity Tethyan Himalaya Series, located ∼700 km from their easternmost outcrop in S‐Tibet, and ∼450 km from Himalayan rocks in the Eastern Himalayan Syntaxis. Metamorphism began at ∼65 Ma, peaked at ∼45 Ma (∼510°C, 0.93 GPa), and exhumation/cooling (∼25°C/Myr) occurred until ∼30 Ma in a subduction‐early collision tectonic setting. When the Burma microplate—part of the intra‐Tethyan Incertus arc—accreted to SE‐Asia, its eastern boundary, the southern continuation of the Indus‐Yarlung suture (IYS), was reactivated as the Sagaing fault (SF), which propagated northward into Indian rocks. In the Katha rocks, this strike‐slip stage is marked by ∼4°C/Myr exhumation/cooling. Restoring the SF system defines a continental collision‐oceanic subduction transition junction, where the IYS bifurcates into the SF at the eastern edge of the Burma microplate and the Jurassic ophiolite‐Jadeite belts that include the Incertus‐arc suture.
(a) Changes of the 3‐dimensional volume‐weighted spatial standard deviation (SSD) of global ocean temperature (SSDT; unit in °C) for 0–2000 m derived from Argo, IAP, Ishii, EN4.2.0, WOA18, SODA2.2.4, and ORAS4. The black thick curve and the gray shading denote the ensemble‐mean and one standard deviation range of the 7 datasets, respectively. The inset shows evolutions for the full‐depth SSDT (from the surface to bottom) from EN4.2.0, SODA2.2.4, and ORAS4. Here SSDT is shown as a 2‐year low‐pass filtered anomaly relative to the 1960–1980 average baseline (anomaly of Argo is relative to the 1960–1980 average of IAP). The red triangles denote major volcanic eruptions. (b) Percent change of 0–2000 m SSDT in 1960–2010 relative to the 1960–1980 average value. The error bars denote the 95% confidence interval. (c), (d) and (e), (f) are the same as (a), (b), but for the salinity SSD (SSDS; unit in psu) and the thermohaline inhomogeneity (THI) index (kg m⁻³), respectively.
Evolutions of SSDT (a), SSDS (b), and THI index (c) of the 0–2000 m global ocean derived from CMIP6 models (37 models for historical simulations of 1850–2014 and 13 models for SSP2‐4.5 projections of 2015–2100), shown as the anomalies relative to the 1960–1980 baseline. The multi‐model mean (MMM) is plotted as a thick curve (black for 1850–2014 and red for 2015–2100), and their one standard deviation ranges are plotted as the shading. Thick blue curves denote the ensemble mean of observational and reanalysis datasets from Figure 1. The red triangles in a denote the major volcanic eruptions. The inset compares the 1960–2014 linear trend from observations (blue), the 1960–2014 linear trend from CMIP6 historical MMM (black), and the 2015–2100 linear trend from CMIP6 SSP2‐4.5 MMM (red), respectively, with the error bars showing 95% confidence intervals. (d) The inter‐model relationship between global mean sea‐surface temperature (GMSST) trend and 0–2000 m SSDT trend during 1960–2014. The correlation coefficient R with its p‐value and the linear fit (blue solid line) are shown. (e) As in (d), but for the inter‐model relationship between the surface salinity contrast (SSC) trend and 0–2000 m SSDS trend during 1960–2014. The model names are listed in corresponding colors.
(a) Horizontal distributions of SSDT,XY trend during 1960–2019. (b) As in (a), but for 0–2000 m average temperature trend. Gray contours in a and b show the climatological temperature deviation δT (°C) from the global mean temperature for 1960–2019. Stippling indicates the insignificant trends at 95%. (c) Vertical distributions of SSDT,Z trend during 1960–2019. (d) As in (c), but for horizontally averaged temperature trend. The shadings denote the 95% confidence interval. The blue curves in c and d denote the climatological global‐mean δT profile. All results are based on IAP data.
Same as Figure 3, but for horizontal distributions of SSDS,XY trend (a) and 0–2000 m average salinity trend (b), and vertical distributions of SSDS,Z trend (c) and horizontally averaged salinity trend (d). Grey contours in a and b show the climatological salinity deviation δS (psu) from the global mean salinity. The blue curves in c and d denote the climatological global‐mean δS profile.
The ocean is inhomogeneous in hydrographic properties with diverse water masses. Yet, how this inhomogeneity has evolved in a rapidly changing climate has not been investigated. Using multiple observational and reanalysis datasets, we show that the spatial standard deviation (SSD) of the global ocean has increased by 1.4 ± 0.1% in temperature and 1.5 ± 0.1% in salinity since 1960. A newly defined thermohaline inhomogeneity index, a holistic measure of both temperature and salinity changes, has increased by 2.4 ± 0.1%. Climate model simulations suggest that the observed ocean inhomogeneity increase is dominated by anthropogenic forcing and projected to accelerate by 200%–300% during 2015–2100. Geographically, the rapid upper‐ocean warming at mid‐to‐low latitudes dominates the temperature inhomogeneity increase, while the increasing salinity inhomogeneity is mainly due to the amplified salinity contrast between the subtropical and subpolar latitudes.
(a) Drought conditions identified by the United States Drought Monitor (USDM) for mid‐June 2021. (b) The National Centers for Environmental Prediction (NCEP) 2 m air temperature composite anomaly for the period from 13 June 2021 to 19 June 2021 over the region of interest. The light green outline shows the USDM D4 (Exceptional Drought) region as identified in Figure 1a.
The observed difference between 2021 and 2019 time averaged for the period from June 13 to June 19 over the region of interest for: (a) MODIS LAI (%), (b) SMAP root zone soil moisture (m/m), (c) MODIS surface albedo, (d) MODIS SSEBop evapotranspiration (W/m²), and (e) MODIS IGBP land cover classification.
Time series of selected variables for the 2021 and 2019 initialization experiments spatially averaged over the region of interest during the month of June.
The simulated difference between the 2021 and 2019 initialization experiments, time averaged and ensemble averaged for the period from June 13th to June 19th over the region of interest for: (a) Leaf Area Index (%), (b) Root zone soil moisture (m/m), (c) Surface albedo, (d) Surface downwelling solar radiation (W/m²), (e) surface upwelling longwave radiation (W/m²), (f) Net radiation (W/m²), (g) latent heat flux (W/m²), and (h) Sensible heat flux (W/m²).
The simulated difference between the 2021 and 2019 initialization experiments time averaged for the period from 13 June to 19 June for: (a) Total precipitation (mm) and (b) Vapor pressure deficit (Pa). The light green outline shows the United States Drought Monitor (USDM) D4 (Exceptional Drought) region as identified in Figure 1a.
In June of 2021 the southwest United States experienced a record-breaking heatwave. This heatwave came at a time when the region was in severe drought. As drought alters the surface energy budget in ways that affect lower atmosphere temperature and circulations, it is possible that the combined drought-heat event was a cascading climate hazard, in which preexisting drought exacerbated the heatwave. We apply satellite observation and numerical experiments with the Weather Research and Forecasting (WRF) model to test for land-atmosphere feedbacks during the heatwave consistent with drought influence. We find a modest positive drought-heat effect, as WRF simulations that include the drought have marginally higher air temperatures than those that exclude the initial drought conditions, with more substantial effects in wetter, forested areas. Evidence of drought-heat-drought coupled feedbacks was similarly modest in our simulations, as accounting for drought preconditioning led to a small reduction in simulated precipitation in the region.
(a) Porosity at the end of model run at about 9.5 Myrs along with location of the vertical profile taken for the plot and (b) time progression of vertical porosity profile taken at 50 km from the ridge axis for model with half spreading rate, U0 = 3 cm/yr, and intrinsic permeability, K0 = 4 × 10⁻⁶.
Estimate of crustal production and corresponding power spectra for models with intrinsic permeability K0 = 4 × 10⁻⁶ and varying half spreading rate, U0. The vertical dashed line in the right column marks 100 Kyr. Note the changing y‐axis scale in the different panels. The last row shows the model for half spreading rate, U0 = 7 cm/yr in which the porosity waves are not persistent as evident by the lack of signals in both the temporal and frequency domains.
Estimate of crustal production and corresponding power spectra for models with half spreading rate, U0 = 2 cm/yr, and varying intrinsic permeability K0. Note the changing y‐axis scale in the different panels.
Modified mobility number, Mo = w0U0UcU0 $\frac{{w}_{0}}{{U}_{0}}\frac{{U}_{c}}{{U}_{0}}$ for varying intrinsic permeability, K0, and half spreading rate, U0. Black dots are the models that have persistent porosity waves. White dots are the models that are lacking in persistent porosity waves. Dots circled yellow are from Parnell‐Turner et al. (2020) and circled red from Sim et al. (2020). The contour of 45 indicates the critical w0/U0 where the models transition from having persistent porosity waves to none.
Scatter plot of spectral peaks observed in bathymetric or crustal thickness data in spreading rate against period domain along with model predictions from this study. Dashed lines correspond to the Milankovitch periods: 23, 41 and 100 kyr. Each color and shape correspond to the studies as laid out in the legend. The smaller circles shaded in gray scale are spectral analysis from this study; black circles have the largest power spectral with amplitudes more than 100. Crowley et al. (2015), Tolstoy (2015) and Boulahanis et al. (2020), had observations limited to less than 100 kyr. The lighter of each shade are the peaks that are ambiguous, that is, 23 kyr peak in Crowley et al. (2015) is not as clear compared to other peaks. Uncertainties and ranges from the studies are not reflected here.
The ocean floor makes up the majority of the Earth's surface and yet, its geomorphology is not fully understood. Recent debate has focused on whether sea level changes—driven by Milankovitch glacial cycles—generate the abyssal hill fabric of the ocean floor by modulating mid‐ocean ridge magma supply. However, periodicities longer than Milankovitch cycles are prominent in the ocean bathymetry. Using crustal thickness estimates from two‐phase flow simulations of ridge magma transport, I show that persistent melt‐rich porosity waves could be responsible for the ocean floor fabric at periods of 100 kyr and longer, except in the case of fast‐spreading ridges. For periods longer than 100 kyr, spectral energy is notably present at large mantle permeabilities regardless of spreading rates. Therefore, two‐phase flow models can provide constraints on elusive mantle parameters such as permeability and viscosity when directly linked to the ocean floor fabric produced.
