Project

Cross-timescale Interactions, Diagnostics and Skill

Goal: + To analyze cross-timescale interference between climate drivers acting at multiple time scales, and their impact on climate hazard predictability.
+ To define and apply a seamless diagnostic framework comparing modeled and observed weather type's statistics, spatial patterns and physical links to climate drivers across timescales.
+ To assess predictive skill of forecast systems at multiple timescales

Methods: Weather Types, Nonparametric Statistics, Climate Modeling

Updates
0 new
8
Recommendations
0 new
1
Followers
0 new
29
Reads
0 new
306

Project log

Ángel G Muñoz
added a research item
Recent research has highlighted the potential for improving predictive skill at the subseasonal timescale, which could be the basis for enhanced, actionable forecasts for climate services involving water and disaster management, health, energy and food security. Projects such as WMOʼs World Weather and World Climate Research Programmeʼs Subseasonal-to- Seasonal Prediction Project (S2S) and NOAAʼs SubX have made available extensive databases with both hindcasts and almost-realtime forecast at this timescale. Lead times are long enough that much of the information in the atmospheric initial conditions is lost, but at the same time are too short for other sources of predictability (e.g., ocean boundary conditions) to have a strong in-fluence in skill. Presently, sub-seasonal skill is still limited beyond 2-3 weeks, and in general uncalibrated forecasts cannot be used to develop climate services. An obvious alternative is to make use of a variety of robust statistical calibration methods ‒also known as Model Output Statistics, MOS‒ available for other timescales, such as the seasonal one. Nonetheless, different methods have different advantages and disadvantages, depending on which forecast attribute to focus on. Here, as a benchmark, we first analyze the spatiotemporal variability of predictive skill in uncalibrated models, and then discuss how local (gridbox-by-gridbox) and non-local (pattern-based) calibration models enhance or decrease skill in different regions of the world.
Ángel G Muñoz
added a research item
Common approaches to diagnose systematic model errors involve the computation of statistical metrics aimed at providing an overall summary of the performance of the model in reproducing the particular variables of interest in the study, normally tied to speciWc spatial and temporal scales. However, the evaluation of model performance is not always tied to the understanding of the physical processes that are correctly represented, distorted or even absent in the model world. As the physical mechanisms are more often than not related to interactions taking place at multiple time and spatial scales, cross-scale model diagnostic tools are not only desirable but required. Here, a recently proposed circulation-based diagnostic framework is extended to consider systematic errors in both spatial and temporal patterns at multiple timescales. The proposed framework, which uses a weather-typing-or ]ow-dependent-dynamical approach, quantiWes spatial biases in the magnitudes, location and tilt of modeled atmospheric circulation patterns, as well as biases associated with their temporal characteristics, such as frequency of occurrence, duration, persistence and transitions. Relationships between these biases and climate teleconnections (e.g., SST patterns, ENSO and MJO) are explored using different models. Some concrete applications are discussed.
Ángel G Muñoz
added 2 research items
K-means cluster analysis of wintertime 500-hPa geopotential height anomalies allowed identifying seven weather regimes (WRs) describing the atmospheric variability over the Euro-Mediterranean domain. The study of transitions between those WRs provided consistent results with the westward displacement of the blocking nearby northern Europe before the onset of the negative phase of the North Atlantic Oscillation (NAO-). The onset of the latter is, indeed, preceded by the North Atlantic blocking regime (NABl). In addition, we detected a preferred transition from the Scandinavian Blocking (ScBl) to NAO+ through the European Ridge regime (EuRG), which is modulated by active phases of the MJO. The examination of the relationship between WRs and precipitation over Morocco showed that the NAO- (NAO+) regime is accompanied by more (less) rainy episodes. The investigation of the lagged relationships between the MJO and the WRs depicted the role of an active MJO in phase 2 as a precursor of the ScBl and of an active MJO in phase 6 as a precursor of the NABl. The exploration of the 10-15 days lagged impact of the MJO on Moroccan rainfall showed an increase (decrease) of wet (dry) conditions 10-to-15 days after the occurrence of an active MJO in phases 6 and 8 (phases 2-3-4). The MJO modulation of the WRs and rainfall patterns over Morocco constitutes an important source of predictability at the medium- and the extended-range (subseasonal) time scales, with potential use by decision makers in key socio economic sectors in the region.
In the original version of the article contained errors in the fig 3 and 4 and captions of Figures 3 and 4 with respect to the panels in the top row.
Ángel G Muñoz
added an update
Just published in JGR. Key Points: 
  • K-means analysis of 500-hpa geopotential height anomalies over the Euro- Mediterranean region was used to identify seven weather regimes using a new information criterion. 