Maps of the Tibetan Plateau and Gongga‐Zheduo granitic massif and conceptual models for crustal thickening and tectonic uplift in eastern Tibet. (a) Distribution of geophysical anomalies and Cenozoic magmatic rocks in the Tibetan Plateau. Vs perturbation at 30 km depth are from Y. Yang et al. (2012). Pn low‐velocity anomalies at the uppermost mantle are from Zhou and Lei (2016). Light‐yellow arrows mark the inferred location of hypothesized flow channels (Bai et al., 2010). The distribution of magmatic rocks is from Chung et al. (2005) and Hou et al. (2006). Numbers represent suture zones and faults: 1‐Eastern Kunlun‐Animaqing suture zone; 2‐Jinshajiang suture zone; 3‐Longmuco‐Shuanghu suture zone; 4‐Bangong‐Nujiang suture zone; 5‐Indus‐Yarlung Zangbo suture zone; 6‐Ganzi‐Litang suture zone; 7‐Kunlun fault; 8‐Longmenshan thrust fault; 9‐Xianshuihe‐Xiaojiang fault; 10‐Ailaoshan‐Honghe fault; 11‐Sagaing fault. (b) Large‐scale crustal flow model with crustal thickening caused by channelized exotic crustal materials (after Clark, Bush, et al., 2005; Clark & Royden, 2000). (c) Soft crust model with crustal thickening caused by local crustal diffusive deformation. (d) A simplified geological map of the Gongga‐Zheduo granitic massif. Yellow circles denote sample locations. The ages in red are data from this study. The ages in black are from previous studies (Lai & Zhao, 2018; H. Li & Zhang, 2013; H. Li, Zhang, et al., 2015; Searle et al., 2016).
Geochemical and isotopic characteristics of the Gongga‐Zheduo granitic massif. (a) K2O/Na2O vs. SiO2 (wt.%) diagram. (b) MgO (wt.%) vs. SiO2 (wt.%) diagram. Fields of metabasaltic and eclogite melt, and metabasaltic and eclogite melt hybridized with peridotite are after Q. Wang et al. (2006). (c) Sr/Y vs. (La/Yb)N diagram. Subscript N denotes chondrite‐normalization. (d) δEu vs. Ba/Nb diagram. δEu = EuN/[(SmN * GdN)^0.5]. (e) εNd(t) vs. ⁸⁷Sr/⁸⁶Sr(t) diagram. The isotopic compositions of potential source rocks were calculated at 10 Ma. (f) δ¹⁸O (‰) vs. εHf(t) diagram. The δ¹⁸O (‰) values for the mantle are from Bindeman (2008). Modeling parameters are listed in Table S6 in Supporting Information S1.
Thermodynamic and trace element modeling. (a, c) Simplified P‐T phase diagram for the average composition of metasedimentary rocks from the Songpan‐Ganzi Basin and Neoproterozoic mafic rocks from the western margin of the Yangtze Craton, calculated with water contents of 3.7 wt.% and 2.2 wt.%, respectively, corresponding to dehydration melting (Table S7 in Supporting Information S1). The red solid line and red dashed line mark calculated solidus and water saturation of the system, respectively. Brown dashed lines show the calculated degree of melting (wt.% of melt); purple dashed lines and orange dashed lines represent garnet and plagioclase proportion (wt.%) in the residue, respectively. Solution models: G‐Green et al. (2016); FL‐Fuhrman and Lindsley (1988), W/WPH‐White et al. (2014), HP‐Holland and Powell (2011). (b, d) Trace element patterns of granitic melts calculated at specific P‐T conditions and the average composition of potential source rocks. The Kd used for modeling are presented in Table S9 in Supporting Information S1. The blue, pale orange, and red shaded areas represent the overall compositional range of the Subgroup 1 Miocene granites, Subgroup 2 Eocene‐Oligocene granites, and Subgroup 2 Pliocene granites, respectively.
Temporal changes in the characteristics of Cenozoic magmatism in the EMTP. (a) Probability density plot of Gongga‐Zheduo granitic massif, Mianning‐Dechang alkaline rocks, and Batang‐Dali alkaline rocks. Data sources are listed in Table S10 in Supporting Information S1. (b) Exhumation history of the EMTP. Data of Longmenshan and Gongga‐Jiulong areas are from E. Wang et al. (2012) and H. Zhang et al. (2016), respectively. (c–f) Initial ⁸⁷Sr/⁸⁶Sr, εNd(t), εHf(t), δ¹⁸O (‰) vs. age (Ma) of Gongga‐Zheduo granitic rocks. Three magmatic episodes, shown by orange, green and purple dashed boxes, are documented. Symbols for rock units are the same as Figure 2.
Plain Language Summary How the Tibetan Plateau grows outward and deformed remains controversial. A large‐scale crustal flow model has been favored for the expansion of the southeast Tibetan Plateau, arguing that crustal materials could flow hundreds of km resulting in crustal thickening and uplift. Detailed geochemical and isotopic investigations on the largest intrusion (Gongga‐Zheduo) in the eastern margin of the Tibetan Plateau show that their magmatic source is local crustal rocks of the Songpan‐Ganzi terrane without the input of crustal materials from central Tibet. Thermodynamic and trace element modeling results show that the Cenozoic magma is derived from ∼30 to 40 km depth, similar to the depth of postulated crustal flow. The results are inconsistent with the large‐scale eastward crustal flow model. A repeated shifting of magmatic sources during the Cenozoic is correlated with crustal uplift. Mantle‐crust interaction plays a primary role in the formation of magmatism and modifying crustal rheology. The continued collision between the Indian and Asian blocks and upwelling of the asthenosphere contribute to the crustal deformation and uplift.
Map illustrating major tectonic features and distribution of seismic stations. Red volcano symbols represent the Holocene volcanic centers: CBV―Changbaishan volcano; LGV―Longgang volcano; JPHV―Jingpohu volcano. Depth contours (in km) of the subducting Pacific slab are outlined by white‐dashed lines. The Tanlu fault zone (TLFZ) is delineated by thin black lines and the Songliao basin (SLB) is located to the northwest of our study region. Seismic stations are marked by triangles with different colors explained in the legend (the top‐right inset). Station MDJ denoted in yellow is chosen to plot the ambient noise cross‐correlation records as displayed in Figure S2 in the Supporting Information S1. The cyan box outlines the simulation region for the adjoint tomography. The top‐left inset shows our study region in a larger scale with the Late Cenozoic volcanoes and westward subduction of Pacific plate indicated by small red dots and the black arrow, respectively. The top‐middle inset is a panoramic view of the nearly 5‐km‐wide caldera of CBV.
(a) Total misfit as a function of the iteration number. Blue stars, purple triangles, and green diamonds represent the total misfit measured at the period bands of 5–10 s, 10–20 s, and 20–40 s, respectively. Red dots denote the averaged total misfit of the three period bands. (b–d) Histograms of traveltime misfit for the initial model (M00) and the final inverted model (M07) at the three period bands. The blue solid bars and the red hollow bars represent misfits for the initial model and the final model, respectively. The mean and standard deviation values of the misfits for the initial model and the final model at different period bands are placed in the upper‐left and upper‐right corners with the same colors specified accordingly.
(a–d) Horizontal slices of Vs at the upper, middle, lower crustal, and uppermost mantle depths. (e–g) Vertical cross sections of Vs transecting two of Changbaishan volcano (CBV), Longgang volcano (LGV), and Jingpohu volcano (JPHV) with their surface locations shown in (a). Please note that the vertical profiles have a V/H ratio of ∼3.5, which means that the vertical dimension is exaggerated by a factor of ∼3.5.
3D view of the crust and uppermost mantle magma plumbing system beneath Changbaishan volcano. The red ellipsoid‐like body, which is enclosed by an iso‐surface of 3.45 km/s, is interpreted as the lower‐crustal mush zone, whereas the yellowish curved surface represented by an iso‐surface of 3.80 km/s indicates the magma ascending path in the uppermost mantle.
Plain Language Summary The enigmatic and active Changbaishan volcano located at the border between China and North Korea is an ideal laboratory for investigating the origin and evolution of continental intraplate volcanism. However, the seismic structures related to the magma plumbing system beneath this volcano are still debated and not well resolved. In this study, we build a high‐resolution seismic velocity model at Changbaishan volcanic region combining seismic data from both China and North Korea. We find significant low velocities in the lower crust of Changbaishan volcano, which is interpreted to be a deep crustal melt‐bearing zone with an estimated melt fraction of ∼1.5%–3.6%. Three narrow channel‐like low‐velocity anomalies observed in the uppermost mantle below Changbaishan, Longgang, and Jingpohu volcanoes are suggested to reflect magma ascending passages that transport deep‐sourced mantle melts upward. Our detailed tomographic images provide tight seismic constraints and deepen our understanding on the magma generation and evolution dynamics associated with the young intraplate volcanism in northeast China.
Annual radiative forcing (RF) (W m⁻²) due to the albedo change and RF relative‐difference between the two model scenarios S1 and S2 from 1983 to 2010 derived from LSR‐land‐use change (LUC) inventory (solid blue line) and HSR‐LUC inventory (solid red line). Gray shading indicates the uncertainty interval estimated by Monte Carlo simulations. The dashed yellow line stands for a relative‐difference of annual RFs using the LSR‐LUC inventory (RFS1) and HSR‐LUC inventory (RFS2). Note that the LUC‐induced RF in the current year was estimated by summing the LUC‐induced RF in the previous year (base year) and the present year.
Radiative forcing due to albedo change derived from model scenario 2 (S2) with the land‐use transition from 1982 to 2010 (solid red line) and model scenario 2 with a fixed land type without transition (S2_land‐use, solid blue line) for six land‐use types, including cropland (cro, Figure 2a), desert (des, Figure 2b), forest (for, Figure 2c), grassland (gra, Figure 2d), shrubland (shr, Figure 2e), and urban (urb, Figure 2f). The inset bar chart represents the relative‐contribution of two‐way land‐use net transition between the land‐use of interested and other LUs from 1983 to 2010. Taking the bar chart in Figure 2a as an example, the bars with different colors show the result of the two‐way transition between cropland to and from other land‐use occurring from 1983 to 2010. Positive bars represent the conversion from other land use to cropland, and negative bars indicate the transition from cropland to other land use.
(a) RFCO2 (W/m²) derived from land‐use change (LUC)‐induced‐CO2 from the LSR‐LUC and updated HSR LUC inventories, and the relative‐difference in RFCO2 calculated using the two inventories from 1983 to 2010; (b) contribution (%) of each of LUC among six LUCs to annual RFCO2 during the same period; (c) contribution (%) of LUC among six LUCs to mean RFCO2 averaged over 1983–2010.
Plain Language Summary Land‐use change (LUC) is considered the second anthropogenic source of climate change after fossil fuel combustion. However, significant uncertainties remain in the estimate of radiative forcing (RF) induced by LUC, partially attributable to the lack of reliable LUC data with a high spatiotemporal resolution. This study incorporated a new LUC data set with a high spatiotemporal resolution into a compact earth‐system model OSCAR to quantify the response of RF to LUCs from 1982 to 2010 in China. We assessed changes in RF for this period subject to the altered surface albedo and carbon emission associated with human‐disturbed land‐use transitions. We compared estimated RF values with those obtained using previously adopted LUC data with a low spatiotemporal resolution, which failed to identify detailed LUCs occurring in China for the past four decades. We show that the updated LUC data set weakens the cooling effect featured by negative RF‐induced by surface albedo variation but significantly enhances positive RF due to CO2 emissions from LU transition. We identify that the LU transition between grassland and cropland and between cropland and forest made the most significant contribution to the changes in RF, attributable to China's national strategies for urbanization, conservation of agricultural resources, and forest expansion.