  • The link found between these weather regimes and rainfall in Morocco suggests that a NAO- (NAO+) state is associated with more (less) rainfall. 
  • The MJO favours a significant enhancement (reduction) of the probability of wet (dry) conditions 10-to-15 days after the occurrence of an active MJO in phases 6 and 8 (phases 2-3-4).  Perhaps of interest is that if one uses the traditional 4 Euro-Mediterranean Regimes, it is not possible to detect an important (not reported until now) ScandinavianBlocking > EuropeanRidge > NAO+ transition, consistent with a south-western propagation of the positive height anomalies, that can actually be characterized once one uses the K-means solution proposed in this paper. 
 
Ángel G Muñoz
added a research item
Cross-timescale interference involves linear and non-linear interactions between climate modes acting at multiple timescales (Muñoz et al., 2015, 2016, 2017; Robertson et al., 2015; Moron et al., 2015), and that are related to windows of opportunity for enhanced predictive skill (Mariotti et al., 2020), with relevant societal impacts (e.g., Doss-Gollin et al., 2018; Anderson et al., 2020). Using a simple mathematical model, reanalysis data and gridded observations, here we analyze plausible mechanisms for cross-timescale interference, describing conditions for coupling of oscillating modes and its impact on extreme rainfall occurrence and predictive skill. Concrete examples for Northeast North America and southern South America are discussed, as well as implications for climate model diagnostics.
Ángel G Muñoz
added a research item
While many Madden–Julian Oscillation (MJO) teleconnections are well documented, the significance of these teleconnections to agriculture is not well understood. Here we analyze how the MJO affects the climate during crop flowering seasons, when crops are particularly vulnerable to abiotic stress. Because the MJO is located in the tropics of the summer hemisphere and maize is a tropical, summer-grown crop, the MJO teleconnections to maize flowering seasons are stronger and more coherent than those to wheat, which tends to be grown in midlatitudes and flowers during the spring. The MJO significantly affects not only daily average precipitation and soil moisture, but also the probability of extreme precipitation, soil moisture and maximum temperatures during crop flowering seasons. The average influence on the probability of extreme daily precipitation, soil moisture, and maximum temperature events is roughly equal. On average the MJO modifies the probability of a 5th or 95th, 10th or 90th, and 25th or 75th percentile event by ∼ 2.5%, ∼ 4% and ∼ 7%, respectively. This means that an exceptionally dry (10th percentile) soil moisture value, for example, would become ∼ 40% more common (happening 14% of the time) during certain MJO phases. That the MJO can simultaneously dry soils and raise maximum air temperatures may be particularly damaging to crops because without available soil water during times of heat stress, plants are unable to transpire to cool leaf-level temperatures as a means of avoiding long-term damage. As a result, even though teleconnections from the MJO last only a few days to a week, they likely affect crop growth.
Ángel G Muñoz
added a research item
Producing probabilistic subseasonal forecasts of extreme events up to six weeks in advance is crucial for many economic sectors. In agribusiness, this time-scale is particularly critical because it allows for mitigation strategies to be adopted for counteracting weather hazards and taking advantage of opportunities. For example, spring frosts are detrimental for many nut trees, resulting in dramatic losses at harvest time. To explore subseasonal forecast quality in boreal spring, identified as one of the most sensitive times of the year by agribusiness end-users, we build a multi-system ensemble using four models involved in the Subseasonal-to-Seasonal (S2S) Prediction Project. Two-meter temperature forecasts are used to analyze cold spell predictions in the coastal Black Sea region, an area that is a global leader in the production of hazelnuts. When analyzed at global scale, the multi-system ensemble probabilistic forecasts for near-surface temperature is better than climatological values for several regions, especially the Tropics, even many weeks in advance; however, in coastal Black Sea skill is low after the second forecast week. When cold spells are predicted instead of near-surface temperatures, skill improves for the region, and the forecasts prove to contain potentially useful information to stakeholders willing to put mitigation plans into effect. Using a cost-loss model approach for the first time in this context, we show that there is added value of having such a forecast system instead of a business-as-usual strategy, not only for predictions released one to two weeks ahead of the extreme event, but also at longer lead-times.
Ángel G Muñoz
added an update
New research led by Stefano Materia shows potential value of multi-model sub seasonal forecasts for agribusiness sector in Turkey
 
Ángel G Muñoz
added a research item
Successful climate services often involve the use of tailored regional climate forecasts at one or multiple timescales. The way those forecasts are implemented is not always straightforward, and depends on several different factors, like which variables, models and calibration methods to use, how to produce the ensemble and tailoring, or even how to present them to the decision makers. Here, NextGen, a systematic general objective approach for designing, calibrating, building an ensemble of, and verifying objective climate forecasts is presented and discussed. NextGen involves the identification of decision-relevant variables by the stakeholders, and the analysis of the physical mechanisms, sources of predictability and suitable candidate predictors (in models and observations) for those key relevant variables. In those cases when prediction skill is deemed high enough, NextGen helps select the best dynamical models for the region of interest through a process-based evaluation, and automates the generation and verification of tailored multi-model, statistically calibrated predictions at seasonal and sub-seasonal timescales, at regional, national or sub-national level.
James Doss-Gollin
added a research item
During the austral summer 2015/16, severe flooding displaced over 170 000 people on the Paraguay River system in Paraguay, Argentina, and southern Brazil. These floods were driven by repeated heavy rainfall events in the lower Paraguay River basin. Alternating sequences of enhanced moisture inflow from the South American low-level jet and local convergence associated with baroclinic systems were conducive to mesoscale convective activity and enhanced precipitation. These circulation patterns were favored by cross-time-scale interactions of a very strong El Niño event, an unusually persistent Madden–Julian oscillation in phases 4 and 5, and the presence of a dipole SST anomaly in the central southern Atlantic Ocean. The simultaneous use of seasonal and subseasonal heavy rainfall predictions could have provided decision-makers with useful information about the start of these flooding events from two to four weeks in advance. Probabilistic seasonal forecasts available at the beginning of November successfully indicated heightened probability of heavy rainfall (90th percentile) over southern Paraguay and Brazil for December–February. Raw subseasonal forecasts of heavy rainfall exhibited limited skill at lead times beyond the first two predicted weeks, but a model output statistics approach involving principal component regression substantially improved the spatial distribution of skill for week 3 relative to other methods tested, including extended logistic regressions. A continuous monitoring of climate drivers impacting rainfall in the region, and the use of statistically corrected heavy precipitation seasonal and subseasonal forecasts, may help improve flood preparedness in this and other regions.
Ángel G Muñoz
added 4 research items
Python interface for IRI’s Climate Predictability Tool (CPT), a widely used research and application Model Output Statistics/Prediction toolbox. Publicly available: GitHub. Automatically downloads required observations (TRMM, CPC Unified) and S2S model data from the IRI Data Library (S2S Database and SubX –ECMWF, CFSv2, GEFS, others are being included). Computes climatologies, anomalies, a variety of skill metrics (uncalibrated and CCA-calibrated hindcasts) and probabilistic sub-seasonal forecasts.
Chapter 4 describes the basic components of weather and climate, and a common theme throughout is that there is considerable variability in space and in time. In this chapter, we start by describing and explaining how climate varies by location, by considering the effects of altitude, latitude and other aspects of geography on the climate. We then examine how climate varies over time, starting with differences between night and day, describing the seasons and how they are affected by location, and then describing how and why climate varies from year-to-year and at even longer timescales. Based on what we have learned in Chapters 1–4 the connection of climate to the spatial and temporal risk of infectious diseases, malnutrition or hydro-meteorological disasters can now be made.
Weather and climate vary on multiple timescales (§§ 3.2 and §§ 5.3), and climate information (historical, current or future) must target the specific time and space scales of the decisions being made. Observed climate is the result of the interaction of natural climate variability and the anthropogenic climate-change signal associated with increasing greenhouse gas emissions (§ 5.4.2). Today’s climate is dominated by natural variability, but the climate-change signal is already emerging and is expected to strengthen as concentrations of greenhouse gases in the atmosphere increase. However, gradual, long-term trends in climate are not the means by which people will experience most aspects of climate change. Instead, impacts will be felt primarily through changes in the weather (including extreme events like heat and cold waves and extreme rainfall), the seasons and potentially through alterations to longer-term components of climate variability, such as the El Niño – Southern Oscillation (see Box 5.1). The predictability of all these different timescales varies by location and period under consideration (see Chapters 7 and 8).