(The left panels) Equatorial electron flux distributions as a function of the equatorial pitch angle and the energy at t = 0 and 60 s (The right panel) The electron pitch angle distributions at four energy channels (491.1 keV, 1. 1, 2, and 3.9 MeV) at t = 0, 20, 40, and 60 s.
The distribution of ργ as a function of the equatorial pitch angle and the energy with whistler wave frequency of 0.2fce,eq, 0.3fce,eq, and 0.4fce,eq. The solid and dashed lines correspond to contour lines of ργ = 1 and 5, respectively.
(Left panel) The ratio of the electron fluxes at t = 0 and t = 60 s (Right panel) The distribution of the origin of electrons scattered into the green square region ranging from 1.5 to 3 MeV in energy and from 56° to 70° in equatorial pitch angle at t = 60 s. The solid and dashed lines in both panels corresponds to the contour lines of ργ = 1 and 5 with whistler wave frequency of 0.3fce,eq, respectively.
(The left panels) The bounce averaged diffusion coefficients in equatorial pitch angle. (The right panels) The bounce averaged diffusion coefficients in energy. (The top panels) The bounce averaged diffusion coefficients in the cyclotron resonance condition satisfying k∥v∥ > 0. (The middle panels) The bounce averaged diffusion coefficients in the cyclotron resonance condition satisfying k∥v∥ < 0. The solid and dashed lines correspond to the lines of ργ = 1 and 5 with whistler wave frequency of 0.3fce,eq, respectively. (The bottom panels) The bounce averaged diffusion coefficients at energy of 512.5 keV.
Plain Language Summary Radiation belt electrons have various pitch angle distributions in response to global/local processes arising in the magnetosphere. Butterfly pitch angle distribution is a characteristic feature of the electron pitch angle distribution, which has the maximum flux intensity at a pitch angle lower than 90°. Wave‐particle interactions have been proposed as a driver for the butterfly distribution in the heart of the radiation belt. However, it is in debate how the wave‐particle interactions contribute to the formation of the butterfly distribution of multi‐megaelectron (MeV) volt electrons that is “killer electrons.” In this Letter, we report that lower band whistler chorus waves play an important role for the electron butterfly distribution at MeV energies. A numerical simulation was carried out and showed that electrons nonlinearly scattered by the whistler chorus waves produce the butterfly distribution at MeV energies. The simulation also showed the upper limit of the rapid electron acceleration in the formation of the butterfly distribution. The simulation results advance our understanding of a formation mechanism of MeV electron butterfly distribution driven by whistler chorus waves.
Example showing lake ice phenologies determined using satellite observations and numerical models. (a) Example of how the dual logistic regression model (DLRM) was used to identify the freeze‐up date, break‐up date, and ice duration using satellite observations. Annual time series of MODIS snow cover fraction within Montreal Lake, Canada, with larger points representing more valid satellite observations and thus greater weight in the dual logistic regression (see Text S1 in Supporting Information S1). The freeze‐up and break‐up dates were determined as 29 October and 14 May 14, respectively. (b) Simulated lake ice phenologies determined using the lake‐specific models and daily mean surface temperatures (ST), which are 2 November and 13 May for freeze‐up and break‐up, respectively. (c) The freeze‐up process for Montreal Lake observed using 10‐m resolution Sentinel MultiSpectral Instrument (MSI) images (RGB true color) and 500‐m resolution MODIS snow cover products. (d) Same as (c), but for the break‐up process. MSI images show that the freeze‐up and break‐up dates were around 30 October and 15 May, which agreed well with the results determined by the dual logistic regression in (a) and numerical model in (b).
Lake ice phenologies detected using MODIS satellite images between 2001 and 2020 (a–c) Median values of the freeze‐up date, break‐up date, and ice duration for each of the 30,063 examined lakes during the 20 years studied. Note that the data are presented as day of year (DOY) (d–f) The associated linear trends (i.e., linear regression slope) and significance in freeze‐up date, break‐up date, and ice duration are shown in the middle panels, with statistically significant (P < 0.05) trends shown using larger dots. Each dot in a–f represents a lake. (g) Histogram distributions for a–c. (h) Histogram distributions for d–f.
Past, present, and future trends of lake ice phenologies. (a–c) Linear slopes (i.e., rates of change) of the median values of freeze‐up, break‐up, and ice duration for global lakes. The slope values and their significance levels (student's t‐test) for three periods are annotated within the panels. The data of the present period are from MODIS satellite observations; past and future data were simulated based on air temperature. The shaded areas associated with the past and future median values of each year represent standard deviations of the simulations from temperatures projected by four climate models. (d) The number of lakes with warming and cooling trends for different lake ice phenologies in the three periods, hatched shades represent significant trends.
Past and future changes in lake ice phenologies. (a, e, i) Past period, (b, f, j) RCP 2.6, (c, g, k) RCP 6.0, and (d, h, l) RCP 8.5. The upper, middle, and bottom panels are the freeze‐up, break‐up, and ice duration, respectively. The changes were estimated as the anomalies of the model‐simulated historical (i.e., 1900–1919) and future (i.e., 2080–2099) median values with respect to satellite‐observed median values between 2001 and 2020 (see Text S1 in Supporting Information S1). Warming induced ice‐free lakes are indicated using black dots. (m) Histograms of past and future changes for all lakes.
Lake ice loss has been detected worldwide due to recent climate warming, yet spatially and temporally detailed information on the changes inglobal ice phenology does not exist. Here, we build a global lake ice phenologydatabase comprising three lake ice phenologies –freeze-up, break-up, and ice duration –for each year acrosstwocenturies (1900-2099). The timing ofall three phenologies experienced mild but statistically significant warming trends in the 20thcentury; continued warming trends were detected in ~60% of the lakes from 2001 to 2020. Under a high emissions scenario (RCP 8.5), future global median ice duration would be shortened by 49.9 days by the end of the 21stcentury; such changecan be substantially reduced under lower emission scenarios. We revealed continuous loss of global lake ice during the observed period, our generated database provides critical baseline information to evaluate the consequences of historical and future lake ice changes.
Bathymetry of Alexandra Canyon at (a) moderate discharge (5,500 m³/s) in June 2016 and (b) low discharge (2,175 m³/s) in August 2016. Distance along the canyon is measured in meters from the canyon entrance. (c) Elevation differences between the two observations. Differences ±0.5 m have been masked as they do not exceed the vertical accuracy. Extreme differences along the channel margins are due to lateral positioning error between surveys.
Centerline transects of velocity and channel geometry through Alexandra Canyon with vertical black lines representing the extent of bedrock confinement. (a) Observations from June at a discharge of 5,500 m³/s with all velocity data collected using the 600 kHz Acoustic Doppler Current Profilers (ADCP). (b) Observations from August at a discharge of 2,175 m³/s with transects 5, 6, and 8 collected using the 1,200 kHz ADCP, others using the 600 kHz ADCP. Transects 4/5, and 6/7 are collected simultaneously on different instruments. Note: Velocity plots in (a and b) do not share color‐bar scaling.
Cross‐sections of (a) horizontal and (b) vertical velocity plotted in Alexandra Canyon at low (2,175 m³/s) and moderate (5,500 m³/s) flow. Bed topography is MultiBeam EchoSounder data (Figure 1) with river meters (RM) from the bedrock canyon entrance.
Surface (Usurf), near‐bed (Ubed), and depth averaged (Uda) velocities at (a) moderate flow (5,500 m³/s) and (b) low flow (2,175 m³/s) in Alexandra Canyon. Also plotted is the relative depth of the maximum velocity in the water column (δmax = 1 − zmax/h where zmax is the height of the maximum velocity and h is depth). Calculations use raw data, but Ubed has a 3‐point running average applied for visualization. Low flow is Transect 4 and moderate flow is Transect 3. The canyon ends at River Meter (RM) 825.
Plain Language Summary Incision in bedrock rivers sets the pace of landscape evolution by controlling the rate of geomorphic responses to climatic and tectonic signals, yet the processes driving incision occur at much finer scale than those captured by landscape evolution models. Local bedrock river incision is driven by flow structures that are not well understood. Rivers typically flow fastest near the surface and slowest near the bed, but many bedrock rivers have channel morphologies that cause this velocity/depth relation to invert. The fastest‐flows submerge toward the bed enhancing near‐bed velocities, sediment transport, and consequently the potential for bedrock incision by particle impacts. However, the first observations of these “plunging flows” were from relatively low discharges and it is not clear if they persist during floods. Here we show that plunging flows get stronger during floods, which clears sediment cover that protects the underlying bedrock and increases bedrock incision potential. The length of the plunging flows matches their coincident pools which are common features of bedrock rivers, explaining why these pools exist. Formation of deep scour pools by complex flow structures in bedrock‐confined rivers is the mechanism that drives incision, begging for a re‐examination of the models used to explore landscape evolution.
Demonstration of the proposed method with event 1 (2011 Tohoku, Mw 9.08) as an example. Panel (a) is the vertical‐component displacement at station TAU (epicentral distance is 80°). The green dotted line marks the beginning of the coda examined at 10,000 s after the event time (at time 0, vertical dark dotted line). Panel (b) shows the corresponding coda energy curve of (a) at the frequency of 5.0 mHz. The linear and curve fittings follow Equations 1 and 2 with the results labeled. The linear fitting only uses points before the vertical blue line (i.e., a window of 10,000 s to 50,000 s). (c and d) Results of ln Ri (Equation 5) and coda moment magnitude Mwo (labeled) derived from real data (c) and synthetics (d). Each dot shows the value at one station. Different colors represent the sampling frequencies we used (3.3, 5.0, and 6.7 mHz, respectively). The gray dashed line is the mean of all the points. The synthetic data in (d) are obtained for the 3D model (s40rts, Ritsema et al., 2011 together with CRUST2.0, Bassin et al., 2000). The calibration model is the 3D model in (c) and the 1D model (PREM, Dziewonski & Anderson, 1981) in (d).
results for different earthquakes in this study. (a and b) Comparisons between log10R from coda energy and log10M0 from Global Centroid Moment Tensor (GCMT) with the synthetic (a) and real data (b). The computing codes and reference models used to acquire the E0M0=1(ω) ${\mathrm{E}}_{0}^{{M}_{0}=1}(\omega )$ are indicated in the legend. A couple of noticeable departures (events 8&9) are labeled. The gray dashed line is the theoretical prediction with a slope of 2 (Equation 7). (c) Mwo calculated in this study (using the 3D model calibration). The solid circles are the results of events in Table S1 of Supporting Information S1 (Mw > 8) and the open circles show the results for the smaller events (Table S2 in Supporting Information S1) in the test of the lower limit of the magnitude. The error bars show the one standard deviation (SD) of the values from different stations. Note the error bars for the smaller events are clearly larger due to smaller signal‐to‐noise ratios, relative to the bigger events. (d) Fitted δ(ω) for different events and frequencies with one SD. The horizontal line is the averaged δ(ω) at the frequency for events with Mw > 8.