Ángel G Muñoz
added a research item
After successful eradication measures in the 1950s, the East African Highlands observed an unexpected resurgence of malaria in the 1990s. Although the epidemic episodes decreased again during the mid-2000s, they caused high mortality and morbidity rates in the population for over a decade. In addition to changes in herd immunity and control operations, climate drivers including anthropogenic global warming, El Niño-Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD) have been suggested to have independently played a role in the re-emergence of malaria in the Highlands. To date there is still debate on which climate drivers were involved, and if malaria transmission could be enhanced again by similar conditions in the future. Here we show that a combination of climate signals acting at multiple timescales, involving global warming, the Pacific Decadal Oscillation (PDO) and the IOD, affected malaria resurgence in the 1990s. We found that the PDO set the stage for the malaria re-emergence during ~1988-1995, impinging a positive trend that was then supported by the IOD during ~1995-2000, all embedded in the global warming background signal. Moreover, the observed malaria cases in the region cannot be explained by the isolated effect of any of these climate drivers, but only via a superposition of their signals. This particular combination is not always present, but its monitoring could help predict similar epidemic episodes in the near future. This analysis underscores the importance of climate and health scientists working together to elucidate climate and malaria interactions.
Ángel G Muñoz
added a research item
The North and South American Monsoon Systems (NAMS and SAMS, respectively) encompass rainfall patterns produced by similar dynamical mechanisms that involve ocean-land-atmosphere processes at multiple timescales. It has even been proposed that there is a cross-equatorial link between both monsoon systems, that is part of the seasonal transition of the large-scale tropical overturning circulation. Using a weather-typing approach to cluster cross-equatorial, meridional Convective Available Potential Energy fluxes, three distinctive but linked atmospheric circulation patterns are found, consistent with NAMS, SAMS and a transition regime. The modulation of the onset, duration and demise of these circulation regimes and their associated rainfall patterns is analyzed, and common sources of predictability are explored. A combination of the frequency of occurrence of the circulation regimes and its temporal derivative could be used as candidate predictors for monsoon characteristics. Our results (a) support the idea of a unified view of the American Monsoon System, (b) suggest that a subseasonal-to-seasonal forecast system could be implemented to predict the timing of onset and demise of the rainfall season(s), and (c) suggest process-based diagnostics to inform climate models improvement.
Ángel G Muñoz
added a research item
During the austral summer 2015/16, severe flooding displaced over 170 000 people on the Paraguay River system in Paraguay, Argentina, and southern Brazil. These floods were driven by repeated heavy rainfall events in the lower Paraguay River basin. Alternating sequences of enhanced moisture inflow from the South American low-level jet and local convergence associated with baroclinic systems were conducive to mesoscale convective activity and enhanced precipitation. These circulation patterns were favored by cross-time-scale interactions of a very strong El Niño event, an unusually persistentMadden–Julian oscillation in phases 4 and 5, and the presence of a dipole SST anomaly in the central southern Atlantic Ocean. The simultaneous use of seasonal and subseasonal heavy rainfall predictions could have provided decision-makers with useful information about the start of these flooding events from two to four weeks in advance. Probabilistic seasonal forecasts available at the beginning of November successfully indicated heightened probability of heavy rainfall (90th percentile) over southern Paraguay and Brazil for December–February. Raw subseasonal forecasts of heavy rainfall exhibited limited skill at lead times beyond the first two predicted weeks, but amodel output statistics approach involving principal component regression substantially improved the spatial distribution of skill for week 3 relative to other methods tested, including extended logistic regressions. A continuous monitoring of climate drivers impacting rainfall in the region, and the use of statistically corrected heavy precipitation seasonal and subseasonal forecasts, may help improve flood preparedness in this and other regions.
Ángel G Muñoz
added an update
In this new paper, we discuss some biases in the representation of the Caribbean Low-Level Jet (CLLJ) by low- and high-resolution models of the Geophysical Fluid Dynamics Laboratory (GFDL). Since the CLLJ is associated with key rainfall characteristics in Central America, northern South America and the Caribbean --the so-called Intra-Americas region, we also discuss biases in the modeled rainfall field.
We found that un-coupled models tend to outperform coupled ones in the Intra-Americas region, and discuss problems with the sea-surface temperature representation of those couple models and related physical mechanisms. In addition, we show that Model Output Statistics (MOS) dramatically improves forecast skill, the improvement being region-dependent. Furthermore, overall, no significant difference in skill is obtained when using low or high-resolution models.
 
Ángel G Muñoz
added a research item
The Caribbean low-level jet (CLLJ) is an important component of the atmospheric circulation over the Intra-Americas Sea (IAS) which impacts the weather and climate both locally and remotely. It influences the rainfall variability in the Caribbean, Central America, northern South America, the tropical Pacific and the continental Unites States through the transport of moisture. We make use of high-resolution coupled and uncoupled models from the Geophysical Fluid Dynamics Laboratory (GFDL) to investigate the simulation of the CLLJ and its teleconnections and further compare with low-resolution models. The high-resolution coupled model FLOR shows improvements in the simulation of the CLLJ and its teleconnections with rainfall and SST over the IAS compared to the low-resolution coupled model CM2.1. The CLLJ is better represented in uncoupled models (AM2.1 and AM2.5) forced with observed sea-surface temperatures (SSTs), emphasizing the role of SSTs in the simulation of the CLLJ. Further, we determine the forecast skill for observed rainfall using both high- and low-resolution predictions of rainfall and SSTs for the July–August–September season. We determine the role of statistical correction of model biases, coupling and horizontal resolution on the forecast skill. Statistical correction dramatically improves area-averaged forecast skill. But the analysis of spatial distribution in skill indicates that the improvement in skill after statistical correction is region dependent. Forecast skill is sensitive to coupling in parts of the Caribbean, Central and northern South America, and it is mostly insensitive over North America. Comparison of forecast skill between high and low-resolution coupled models does not show any dramatic difference. However, uncoupled models show improvement in the area-averaged skill in the high-resolution atmospheric model compared to lower resolution model. Understanding and improving the forecast skill over the IAS has important implications for highly vulnerable nations in the region.
Ángel G Muñoz
added 2 research items
This study explores the impact of El Niño and La Niña events on precipitation and circulation in East Asia. The results are based on statistical analysis of various observational datasets and Geophysical Fluid Dynamics Laboratory’s (GFDL’s) global climate model experiments. Multiple observational datasets and certain models show that in the southeastern coast of China, precipitation exhibits a nonlinear response to Central Pacific sea surface temperature anomalies during boreal deep fall/early winter. Higher mean rainfall is observed during both El Niño and La Niña events compared to the ENSO-Neutral phase, by an amount of approximately 0.4–0.5 mm/day on average per oC change. We argue that, in October to December, while the precipitation increases during El Niño are the result of anomalous onshore moisture fluxes, those during La Niña are driven by the persistence of terrestrial moisture anomalies resulting from earlier excess rainfall in this region. This is consistent with the nonlinear extreme rainfall behavior in coastal southeastern China, which increases during both ENSO phases and becomes more severe during El Niño than La Niña events.
Ángel G Muñoz
added an update
Journal of Climate just accepted our paper on "Heavy rainfall in Paraguay during the 2015-2016 austral summer: causes and sub-seasonal-to-seasonal predictive skill", led by James Doss-Gollin
We discuss the physical mechanisms behind a set of heavy rainfall events in southeastern South America, which turned out to be related to a cross-timescale interference between ENSO, the South Central Atlantic Dipole (SCAD) and MJO, impacting the occurrence and frequency of atmospheric circulation regimes conducive to --among others-- more "No-Chaco" South American Low Level Jet events, and thus heavy precipitation over Paraguay.
We then analyze if those events could have been skillfully forecast, using both the IRI seasonal "flexible forecast system" and ECMWF sub-seasonal forecasts. We found that IRI's flexible forecast system, which provides exceedance probabilities for a given threshold, successfully predicted above normal wet conditions for the trimester under study, although with spatial biases that suggest that models are not capturing well the physical interactions between the Pacific and the Atlantic.
The uncalibrated ECMWF sub-seasonal forecasts misplaced the heavy rainfall in the region, but pattern-based Model Output Statistics (e.g., Canonical Correlation Analysis) greatly improves the overall predictive skill of this kind of events.
 
Ángel G Muñoz
added 3 research items
Traditional dynamical downscaling approaches in the climate change context use coupled general circulation model (CGCM) outputs to provide realizations of the expected future changes in variables such as temperature and precipitation for a specific region. The coarse resolution CGCMs do not resolve weather transients or the physical mechanisms related to extreme precipitation events. Downscaled climate models may be able to capture some of the higher resolution processes (e.g., weather transients) and may even represent how these are impacted by large-scale climate. However, the regional models cannot correct most errors and biases in the large-scale climate fields. Both global and regional climate models may be able to provide some useful information for current and future climate, but it is increasingly clear that neither global or regional climate model outputs alone are adequate to directly drive impacts models, such as crop models. Here we explore a different approach to the provision of multi-scale climate information for agricultural planning wherein the weather and climate information inferred by the models is interpreted through the observations to inform changes to weather characteristics for use by crop models. To explore this methodology for Southeastern South America, we start with a decomposition of the climate time series into inter-annual, decadal and long-term signal components for the NCAR-CCSM4 global model and NCEP/NCAR Re-Analysis. The first result is that the global model does not capture the magnitude of the wetting trend that has been observed, and the dynamical downscaling does not improve that simulation. We then examine changes in the weather characteristics and interannual-to-decadal climate in the observations, NCAR-CCSM4, and NCAR-WRF regional model for 1981-1990, 2001-2010 and 2021-2030. We find that global and regional simulations adequately represent the statistical characteristics of decadal-scale rainfall when compared to observations. We also find that the models represent well the regional rainfall response to large-scale climate such as ENSO and the Southern Annual Mode, which allows us to explore the possible connection between the large-scale climate variability and variability in the weather characteristics. Due to the timing of decadal variability for our chosen periods in the model and observations, and the lack of notable trend in the model, any changes in weather characteristics in the model is presumed to be due to decadal variability and in the observations is presumed to be due to trend. This information can potentially be merged to produce synthetic, but representative time series, that would provide better constrained weather and climate inputs for crop models. For the 2020s, the mean decadal precipitation and days with extreme precipitation (R95p) are projected to be lower than in the 2000s, while the frequency of consecutive dry days is expected to be higher.