Synthetic tests on factors influencing the R value. The synthetics are for the 3D model but the calibration model is 1D. (a and b) Tests on source time function (STF) and reference Q model, respectively (event 1 as an example). Each point is the result of one station and colors represent the three frequencies. Solid and dashed lines represent averaged ln R values before and after the change, respectively. The distribution of ln Ri in (a) is calculated when the STF is underestimated by 60% and in (b) when model Q values are 2/3 of the PREM values. (c) Test results of Mwo on STF (circles) using event 1 as an example and Earth Q model (pluses) using events 1&10 as examples. The short/long STF are the shortest/longest durations among the 10 events. The test Q model is 2/3 of the PREM values. (d) Results of tests by placing the moment tensor (MT) of event 3 at the location of event 8 (case 1) or 9 (case 2) to demonstrate the influence of both Earth structure and MT on Mwo (see text).
We present a novel and robust method for estimating moment magnitudes (Mw) of large earthquakes with long‐period and long‐lasting coda energy. Fitting the energy with a simple decay model, we derive a straightforward relationship between the coda energy and the Mw. Tests with both real and synthetic data of 10 globally distributed large earthquakes (Mw > 8) verify the method and the results are stable and reliable even with a fast calculation of synthetics for a 1D model. Tests also show that the method is applicable for earthquakes with Mw above 7.5–8.0 with energy sufficiently greater than the ambient noise. The method removes or reduces the effects of geometric spreading, focal mechanism, source rupture process, and the actual Earth structure, making it advantageous for estimating the magnitudes of large earthquakes. The new long‐period coda moment magnitude (Mwo) estimations are similar to the conventional solutions but are slightly larger (by 0.04 on average).
Analysis of different properties of the rocks used in this study: (a) Pore size distribution as determined by MICP method (b) Capillary entry pressure versus mercury saturation as determined by MICP method (c) Capillary entry pressure for air versus water saturation as determined by centrifuge capillary pressure test and (d) Bulk mineral composition of each sample as determined by XRD analysis (see Table S1 in Supporting Information S1).
Schematic of the setup used for gas‐brine relative permeability measurements; to restrain gravity segregation into the core sample during the gas injection process, the core holder is placed in a vertical mode. The sizes of different objects have been re‐scaled to make them visible.
(a–g) Drainage H2 and different brine relative permeability curves for 2 sandstone and 1 carbonate samples at different pressures. (h) Drainage N2 and brine relative permeability curves for sample S2(i) Drainage CH4 and brine relative permeability curves for sample S2.
Drainage gas relative permeability curves versus gas saturation to compare (a) effect of pressure on drainage hydrogen relative permeability (b) effect of salinity on hydrogen drainage relative permeability (c) effect of rock type and rock pore structure on hydrogen drainage relative permeability and (d) hydrogen drainage relative permeability with that of N2 and CH4.
Geological hydrogen storage in depleted gas fields represents a new technology to mitigate climate change. It comes with several research gaps, around hydrogen recovery, including the flow behavior of hydrogen gas in porous media. Here, we provide the first‐published comprehensive experimental study of unsteady state drainage relative permeability curves with H2‐Brine, on two different types of sandstones and a carbonate rock. We investigate the effect of pressure, brine salinity, and rock type on hydrogen flow behavior and compare it to that of CH4 and N2 at high‐pressure and high‐temperature conditions representative of potential geological storage sites. Finally, we use a history matching method for modeling relative permeability curves using the measured data within the experiments. Our results suggest that nitrogen can be used as a proxy gas for hydrogen to carry out multiphase fluid flow experiments, to provide the fundamental constitutive relationships necessary for large‐scale simulations of geological hydrogen storage.
The evolution of a current sheet without foreshock ions. Panel (a) shows the positions of THEMIS A (magenta), B (red), C (green), D (cyan), and E (blue) in the Geocentric solar ecliptic (GSE) coordinate at about 5:00 UT on 10 June 2009. Here, the positions of bow shock and magnetopause are obtained from the models in the previous papers (Shue et al., 1998; Wu et al., 2000). Panels (b–j) show (b) magnetic field in the LMN coordinate, (c) total magnetic field, (d) electron density, (e) ion temperature, (f) ion velocity in the LMN coordinate, (g) ion total speed, (h) dynamic pressure, (i) ion energy fluxes, and (j) electron energy fluxes, respectively, that were observed by THEMIS B in the solar wind. Panels (k–q) show (k) magnetic field, (l) total magnetic field, (m) ion density, (n) ion velocity in the GSE coordinate, (o) total pressure (black), dynamic pressure (blue), magnetic pressure (green) and thermal pressure (red), (p) ion energy fluxes, and (q) electron energy fluxes, respectively, that were observed by THEMIS C in the magnetosheath. Panels (r–v) show (r) magnetic field, (s) ion density, (t) ion velocity in the GSE coordinate, (u) ion energy fluxes, and (v) electron energy fluxes, respectively, observed by THEMIS A in the magnetosphere.
Panel (a) shows the positions of ACE (black), THEMIS B (red) and THEMIS C, respectively, at about 19:40 UT on 03 August 2008. Panel (b) shows the positions of THEMIS A (magenta), B (red), C (green), D (cyan), E (blue), and GOES satellites (orange), respectively, at about 19:40 UT on 03 August 2008. Panels (c–f) show (c) magnetic field in the LMN coordinate, (d) total magnetic field, (e) magnetic field cone angle in the GSE coordinate, and (f) ion velocity in the LMN coordinate, respectively, that were observed by ACE in the solar wind. Panels (g–p) show (g) magnetic field in the LMN coordinate, (h) total magnetic field, (i) magnetic field cone angle in the GSE coordinate, (j) electron density, (k) ion temperature, (l) ion velocity in the LMN coordinate, (m) ion total speed, (n) dynamic pressure, (o) ion energy fluxes, and (p) electron energy fluxes, respectively, that were observed by THEMIS B in the solar wind. Panels (q–z) show the same format of panels (g–p), except that these were observed by THEMIS C.
Panels (a–g) show the same format of Figures 1k‐1q except that these were observed by THEMIS D on 03 August 2008. Panels (h–k) show (h) magnetic field, (i) ion density, (j) ion velocity in the GSE coordinate, and (k) ion energy fluxes, respectively, observed by THEMIS A in the magnetosheath. Panels (l–n) show the detrended magnetic fields observed by GOES 10–12, respectively in the magnetosphere.
The illustration of the geoeffectiveness of solar wind current sheets with and without foreshock ion modulations. The gray dots are rough locations of in‐situ observations in this study. The pink and light green lines are IMF, and the arrows on them showing the directions of IMF.
Plain Language Summary Previous observations and models have shown that current sheets are associated with plateaus in plasma density, which result in dynamic pressure increases and potentially impact the Earth's magnetosphere. Thus, the evolution of solar wind current sheets in the magnetosheath and their potential effects in the magnetosphere are examined based on the in‐situ observations of the Time History of Events and Macroscale Interactions during Substorms (THEMIS) probes and the Geostationary Operational Environmental Satellites (GOES) satellites. Since these solar wind current sheets are associated with magnetic field changes, they can also control the appearance and disappearance of foreshock ions which, in turn, may modulate the geoeffectiveness of solar wind current sheets. In this study, we found that the current sheets without foreshock‐ion modulations generated dynamic pressure plateaus, which further evolved to dynamic pressure plateaus in the magnetosheath and likely induced waves in the magnetosphere. Some current sheets were modulated by foreshock ions. These foreshock ions enhanced the perturbations within the current sheets, which further compressed the magnetosphere and caused more oscillations than the current sheets without foreshock‐ion modulations. Thus, we conclude that solar wind current sheets can be geoeffective and their geoeffectiveness can be amplified by foreshock ions inside their structure.
(a) Schematic map of the Beaufort Sea with the observed winter sea ice thickness from CS2/SMOS (shading), ice flow from neXtSIM (arrows), and mean sea‐level pressure from ERA5 (solid and gray lines) all shown on 23 February 2013. (b) Daily categorical lead map following Willmes and Heinemann (2015) based on the Moderate Resolution Imaging Spectroradiometer imagery. (c) Simulated lead fraction using WRF10 as the atmospheric forcing. (d) Time series of lead area fraction in the Beaufort Sea for the model (blue) and Arcleads (gray‐dashed line). Leads are defined as areas where the lead fraction exceeds 5%. The shading shows the sensitivity to using a threshold value of 3% and 7%, respectively. The r‐value is the correlation coefficient between the observed and modeled lead fraction. Both (b and c) for 18, 23 February and 1 March 2013 are marked by green triangles in (d).
Time series during the breakup event of (a) wind speed and direction (red arrows; up = away from Banks Island), (b) sea ice velocity and lead propagation indicated by the 5‐cm s⁻¹ ice velocity contour, and (c) observed ice velocity from OSISAF and lead propagation from Arcleads data (purple lines). All variables are calculated along the transect indicated in (b) running from the western Beaufort Sea (close to Point Barrow) to Banks Island. Black lines in (a and b) represent the 10‐m s⁻¹ wind speed and 5‐cm s⁻¹ ice velocity, respectively. The red‐dashed line corresponds to the WRF80 experiment.
Time series of (a) mean sea ice velocities in the Beaufort Sea and maximum winds (gray‐dashed line) in the along‐transect direction (inset in Figure 2b). (b) Average lead fraction shown as a % of the total Beaufort Sea area. The area is outlined in Figure 4a.
(a) Histograms of the normalized sea ice thickness in the Beaufort Sea before (blue) and after (orange) the breakup event. (b) Cumulative thermodynamic ice growth in the Beaufort Sea, calculated for new ice (blue line), young ice (orange line), and old ice (green line). (c) Total thermodynamic ice growth (solid line) and ice volume change (ΔSIV; dashed line) in the Beaufort Sea for WRF10 (blue) and no_motion (gray). In no_motion, the sea ice dynamics are turned off. (d) Time series of sea ice volume flux, where positive values correspond to an export out of the Beaufort Sea. The total flux is split into contributions from newly formed sea ice (SIT < 1 m), first‐year ice (1 > SIT < 1.6 m), and multiyear ice (SIT > 1.6 m). The spatial distribution of the thickness classes is shown in (a) for February 13.
Plain Language Summary The loss of thick multiyear sea ice in the Arctic leads to weaker sea ice that is more easily broken up by strong winds. As a consequence, extreme sea ice breakup events may become more frequent, even during the middle of winter when the sea ice cover is frozen solid. This can lead to an earlier onset of the melt season and potentially accelerate Arctic sea ice loss. Such extreme breakup events are generally not captured by climate models, potentially limiting our confidence in projections of Arctic sea ice. We investigated the driving forces behind sea ice breakup events during winter and how they change in a future climate. Our sea ice model is the first to reproduce such breakup events and reveals that the combination of strong winds and thin sea ice is a key factor for these breakups. We found that winter breakups have a large effect on local heat and moisture transfer and cause enhanced sea ice production, but also increase the overall movement of the sea ice cover, making it more vulnerable. Finally, we show that if the Arctic sea ice continues to thin, these extreme breakup events could become even more frequent.