The physical mechanisms and predictability associated with extreme daily rainfall in South East South America (SESA) are investigated for the December-February season. Through a k-mean analysis, a robust set of daily circulation regimes is identified and then it is used to link the frequency of rainfall extreme events with large-scale potential predictors at subseasonal-to-seasonal scales. This basic set of daily circulation regimes is related to the continental and oceanic phases of the South Atlantic Convergence Zone (SACZ) and wave train patterns superimposed on the Southern Hemisphere Polar Jet. Some of these recurrent synoptic circulation types are conducive to extreme rainfall events in the region through synoptic control of different meso-scale physical features and, at the same time, are influenced by climate phenomena that could be used as sources of potential predictability. Extremely high rainfall (as measured by the 95th- and 99th-percentiles) is preferentially associated with two of these weather types, which are characterized by moisture advection intrusions from lower latitudes and the Pacific; another three weather types, characterized by above-normal moisture advection toward lower latitudes or the Andes, are preferentially associated with dry days (days with no rain). The analysis permits the identification of several subseasonal-to-seasonal scale potential predictors that modulate the occurrence of circulation regimes conducive to extreme rainfall events in SESA. It is conjectured that a cross-timescale interference between the different climate drivers improves the predictive skill of extreme precipitation in the region. The potential and real predictive skill of the frequency of extreme rainfall is then evaluated, finding evidence indicating that mechanisms of climate variability at one timescale contribute to the predictability at another scale, i.e., taking into account the interference of different potential sources of predictability at different timescales increases the predictive skill. This fact is in agreement with the Cross-timescale Interference Conjecture proposed in the first part of the thesis. At seasonal scale, a combination of those weather types tends to outperform all the other potential predictors explored, i.e., sea surface temperature patterns, phases of the Madden-Julian Oscillation, and combinations of both. Spatially averaged Kendall’s τ improvements of 43% for the potential predictability and 23% for realtime predictions are attained with respect to standard models considering sea-surface temperature fields alone. A new subseasonal-to-seasonal predictive methodology for extreme rainfall events is proposed, based on probability forecasts of seasonal sequences of these weather types. The cross-validated realtime skill of the new probabilistic approach, as measured by the Hit Score and the Heidke Skill Score, is on the order of twice that associated with climatological values. The approach is designed to offer useful subseasonal-to-seasonal climate information to decision-makers interested not only in how many extreme events will happen in the season, but also in how, when and where those events will probably occur. In order to gain further understanding about how the cross-timescale interference occurs, an externally-forced Lorenz model is used to explore the impact of different kind of forcings, at inter-annual and decadal scales, in the establishment of constructive interactions associated with the simulated “extreme events”. Using a wavelet analysis, it is shown that this simple model is capable of reproducing the same kind of cross-timescale structures observed in the wavelet power spectrum of the Niño3.4 index only when it is externally forced by both inter-annual and decadal signals: the annual cycle and a decadal forcing associated with the natural solar variability. The nature of this interaction is non-linear, and it impacts both mean and extreme values in the time series. No predictive power was found when using metrics like standard deviation and auto-correlation. Nonetheless, it was proposed that an early warning signal for occurrence of extreme rainfall in SESA may be possible via a continuous monitoring of relative phases between the cross-timescale leading components.