Fuel types, management zones, and historical fire distribution in Banff, Kootenay, and Yoho National Parks. Fires were derived from the Canadian Landsat Burn Severity (CanLaBS) product from 1985 to 2015 (Guindon, Villemaire, et al., 2020). Fuel types in the Canadian Forest Fire Behavior Prediction (FBP) System (Forestry Canada Fire Danger Group, 1992): C1—Spruce‐Lichen Woodland; C2—Boreal Spruce; C3—Mature Lodgepole Pine; C4—Immature Lodgepole Pine; C7—Douglas Fir; D1—Leafless Aspen; M1/2—Boreal Mixedwood—Leafless/Green; O1—Standing Grass.
Relative importance of the predictors (Table S1 in Supporting Information S1) for predicting burn severity using random forest regression models. Mgmt_Zone represents the management zone.
Predicted burn severity (dNBR) in the study area in 2002 and 2012 for different fire causes and the dNBR differences of different predictions. Three severity levels were identified based on the thresholds given by Key and Benson (2006): low‐moderate severity (0.100–0.439), moderate‐high severity (0.440–0.659), and high severity (0.660–1.300).
Examples of the predicted burn severity with relatively large changes for different fire causes and prediction years in (a) Banff National Park, (b) Kootenay National Park, and (c) Yoho National Park.
Plain Language Summary Understanding the spatial pattern of burn severity is crucial for fire‐related ecological research and effective fire management. The Canadian Rocky Mountain region is characterized by mixed‐severity fires, which makes fine‐scale burn severity investigation a challenge. This study used random forest models to establish the relationships between observed burn severity and various environmental predictors (fuel type, fire cause, management zone, topography, vegetation, and climate) and identify key drivers of burn severity in three Canada's mountain national parks (Banff, Kootenay, and Yoho). The prediction models were applied to predict the burn severity potentials by human‐ and lightning‐caused fires for all forest locations in the study area in 2002 and 2012, that is, the 2 years with available data. The results contribute to a more comprehensive understanding of regional fire behavior. The estimated important influences of fuel type, topography, vegetation, and climate on regional burn severity indicate the complex mechanism of environmental controls on fire behavior. The predictions of burn severity in the parks showed an overall consistent spatial pattern over time, which provide a baseline for relevant fire ecology research and useful information for park conservation.
(a) Map of ice‐covered areas in Greenland. Greenland ice sheet (white). Peripheral glaciers in the north (black dots), northeast (green dots), southeast (red dots), southwest (blue dots), and northwest (purple dots). Mean surface air temperature in °C during May–September in (b) the north, (c) northwest, (d) southeast, (e) southwest, and (f) northwest Greenland from RACMO2.3p2. The straight line in panel (b–f) denotes 1990–2021 trend with the rate listed above the line. Area size in km² for each peripheral glacier region and the Greenland ice sheet are listed in Table 1.
Elevation change rates, in m/yr, during (a) February 2003–October 2009, (b) October 2008–April 2019, and (c) October 2018–December 2021.
Top row: Detailed elevation change rates of Figure 2 during (a) February 2003–October 2009, (b) October 2008–April 2019, and (c) October 2018–December 2021 for peripheral glaciers in the North. Bottom row: Detailed elevation change rates for peripheral glaciers in the northeast during (d) February 2003–October 2009, (e) October 2008–April 2019, and (f) October 2018–December 2021.
Plain Language Summary The Arctic is warming more rapidly than the rest of the world. This warming has had an especially profound impact on Greenland's ice cover. Only 4% of Greenland's ice cover are small peripheral glaciers that are distinct from the ice sheet proper. Despite comprising this relatively small area, these small peripheral glaciers are responsible for 11% of the ice loss associated with Greenland's recent sea level rise contribution. Using the satellite laser platforms Ice, Cloud, and land Elevation Satellite (ICESat) and ICESat‐2, we estimate that ice loss from these Greenland glaciers increased from 27 ± 6 Gt/yr (2003–2009) to 42 ± 6 Gt/yr (2018–2021). We find that the largest acceleration in ice loss is in North Greenland, where we observe ice loss to increase by a factor of four between 2003 and 2021. In some areas, it appears that recent increases in snowfall at high altitudes have partially counteracted recent increases in melt at low altitudes. While many recent Greenland ice loss assessments have focused on only the ice sheet, the recent sharp increase in ice loss from small peripheral glaciers highlights the importance of accurately monitoring Greenland's small peripheral glaciers. These small peripheral glaciers appear poised to play an outsized role in Greenland ice loss for decades to come.
CMIP5 multi‐model mean of average NDJFMA changes in (a) precipitation minus evapotranspiration, (b) dynamic component, and (c) thermodynamic component (d–f) Same as (a–c) for CMIP6 multi‐model mean (g–i) same as (a–c) for CMIP5 and CMIP6 combined multi‐model mean. Boxes show the boundaries of the Mediterranean region, Pacific Northwest region and subtropical Pacific region for Figure S3 in Supporting Information S1. The contours and colorbars are chosen to emphasize midlatitude changes.
(top) Correlation coefficients across all 48 models of the NDJFMA change in precipitation minus evapotranspiration with the NDJFMA change in the (a) dynamic component and (b) thermodynamic component (bottom) Correlation of the change in ΔTglobal with the change in the (c) thermodynamic term and (d) P − E for each gridpoint. Boxes show the boundaries of the Mediterranean region, Pacific Northwest region, and subtropical Pacific region for Figure S3 in Supporting Information S1. Correlation coefficients exceeding ±0.29 can lead to the rejection of a null hypothesis of no relationship at the 95% confidence level using a two‐tailed Student‐t test given 48 distinct models.
Correlation coefficients across all 48 models of the (top) NDJFMA and (bottom) MJJASO change in precipitation minus evapotranspiration with (blue) Δ dynamic component and (red) Δ thermodynamic components, after (left) first performing a zonal average, and (right) first performing a zonal average and a meridional averaging over a window of 10°. Gray shading indicates a correlation not statistically significant at the 95% level using a two‐tailed Students‐t test.
End of century projections from Coupled Model Intercomparison Project (CMIP) models show a decrease in precipitation over subtropical oceans that often extends into surrounding land areas, but with substantial intermodel spread. Changes in precipitation are controlled by both thermodynamical and dynamical processes, though the importance of these processes for regional scales and for intermodel spread is not well understood. The contribution of dynamic and thermodynamic processes to the model spread in regional precipitation minus evaporation (P − E) is computed for 48 CMIP models. The intermodel spread is dominated essentially everywhere by the change of the dynamic term, including in most regions where thermodynamic changes drive the multi‐model mean response. The dominant role of dynamic changes is insensitive to zonal averaging which removes any influence of stationary wave changes, and is also evident in subtropical oceanic regions. Relatedly, intermodel spread in P − E is generally unrelated to climate sensitivity.
(a) 2004–2018 mean dynamic height at the surface relative to 1975‐dbar from Argo (Roemmich & Gilson, 2009), high‐resolution expendable bathythermograph (HR‐XBT) nominal transects (black), and associated western boundary currents (blue). Abbreviations are: AC ‐ Agulhas Current; ARC ‐ Agulhas Return Current; EMC ‐ East Madagascar Current; EAC ‐ East Australian Current; EACx ‐ East Australian Current Extension; TF ‐ Tasman Front; Ku ‐ Kuroshio; KE ‐ Kuroshio Extension. (b–d) 2004–2019 mean depth‐integrated (0–1975 m) absolute geostrophic velocity across mean HR‐XBT transects (b) IX21, (c) PX30, and (d) PX40. Depth contours are 200 m (green), 1000 m (blue‐green), and 2000 m (blue). Izu Ridge (referred to in the text) is labeled in (d).
Longitude‐time Hovmöller plots of depth‐integrated (0–1975 m) absolute geostrophic velocity for the western section of high‐resolution expendable bathythermograph (HR‐XBT) transects (a) IX21, (b) PX30, and (c) PX40 sampling the Agulhas Current, East Australian Current (EAC), and Kuroshio respectively. Negative values (blue) indicate southward flow. Yellow and black lines identify the core and offshore edge of the western boundary current. Filled circles on the right indicate the dates of HR‐XBT transects. Low depth‐integrated velocities near 160°E (PX30) are due to a shallow seamount (Figure 1c).
Annual cycles in the Agulhas Current (left), East Australian Current (EAC, middle), and Kuroshio (right) for: (a–c) western boundary current transport (positive values northward); (d–f) core depth‐integrated velocity (positive values northward); (g–i) offshore edge deviations (positive values eastward). Shading is ±1 standard error. Dashed line is the 2004–2019 mean.
(a) Monthly time series of Agulhas Current core longitude. Dashed line indicates +1 standard deviation from the mean. (b) Absolute cross‐transect geostrophic velocity composite for Natal Pulse (NP) events. Negative velocities are poleward. (c) As in (b) but for non‐NP periods. (d) Difference between the two composites (NP minus non‐NP). Hatching indicates where differences are not significant at the 95% confidence level. For (b–d) the black contour denotes 0 m s⁻¹ and the solid and dashed gray contours are at intervals of +0.1 and −0.1 m s⁻¹. (e) Composite of monthly sea level anomaly (SLA) for NP months. The black line is the IX21 nominal transect and the yellow marker the mean location of the Agulhas Current core. The white SLA contour denotes 0 m and the solid and dashed black SLA contours are at intervals of +0.05 and −0.05 m.
(a) Absolute cross‐transect geostrophic velocity composite for periods when the Kuroshio Extension Index (KEI) is positive >12σ $\left( > \frac{1}{2}\sigma \right)$. Positive velocities are poleward. (b) As in (a) but for negative KEI <−12σ $\left(< -\frac{1}{2}\sigma \right)$. (c) Difference between the two composites (positive minus negative). Hatching indicates where differences are not significant at the 95% confidence level. For (a–c) the black contour denotes 0 m s⁻¹ and the solid and dashed gray contours are at intervals of +0.1 and −0.1 m s⁻¹.
Plain Language Summary Western boundary currents are major ocean currents located on the western side of the world's oceans. These currents transport warm water toward the poles, which influences regional weather and climate. However, despite their importance, the fast speeds, high variability, and narrow width of these currents makes them difficult to observe. Here we examine the western boundary currents (WBCs) of the Indian Ocean (Agulhas Current) and Pacific Ocean (East Australian Current and Kuroshio) using measurements from three different ocean observing networks. The same method is applied to each current, allowing us to compare variability in the three currents. Between the start of 2004 and end of 2019 we find that transport of water has decreased in the Kuroshio but has not changed in the Agulhas or East Australian Current. We find changes in the path of the Kuroshio on interannual time scales, and shorter and irregular changes in the path of the Agulhas. On annual time periods, all three currents transport more water in summer than winter, which is related to faster speeds during the summer. These long subsurface time series can help us better understand how WBC variability impacts society at the western boundaries of the ocean.
The spatial MLAT‐MLT distributions of the dayside cold (blue) and hot (red) patches superimposed on the statistical ionospheric convection patterns (electrostatic potentials in kV, gray lines) for eight 45°‐wide clock angles centered about θ = (a) −45°, (b) 0°, and (c) 45°, and so on. The yellow lines highlight the statistical poleward boundary of the auroral oval. The numbers of cold and hot patches are marked in the upper right corner of each panel.
Magnetic latitude (MLAT) versus clock angle distributions of (a1–a4) patch occurrence, (b1–b4) O⁺ density, (c1–c4) electron temperature (Te), (d1–d4) soft‐electron (<1 keV) energy flux, and (e1–e4) the cross‐track velocity (Vcross‐track). The two columns on the left correspond to cold patches, and the two columns on the right correspond to hot patches. The distributions are binned over 1° MLAT and 10° clock angle. The black/magenta lines in panels a1–a4 identify the MLAT with the highest patch occurrence versus clock angle (the magenta dashed lines in panels a1–a2 are overlaid from the magenta solid lines panels a3–a4 for easier comparison between hot and cold patches).
Occurrence of polar cap patches versus (a and c) anti‐sunward velocity and (b and d) soft‐electron energy flux. Panels a and b correspond to southward interplanetary magnetic field (IMF) (i.e., more negative velocity), and panels c and d correspond to northward IMF. Blue bars indicate cold patches, red bars indicate hot patches, and gray bars show the background. The dotted blue and red lines show the relative occurrence rates of the cold and hot patches, respectively.
The Mean MLAT of Cold Patches, Hot Patches, and the Poleward Auroral Boundary in the Dawnside and Duskside Sector for Each of the Eight Different IMF Orientations Presented in Figure 1
By using a database of 4,634 cold patches (high density and low electron temperature) and 4,700 hot patches (high density and high electron temperature) from Defense Meteorological Satellite Program F16 in 2005–2018 winter months (October–March), we present a statistical survey of the distributions of polar cap patches for different interplanetary magnetic field (IMF) orientations and ionospheric convection geometries. We investigate the dependence of cold and hot patches on local plasma transport and soft‐electron precipitation. Our results indicate that: in winter, (a) more cold and hot patches occur in the stronger anti‐sunward flow organized by different IMF orientations. (b) cold patches are frequent near the central polar cap, while hot patches are closer to the auroral oval. (c) enhanced anti‐sunward flow (E × B drift) mainly contributes to cold patch occurrence under Bz < 0, and soft‐electron precipitation contributes to hot patch occurrence both under southward and northward IMF.
Radiative and temperature response to Stratospheric Aerosol Injection and solar dimming. Shading: Annual mean differences in (top) short wave heating rates (K day⁻¹) and (bottom) in temperature (K), averaged over 2070–2089, between CESMsulfur and RCP8.5 experiments (left), and between CESMsolar and RCP8.5 experiments (right). Hatching indicates regions where the difference is not statistically significant (±2 standard errors). Contours show the RCP8.5 climatology for reference.
Stratospheric‐tropospheric dynamical response to Stratospheric Aerosol Injection and solar dimming. Shading: Monthly mean differences in zonal wind (ms⁻¹) at 60°S (a, b) and December‐to‐February mean (DJF) mean differences in zonal wind (ms⁻¹) (c, d), averaged over 2070–2089, between CESMsulfur and RCP8.5 (a, c), and between CESMsolar and RCP8.5 (b, d). Thick white line in top panels marks the regions where the response is statistically significant (±2 standard errors). Hatching in the bottom panels indicates regions where the difference is not statistically significant (±2 standard errors). Black contours in all panels show the corresponding RCP8.5 climatology for reference.
Impacts on the Southern Hemisphere (SH) tropospheric climate. (a) The DJF differences between 2,070 and 2,089 mean Southern Annular Mode index (see Section 2), and the location of the SH eddy‐driven jet, the edge of the SH Hadley Cell and the edge of the SH subtropical dry zone (PE) between each of the CESMsulfur (blue), CESMsolar (red), CESMsolar_grad (orange) experiments and RCP8.5. Whiskers indicate ±2 standard errors of the difference in means. (b) The DJF differences in precipitation (mm/day) between 2,070 and 2,089 CESMsulfur (blue), CESMsolar (red), CESMsolar_grad (orange) experiments and RCP8.5. The dashed black line (with scale on the right hand side) indicates climatological mean precipitation in RCP8.5 for reference. The analogous figure, but for the differences compared against BASE (i.e., 2,011–2,030 mean of RCP8.5) is shown in Figure S11 in Supporting Information S1.
Verification of the impacts on the Southern Hemisphere (SH) tropospheric climate in a multi‐model framework. The December‐to‐February mean (DJF) differences between 2080–2099 mean (a) Southern Annular Mode index (see Section 2), (b) the location of the SH eddy‐driven jet, (c) the location of the edge of the SH Hadley Cell and (d) the location of the edge of the SH subtropical dry zone (PE) between each of the GeoMIP G6sulfur (blue) and GeoMIP G6solar (red) experiments and SSP5‐8.5. Results are calculated separately for each model. Whiskers indicate ±2 standard errors of the difference in means. The analogous figure, but for the differences compared against the present‐day reference period are shown in Figure S12 in Supporting Information S1.
Modeling experiments reducing surface temperatures via an idealized reduction of the solar constant have often been used as analogs for Stratospheric Aerosol Injection (SAI), thereby implicitly assuming that solar dimming captures the essential physical mechanism through which SAI influences surface climate. While the omission of some important processes that otherwise operate under SAI was identified before, here we demonstrate that the imposed reduction in the incoming solar radiation also induces a different stratospheric dynamical response, manifested through a weakening of the polar vortex, that propagates from the upper stratosphere down to the troposphere. The coupled stratospheric‐tropospheric response exerts a previously overlooked first‐order influence on southern hemispheric surface climate in the solar dimming experiments, including on the position of the tropospheric jet and Hadley Circulation and thus, ultimately, precipitation patterns. This perturbation, opposite to that expected under SAI, highlights the need for caution when attributing responses in idealized experiments.
Stations distribution of the DONET1 network in southwestern Japan.
(a) Monthly power spectrum density (PSD) as a function of frequency for the X component of station KMA03. Colorful curves stand for the mean PSD amplitude for each 3 days. SFMs are highlighted by yellow bars. For reference, black dashed curves are the new low‐noise model (NLNM) and the new high‐noise model (NHNM). (b) The averaged PSD values of single‐frequency microseisms (SFMs) with depth of OBSs. Colorful straight lines are the linear regressions between the PSD of corresponding components and depth of ocean bottom seismometers (OBSs). (c)The PSD versus mean correlation coefficients (CCs) at each OBS.
Monthly back azimuth (BAZ) variations of Rayleigh‐wave single‐frequency microseisms (SFMs) at stations KMA03 and KMB06 throughout the study period. Every polarization diagram is characterized by its BAZs (indicated by the angle with respect to north) and its frequency (given by the radius), with intervals of frequency and azimuth bins of 0.01 Hz and 3° wide, respectively. Two white circles constrain the frequency band of SFMs.
3D topography of ocean bottom surrounding the ocean bottom seismometers. Pink circles and irregular rectangles stand for the possible source regions whose boundaries cannot be determined precisely.
Single‐frequency microseisms (SFMs) have been revealed to be only generated in shallow water for decades, while some recent studies reported them in a deep ocean. Using the continuous waveform data recorded by ocean bottom seismometers, we investigate the deep ocean SFMs in the northern Philippine Sea. Based on the spectrum analysis, we find that the SFMs can be detected in deep ocean and the detection is time variable. To determine the source locations of SFMs, we perform polarization analysis and calculate the cross‐correlation coefficients between SFMs on vertical components and ocean wave energy spectra considering the attenuation of SFMs. Both the polarization and correlation results show that the sources nearby stations dominate the observed Rayleigh‐wave SFMs though distant sources also contribute. Our investigation suggests that the SFMs can be generated in a deep ocean likely by infragravity ocean waves interacting with seafloor topography, which is strengthened by strong ocean storms.
Illustration of the method used to determine the inner core (IC) rotation. The basic data are South Sandwich Islands (SSI) doublets (repeating earthquakes, in stars) recorded at the Kyrgyzstan twin stations (AAK and KZA, in triangles). (a) Ray paths of seismic PKP waves (P waves through the core), including IC reflected PKiKP (CD branch, blue) and refracted PKIKP (DF branch, red) waves. (b) Surface projections of the raypaths. The red segment of the raypath represents the DF leg traversing the IC. (c and d) Cartoon illustration of the detection method. Different colors of the IC (center sphere) represent lateral velocity variations of the IC. A local IC structure (the red patch) is sampled by the path from an earlier (old) event to station AAK in (c). The same structure (the red patch) is subsequently captured by the path from a later (new) repeating event to station KZA east of AAK in (d), after an eastward IC rotation. Note that the scales for the separation between the two stations and the IC structure and the amount of the IC rotation are greatly exaggerated for the illustration purpose.
Example waveform comparisons from an South Sandwich Islands doublet (ID 99_07, with events in 1999 and 2007 and lapse of 7.2 years) to the twin stations (AAK and KZA). The waveforms are aligned with CD starting at reference time zero using cross‐correlation and are normalized by their peak‐to‐trough amplitudes. The cross‐correlation coefficient of DF at the zero lag and time shift of DF (ddt in seconds) from the best cross‐correlation is labeled in the parentheses, respectively. (a) Temporal change at AAK between the two events of the doublet. (b) Spatial lateral variation sampled by the new (later) event to the twin stations. (c) Waveform comparison between the old (earlier) event to station AAK and the new event to station KZA.
Similar to Figure 2, but for 8 South Sandwich Islands doublets. The pairs are in the order of increasing time lapse from top to bottom. The gray vertical bar marks the rough time separating the DF and CD arrivals. The waveforms in the purple box in (c) for the doublets of time lapses 6–10 years show relatively good matches, even though they are from different events and different stations. The measurements of the cross‐correlation coefficient at zero lag and the ddt value are labeled in the parenthesis of each panel (similar to Figure 2). When the cross‐correlation coefficient is negative, we set it to zero for better visualization.