James Doss-Gollin
added a research item
During the austral summer 2015-16, severe flooding displaced over 170 000 people on the Paraguay River system in Paraguay, Argentina, and Southern Brazil. These floods were driven by repeated heavy rainfall events in the Lower Paraguay River Basin. Alternating sequences of enhanced moisture inflow from the South American Low-Level Jet and local convergence associated with baroclinic systems were conducive to mesoscale convective activity and enhanced precipitation. These circulation patterns were favored by cross-timescale interactions of a very strong El Niño event, an unusually persistent Madden-Julian Oscillation in phases four and five, and the presence of a dipole SST anomaly in the central southern Atlantic Ocean. The simultaneous use of seasonal and sub-seasonal heavy rainfall predictions could have provided decision makers useful information about the start of these flooding events from two to four weeks in advance. Probabilistic seasonal forecasts available at the beginning of November successfully indicated heightened probability of heavy rainfall (90th percentile) over southern Paraguay and Brazil for December-February. Raw sub-seasonal forecasts of heavy rainfall exhibited limited skill at lead times beyond the first two predicted weeks, but a Model Output Statistics approach involving principal component regression substantially improved the spatial distribution of skill for week 3 relative to other methods tested including extended logistic regressions. A continuous monitoring of climate drivers impacting rainfall in the region, and the use of statistically corrected heavy precipitation seasonal and sub-seasonal forecasts, may help improve flood preparedness in this and other regions.
Ángel G Muñoz
added a research item
The International Research Institute for Climate and Society Data Li- brary (IRIDL) is a powerful and freely accessible online data repository and analysis web-service that allows a user to view, analyze, and down- load hundreds of terabytes of climate-related data through a standard web browser in a computer or a smartphone. A wide variety of opera- tions, from simple anomaly calculations to more complex EOF or cluster analyses can be performed with just a few clicks. About 75% of the S2S Database forecasts and reforecasts, including all models’ RMM indices for MJO analysis, are presently available in the IRIDL from the ECMWF server and kept up to date as new forecasts & re- forecasts are made. It provides a flexible and fully online interface to the S2S database, for easy subsetting, analysis & visualization, and download in a variety of formats, including NetCDF, GoogleEarth’s KML and GIS- compatible layers. Furthermore, the IRIDL is an OpenDAP server, which means local client programs –e.g., written in Python, R or Matlab– can read the desired data online, avoiding the
Ángel G Muñoz
added an update
We just submitted two contributions to the 2nd International Conferences on Subseasonal-to-Decadal Prediction (Sep 2018, Boulder, CO).
 
Ángel G Muñoz
added a research item
Common approaches to diagnose systematic errors involve the computation of metrics aimed at providing an overall summary of the performance of the model in reproducing the particular variables of interest in the study, normally tied to specific spatial and temporal scales. However, the evaluation of model performance is not always tied to the understanding of the physical processes that are correctly represented, distorted or even absent in the model world. As the physical mechanisms are more often than not related to interactions taking place at multiple time and spatial scales, cross-scale model diagnostic tools are not only desirable but required. Here, a recently proposed circulation-based diagnostic framework is extended to consider systematic errors in both spatial and temporal patterns at multiple timescales. The framework, which uses a weather-typing dynamical approach, quantifies biases in shape, location and tilt of modeled circulation patterns, as well as biases associated with their temporal characteristics, such as frequency of occurrence, duration, persistence and transitions. Relationships between these biases and climate teleconnections (e.g., ENSO and MJO) are explored using different models.
Ángel G Muñoz
added a research item
Recent research has highlighted the potential for improving predictive skill at the sub-seasonal timescale, which could be the basis for enhanced, actionable forecasts for climate services involving water and disaster management, health, energy and food security. The WMO's World Weather and World Climate Research Programmes Subseasonal-to-Seasonal Prediction Project (S2S) has made available an extensive database with both hindcasts and almost-realtime forecast at this timescale. Lead times are long enough that much of the information in the atmospheric initial conditions is lost, but at the same time are too short for other sources of predictability (e.g., ocean boundary conditions) to have a strong influence in skill. Presently, sub-seasonal skill is still limited, and in general raw uncalibrated forecasts cannot be used to develop climate services. An obvious alternative is to make use of a variety of robust bias-correction and calibration methods also known as Model Output Statistics (MOS) available for other timescales, such as the seasonal one. Nonetheless, some technical issues can hinder this approach. We discuss problems and advantages of applying MOS to sub-seasonal forecasts, analyzing the spatio-temporal variability of skill in several models and methods.