Double difference time (ddt) measurements and estimations of the inner core (IC) rotation rate. (a) The ddt measurements as a function of the time lapse dT. The measured temporal and spatiotemporal ddt values are marked in Figure 3. Their values are close to zero after correcting for the preferred eastward IC rotation of 0.127 deg/year (see text). (b) Enlarged view of spatiotemporal ddt. As dT increases, the ddt goes from slightly positive to slightly negative values, crossing the zero line at about 6.2 years. This occurs when the two spatiotemporal records match exactly, which gives a direct estimate of the rotation rate (see text). (c and d) Two estimates are made, yielding similar results (labeled, ∼0.13°/year). The annual ddt in (c) is the temporal ddt divided by the dT of the corresponding doublet; the local gradient is the corrected spatial ddt divided by the longitudinal separation of the DF raypaths to the twin stations. In (c), the dashed line represents the linear regression (labeled) of the seven doublets with dT > 5 years (solid dots) with a zero intercept. The error is two standard deviations of the slope. The eastward rotation rate is equal to the slope with the opposite sign. The open circle is from the doublet 05_08 with a short lapse of 3.3 years. The two triangles near the origin are from two additional doublets with little temporal and spatial variations (Figure S12 in Supporting Information S1). In (d), a rotation rate is obtained for each of the seven doublets by dividing the annual temporal change by the local spatial gradient. The vertical bars show the mean ± two standard deviations (gray) and the mean ± two standard deviations of the mean (dark, labeled), respectively.
The Earth's solid inner core (IC) is generally believed to rotate relative to the mantle, but the proposal remains controversial. Here we use seven waveform doublets in the South Sandwich Islands region with time lapses of 5.8–17.0 years that are recorded by two close stations in Kyrgyzstan with virtually the same epicentral distance. The fortuitous geometry allows precise measurements of the IC temporal changes and the underlying local structure at the same time. The remarkable observations in waveforms and spatial‐temporal measurements show unequivocally that the IC must have shifted (rotated) eastward in 1991–2010 and help determine accurately the average rotation rate as 0.127 ± 0.006°/yr at 95% confidence level during the time span.
Scaling rates (% K⁻¹) were estimated using hourly precipitation (PPT) from (a) NOAA/NCDC observations with initial quality control (RAW‐DATA), (b) a quality‐controlled (GSDR‐QC) version of the Global Sub‐daily Rainfall (GSDR) data set (Lewis et al., 2021), and daily dewpoint temperature (DPT) from the HadISD data set (Dunn, 2019), (c) scaling difference (a–b; % K⁻¹) using RAW‐DATA and GSDR‐QC versions of data. Scaling is estimated using the binning method at the 99th percentile from the second to the second last bin (2 to 2last) for 2,905 gauges, which have at least 12 years of hourly precipitation data. The numbers below each panel show positive (red), super‐CC (brown) and negative (blue) median scaling rates respectively (numbers in parentheses show the percentage of locations with positive/super‐CC/negative scaling respectively). This figure and subsequent figures were plotted using the Generic Mapping Tool (GMT).
Scaling rates (% K⁻¹) were estimated using hourly precipitation (PPT) from the updated version of Global Sub‐daily Rainfall (GSDR) (V2; Lewis et al.(2021), under preparation) at (a) 2.54 mm precision and (b) 0.25 mm precision, (c) scaling difference (b–a; % K⁻¹) using 0.25 and 2.54 mm precision data. Scaling is estimated using the binning method at the 99th percentile from the second to the second last bin (2 to 2last) for 475 gauges that have at least 7 years of PPT data. The numbers below each panel show positive (red), super‐CC (brown) and negative (blue) median scaling respectively (numbers in parentheses show the percentage of locations with positive/super‐CC/negative scaling respectively).
Scaling rates (% K⁻¹) were estimated using hourly precipitation (PPT) from the Global Sub‐daily Rainfall (GSDR) data set (Lewis et al., 2021; 2.54 mm precision) and daily dewpoint temperature (DPT) from the HadISD data set (Dunn, 2019). Scaling is estimated using the second bin to the breakpoint (2 to BP) in the binning method (first row), second bin to the second last bin (2 to 2last) in the binning method (second row), quantile regression (QR; third row) and Zhang et al. (2017) method (ZM; fourth row) respectively. The scaling has been estimated at an individual location (1‐Station; first column), pooling three nearest locations (3‐Stations; second column), and pooling five nearest locations (5‐Stations; third column) respectively. The numbers below each panel show positive (red), super‐CC (brown) and negative (blue) median scaling rates respectively (numbers in parentheses show the percentage of gauges with positive/super‐CC/negative scaling respectively).
Short‐duration precipitation extremes (PE) increase at a rate of around 7%/K explained by the Clausius‐Clapeyron relationship. Previous studies show uncertainty in the extreme precipitation‐temperature relationship (scaling) due to various thermodynamic/dynamic factors. Here, we show that uncertainty may arise from the choice of data and methods. Using hourly precipitation (PPT) and daily dewpoint temperature (DPT) across 2,905 locations over the United States, we found higher scaling for quality‐controlled data, all locations showing positive (median 6.2%/K) scaling, as compared to raw data showing positive (median 5.3%/K) scaling over 97.5% of locations. We found higher scaling for higher measurement precision of PPT (0.25 mm: median 7.8%/K; 2.54 mm: median 6.6%/K). The method that removes seasonality in PPT and DPT gives higher (with seasonality: median 6.2%/K; without seasonality: median 7.2%/K) scaling. Our results demonstrate the importance of quality‐controlled, high‐precision observations and robust methods in estimating accurate scaling for a better understanding of PE change with warming.
(a) Map of biogeochemical (BGC) float data used in this study. Colored dots indicate Chl estimates (ChlKd) based on the diffuse attenuation coefficient (Kd(490)) between 10 and 40 dbar. (b–d) Results of spatiotemporal comparisons (n ∼ 150) between estimates of Kd(490) and ChlKd obtained from BGC‐Argo floats and from MODIS satellite ocean color (see Section 2), color‐coded by the day of the year (DOY), showing no temporal bias. Ocean color algorithms are MODIS OC3M and the Southern Ocean (SO) algorithm developed by R. Johnson et al. (2013). Also shown are a robust regression (blue), the 1:1 line (black), the %bias, and mean percentage error (MPE) relative to BGC‐Argo estimates.
F/Chl estimated from biogeochemical (BGC)‐Argo float data around the Kerguelen plateau (a and b) and measured in laboratory experiments with Fe‐replete and Fe‐limited phytoplankton (c and d). (a) Boxplot of data corresponding to the two boxes indicated in the map in panel (b): upstream (west) and downstream (east) of the Kerguelen plateau. (b) F/Chl estimates from BGC‐Argo floats overlain with currents from a near‐surface velocity climatology (Laurindo et al., 2017). (c) F/Chl of Phaeocystis antarctica as a function of Fe availability and irradiance; (d) F/Chl of five Southern Ocean diatom species: Proboscia inermis, Eucampia antarctica, Fragilariopsis cylindrus, Chaetoceros neogracilis and Chaetoceros flexuosus.
Exploring the relationship between F/Chl and other estimated and measured variables, with colors representing data density. The Fe‐limitation proxy αNPQ (a), NPQmax (b), and latitude (c) are all significantly correlated with F/Chl, with αNPQ the best‐ranked predictor (see Section 3.3). The relationship with Tmean (d) is not significant. Data correspond to those used in the statistical model (n = 1,734); full model results are given in Table S3 in Supporting Information S1.
Plain Language Summary Phytoplankton fluorescence is a relatively easy and consistent measurement that can be made in the ocean. The measured fluorescence stems directly from the chlorophyll contained in phytoplankton cells and is therefore very specific to the tiny plants. Great oceanic coverage of fluorescence measurements is achieved because fluorometers can be mounted on autonomous platforms such as gliders and floats, yielding reliable data for years without human intervention. But there is a problem with this measurement: fluorescence does not equal chlorophyll, and the factor for the conversion of fluorescence to chlorophyll varies almost 10‐fold across the world's oceans. The largest conversion factors, with the highest variability, are observed in the Southern Ocean. We know that things such as phytoplankton species composition, nutrient status, and acclimation to light all affect the conversion factor. Using data from biogeochemical Argo floats and from laboratory studies, we show in this study that iron limitation also significantly increases the fluorescence‐chlorophyll conversion factor, and that iron limitation may indeed be a key driver for the high conversion factors observed in the Southern Ocean.
(a and b) 630 nm All‐Sky imager data projected to 250 km altitude at t = 08:38 UT and t = 08:40 UT (color‐coded) and enhanced 1 s phase scintillation indices (black circles) during t ± 1 min. Payload trajectories (T2‐H: high‐flyer, T2‐L: low‐flyer) are shown in magenta, with thicker lines for t ± 1 min. (c and d) Super Dual Auroral Radar Network spectral width with trajectories (black lines) and payload location (black crosses).
(a) Ne and payload altitude. (b) Wavelet spectrogram of Ne fluctuations. (c) Eastward (EE) and Northward (EN) components of the E‐field. Periodic spikes occur when an antenna probe enters the rocket's shadow. (d) Wavelet spectrogram of ΔEE. (e) Time shifts Δτ of maximum cross‐correlation between currents fluctuations measured by two multi‐needle Langmuir probes. The red curve exhibits (median‐filtered) magnetic field fluctuations reflecting the payload spin.
(a and b) Magnitude squared coherence (Cxy) of relative multi‐needle Langmuir probe current fluctuations. (c and d) Relative phase angles θ of the current fluctuations. The black and magenta lines show θ obtained assuming Doppler shift due to vE×B and from a fit, respectively. (e and f) PSD of the current fluctuations. (g–l) Similar analysis for the E‐field. Filled dots are used for frequency components with Cxy > 0.8.
(a) Ne and BEast. (b) Velocities obtained using Cross‐spectral density analysis on multi‐needle Langmuir probe (magenta) and E‐field (blue/cyan) data for different probe pairs. (c) Local intermittency Analysis of mNLP3 current with the electron inertial length λe and λ ≈ 15 m superimposed. (d) Electron number flux parallel to B (color) and total number flux (black line, a.u.). (e) Omnidirectional ion number flux. Times from Figure 1 are annotated at the top.
Plain Language Summary Ionospheric plasma are known to be highly irregular, with fluctuations evolving both in space and time. Irregular structures can reach hundreds of kilometers to a few meters and, despite being common and having space weather impacts, the details of their source(s) and behavior are still unclear, especially at smaller scales. In this work, we investigate small‐scale plasma density and electric field fluctuations observed by a sounding rocket where ground‐based instruments also detected irregularities. To circumvent ambiguities of interpreting measurements made by single probes, we take advantage of the fact that the fluctuations were detected by spatially separated probes and use multi‐point analysis techniques to separate the spatial and temporal scales of the observed structures. The analysis allows to estimate the phase velocities and wavelengths of the fluctuations and reveals spatial irregularities from tens of meters to a meter, that is, irregularities that are slow in the plasma frame. Additionally, we show that these small‐scale structures are concentrated outside of regions where most electrons are precipitating downward along the Earth's magnetic field and discuss the observations in the context of irregularity creation. Altogether, this study provides new insights into the sources and behavior of high‐latitude ionospheric irregularities.
Ratios of surface to bottom water concentration of N2 and N2O (a), and variation in the ratios (b), DIN concentrations (c), river depth (d), SPS concentration (e), and SW/SB (f) with stream order. The gray dots in c represent urban sites that were excluded from the analysis. The solid red lines in c‐e represent the fit of power or exponential regressions through observed data. SW/SB in f represent the ratio of SPS‐water contact area (SW) to sediment‐water contact area (SB, details in Supporting information S1), and the solid red line represents SW/SB = 1. Boxes in a and f are bounded by the 25th (Q1) and 75th percentiles (Q3), and whiskers represent 1.5× the interquartile range (IQR), and the solid line is the median value. The values greater than Q3 + 1.5 × IQR or Q1 − 1.5 × IQR are identified as outliers. The ratios of surface to bottom water concentration, DIN concentration, depth, and SPS concentration in b to e are the mean values of the sampling sites for each stream order. All error bars show mean ± 1 s.e.m.
Relationship between stream order and water‐air fluxes, sediment‐water fluxes, and water column production rates of N2 and N2O, as well as water column contributions. The blue and red line in (a) represents the fit of a power function, and the green line represents the fit of an exponential function. The red solid line in (b) represents the fit of an exponential function, and the red dash line represents water column contributions equal to 50%, above which water column is dominant for production of N2 and N2O. Data points in (a) and (b) are the mean for each stream order, and error bars are +1 s.e.m in (a), and ±1 s.e.m in (b).
Seasonal and spatial variations of the water column contributions to N2 (a) and N2O (b) fluxes. Boxes are bounded by the 25th and 75th percentiles, and whiskers represent 1.5× the interquartile range, and the solid line is the median value. The black dots are arithmetic means, and the gray dots are outliers that are greater than Q3 + 1.5 × IQR or Q1 − 1.5 × IQR. The red dash lines represent water column contributions equal to 50%.
A conceptual framework for assessing the relative role of different zones in riverine N processing along a theoretical stream‐river continuum. N2 and N2O emissions decrease with river size and shift from benthic dominance to water column‐dominance.
Plain Language Summary The entire global N budget remains out of balance, with total N inputs exceeding N losses. Streams and rivers serve as substantial recipients and processors of reactive N transported from terrestrial landscapes. The benthic zone is traditionally identified as a hotspot for N processing in fluvial systems. However, the role of the water column is poorly understood. Here, we found that the water column area‐basis production rates of N2 and N2O increased with stream order although volumetric‐basis production rates did not change significantly with stream order through 4‐year observations across six river networks in China. The water column contribution increased with stream order and became dominant in large rivers. The increase in the contact area of SPS‐water caused by higher SPS concentrations and water volume accounted for the shift as river size increased. The current estimates would underestimate riverine N removal and N2O emissions by approximately 50% if neglecting water column processes based on the upscaling results for the six large river networks ranging from first to eighth order. Thus, our findings provide insight into the understanding of riverine N dynamics and highlight the important role of water column processes in N upscaling for closing regional and global N budgets.
Fast blue discharge observed by ASIM. No present signature in the red or UV photometer. (a) The photometer data for the event showing the rise time of 30 μs and a duration of 1.2 ms (b) The photometer data in log scale showing the photometer data for the full camera frame (83 ms). (c) Satelite data for the cloud top temperature and the event location showing that it occurs close to the coldest part of the cloud cell. The white cross determines the location of the event while the circle is the location uncertainty, black points are GLD360 flashes close in time to the event (±2 min). (d) The 337 nm camera image for the event with the CAPE of the event.
A slow blue discharge with a rise time of 250 μs and a duration of 2.02 ms. The cloud top temperature shows the event originating from the coldest part of the cloud and the rise time suggests an event deeper within the cloud. Panels and legend as Figure 1.
The global distribution of blue discharges detected by ASIM normalized by observation time with smoothing, see Figure S5 in Supporting Information S1 for unsmoothed distribution. (a) The global distribution in 2° × 2° bins in the region covered by the ISS. (b–e) Zoom to selected regions showing events with black markers.
Rise time and duration of all blue discharge observed in the 3 years (1 July 2018–1 July 2021) interval with binsizes of 10 and 50 μs, respectively. Red colored bins indicate a fast blue discharge with rise time less than 40 μs and green a rise time greater than 40 μs. (a) The rise times for the blue discharges, with medians for the two distributions at 20 and 170 μs. (b) Distribution of the duration of blue discharges, resulting in a median of 1 ms. (c) The calculated cloud depth of the event sources given the detected rise times, the medians of the two distributions have cloud depth of 1.4 and 3.2 km respectively. (d) The cloud depth calculated from the camera image using a subset of the available events showing a peak for the fast events at 0.8 km, binsize is 200 m. The two peaks indicate two separate types of events, one at cloud tops close to the tropopause with a fast rise time and another originating deeper into the cloud system with a correspondingly lower rise time.
Cumulative distribution of CAPE and CTT difference of the blue discharges compared to regular lightning detected by MMIA normalized by total number of events. Individual blue discharge observations are clustered into cells via density clustering and considered coming from a cell generating fast blue discharges if there is at least one fast blue discharge in its cell. Of the cells generating fast blue discharges, 50% are overshooting tops, compared to 34% of the cells only with slow discharges. Cells containing only lightning consists of 21% overshooting tops. About 25% of cells with fast discharges occur during deep convection (CAPE greater than 2000 Jkg⁻¹), compared to 17% of cells only with slow discharges and 10% of lightning cells.
Blue electric streamer discharges in the upper reaches of thunderclouds are observed as flashes of 337.0 nm (blue) with faint or no emissions of 777.4 nm (red). Analyzing 3 years of measurements by the Atmosphere‐Space Interactions Monitor on the International Space Station, we find that their distribution in rise time falls into two categories. One with fast rise times of 30 μs or less that are relatively unaffected by cloud scattering and emanate from within ∼2 km of the cloud tops, and another with longer rise times from deeper within the clouds. 50% of cells generating shallow events are associated with overshooting tops compared to 34% of cells generating deeper events. The median Convective Available Potential Energy of the cells is ∼70% higher for the shallow events and ∼38% higher for the deeper events than for lightning cells, suggesting the discharges are favored by strongly convective environments.
A two‐layer mass balance model of ²²⁶Ra for the global ocean. The numbers represent the mean fluxes of each term in 10¹⁶ dpm/yr. The probability distribution for each term is shown in Figure S1 in Supporting Information S1.
Particle scavenging contribution as a percent of the total sink terms versus water residence time (in days). PS% for ²²⁶Ra (a) and ²²⁸Ra (b) with different scavenging coefficient k values ranging from 0.02 to 2 yr⁻¹.
Radium isotopes are powerful proxies in oceanography and hydrology. Radium mass balance models, including assessments of submarine groundwater discharge (SGD), often overlook particle scavenging (PS) as a pathway for dissolved radium removal from the world ocean. Here, we build a global ocean ²²⁶Ra mass balance model and reevaluate the potential importance of PS. We find that PS is the major ²²⁶Ra sink for the upper ocean, removing about 96% of the total input from various sources. Aside from vertical exchange with the lower ocean, SGD is the largest ²²⁶Ra source into the upper ocean. The biological pump transfers particles to the deep ocean, resulting in a major but often overlooked impact on the global ²²⁶Ra marine budget. Our findings suggest that radium mass balance models should consider PS in systems with high siliceous algae production and export fluxes and long water residence times to prevent underestimation of large‐scale SGD fluxes.
Trophic amplification due to seasonally varying trophic transfer efficiency (TTE): (a) Seasonal cycles of net phytoplankton (NPphy, green), mesozooplankton production (NPlzoo, purple) and FCE (NPlzoo/NPphy, yellow). The y axis range spans 100% of change relative to the annual mean for each variable; (b) seasonal cycles of the terms in Equation 8 to assess the relative contributions from trophic transfer efficiency between trophic levels 1 and 2 (TTE1, light blue) and trophic levels 2 and 3 (TTE2, dark blue), and food chain length (FCL; as 1/(FCL − 2), red) over the water column within the focus region. The y axis range spans 50% of change relative to the annual mean for each variable.
Reduced trophic transfer efficiency in winter (TTE) due to reduced efficiency of grazing: Fate of phytoplankton production in austral (a) summer; and (b) winter. The sizes of the pie charts are representative for the magnitude of phytoplankton production. (c) Biomass‐specific rate of respiration (purple), fecal pellet production (blue) and production available to the next trophic level (red) in austral summer and winter. The sum of all colored components (respiration, fecal pellets and the production available to the next trophic level) represents zooplankton grazing; only the grazed (colored) parts are being processed by zooplankton.
Reduced efficiency of grazing (specific grazing rate) due to prey dilution in deep winter mixed layers: (a) Vertical profiles of predator biomass specific grazing rate (d⁻¹) in February and June with color indicating the prey concentration (TL1, mmol N m⁻³) and size of the circle indicating the predator concentration (TL2, mmol N m⁻³) within the focus region; inserts show prey vertically integrated over the water column (green), the mixed layer depth (yellow), and the average mixed layer biomass specific ingestion rate (light blue); (b) Correlation of mixed layer depth and vertical encounter efficiency (EE, Equation 9) with colors indicating the time of the year (months). R² value of the correlation is shown on the left side.
Indication for seasonal mixed layer depth (MLD) variations driving food chain efficiency (FCE) beyond the Humboldt system: Ocean biomes calculated from (a) observations, (b) UVic‐model and (c) GFDL‐model with productive (green, above 0.1 mg Chl m⁻³ (observations)/0.15 mg Chl m⁻³ (model)) and oligotrophic (yellow, below 0.1 mg Chl m⁻³ (observations)/0.15 mg Chl m⁻³ (model)) regions. Dots in (a) indicate the locations of observations. Average seasonal cycles of MLD (black) and food chain efficiency (FCE, the ratio of net mesozooplankton production to phytoplankton production, Equation 6, color) for (d and g) observations (e and h) UVic‐model and (f and i) GFDL‐model, normalized by the mean MLD and FCE for each biome: productive (green, (d to f) and oligotrophic (yellow, g to i) regions. Values in brackets indicate the absolute values of annual means. R² values of the correlations between MLD and FCE are shown in the top‐left side of each panel.
Plain Language Summary The Humboldt Upwelling System is a fishery‐important region. A common assumption is that a certain amount of phytoplankton supports a proportional amount of fish. However, we find that a small seasonal change in phytoplankton can trigger a larger variation in zooplankton. This implies that one may underestimate changes in fish solely based on phytoplankton. Using ecosystem model simulations, we investigate why changes of phytoplankton are not proportionally reflected in zooplankton. The portion of phytoplankton that ends up in zooplankton is controlled by the changing depth of the surface ocean “mixed layer.” The “mixed layer” traps both the phytoplankton and zooplankton in a limited amount of space. When the “mixed layer” is shallow, zooplankton can feed more efficiently on phytoplankton as both are compressed in a comparatively smaller space. We conclude that in the Humboldt System, and other “food‐rich” regions, feeding efficiently, determined by the “mixed layer,” is more important than how much food is available.
Top-cited authors
E. Rignot
  • University of California, Irvine
John Gosling
  • University of Colorado Boulder
Michiel Roland Van den Broeke
Andre Balogh
  • Imperial College London
David Mccomas
  • Southwest Research Institute