James Doss-Gollin
added a research item
During the austral summer 2015-16, severe flooding displaced over 170000 people on the Paraguay River system in Paraguay, Argentina, and Southern Brazil. These floods were driven by repeated heavy rainfall events in the Lower Paraguay River Basin. Alternating sequences of enhanced moisture inflow from the South American Low-Level Jet and local convergence associated with baroclinic systems were conducive to mesoscale convective activity and enhanced precipitation. These circulation patterns were favored by cross-timescale interactions of a very strong El Niño event, an unusually persistent Madden-Julian Oscillation in phases four and five, and the presence of a dipolar SST anomaly in the central southern Atlantic Ocean. The simultaneous use of seasonal and sub-seasonal heavy rainfall predictions could have provided decision makers useful information about the start of these flooding events from at least two-to-four weeks in advance. Probabilistic seasonal forecasts available at the beginning of November successfully indicated heightened probability of heavy rainfall (90th percentile) over southern Paraguay and Brazil for December-February. Raw sub-seasonal forecasts of heavy rainfall exhibited limited skill at lead times beyond the first two predicted weeks, but a Model Output Statistics approach involving principal component regression substantially improved the spatial distribution of skill for Week 3 relative to other methods tested including extended logistic regressions. A continuous monitoring of climate drivers impacting rainfall in the region, and the use of bias-corrected heavy precipitation seasonal and sub-seasonal forecasts, may help improve flood preparedness for the austral summer season in this part of the world.
Ángel G Muñoz
added an update
Final version of the cross-timescale diagnostic paper is finally online in JCLI
 
Ángel G Muñoz
added an update
Our most recent paper on cross-timescale diagnostics of climate models using weather types as building blocks of the framework is online:
 
Ángel G Muñoz
added a research item
This study proposes an integrated diagnostic framework based on atmospheric circulation regime spatial patterns and frequencies of occurrence to facilitate the identification of model systematic errors across multi- ple timescales. To illustrate the approach, three sets of 32-year-long simulations are analyzed for northeastern North America and for the March-May season using the Geophysical Fluid Dynamics Laboratory’s Low Ocean-Atmosphere Resolution (LOAR) and Forecast-oriented Low Ocean Resolution (FLOR) coupled mod- els; the main difference between these two models is the horizontal resolution of the atmospheric model used. Regime-dependent biases are explored in the light of different atmospheric horizontal resolutions and under different nudging approaches. It is found that both models exhibit a fair representation of the observed circu- lation regime spatial patterns and frequencies of occurrence, although some biases are present independently of the horizontal resolution or the nudging approach, and are associated with a misrepresentation of troughs centered north of the Great Lakes, and deep coastal troughs. Moreover, the intra-seasonal occurrence of cer- tain model regimes is delayed with respect to observations. On the other hand, inter-experiment differences in the mean frequencies of occurrence of the simulated weather types, and their variability across multiple timescales, tend to be negligible. This result suggests that low-resolution models could be of potential use to diagnose and predict physical variables via their simulated weather type characteristics.
Ángel G Muñoz
added an update
Nathaniel C. Johnson presented in AMS' 30th Conference on Climate Variability and Change our most recent work on pattern-dependent bias correction for seasonal forecasts, using a dynamical-statistical approach involving weather types. The poster is available as a document in this project.
 
Ángel G Muñoz
added a research item
Dynamical forecast models provide a foundation for seasonal forecast systems, but systematic errors may arise for various reasons, including insufficient spatial resolution, insufficient ensemble size, and errors in physical parameterizations. Despite these flaws, the ability of dynamical models to simulate the sources of prediction skill (e.g., ENSO) and their large-scale circulation responses allows us to draw from empirical predictor/large-scale circulation relationships to compensate for these shortcomings. In this study we use the framework known as weather types (WTs) to act as the mediator for a hybrid dynamical-statistical seasonal forecast system. WTs are large-scale circulation patterns that, in this application, are determined by k-means clustering of geopotential height. We generate seasonal forecasts for December – February over eastern North America by taking dynamical model forecasts of WTs and then using empirical relationships to translate these WT forecasts into probabilistic temperature and precipitation forecasts. We use hindcasts from both a lower resolution (CM2.1) and higher resolution (FLOR) dynamical forecast model from the Geophysical Fluid Dynamics Laboratory (GFDL), considering different initialization strategies. This application of WTs essentially serves as a pattern-dependent bias correction and downscaling approach. We evaluate the performance of the hybrid dynamical-statistical forecasts in the context of more conventional post-processing methods.
Ángel G Muñoz
added a research item
Potential and real predictive skill of the frequency of extreme rainfall in South East South America for the December-February season are evaluated in this paper, finding evidence indicating that mechanisms of climate variability at one timescale contribute to the predictability at another scale, i.e., taking into account the interference of different potential sources of predictability at different timescales increases the predictive skill. Muñoz et al. (2015) suggested that a set of daily atmospheric circulation regimes, or weather types, was sensitive to these cross-timescale interferences, conducive to the occurrence of extreme rainfall events in the region, and could be used as potential predictor. At seasonal scale, a combination of those weather types indeed tends to outperform all the other candidate predictors explored, i.e., sea surface temperature patterns, phases of the Madden-Julian Oscillation, and combinations of both. Spatially averaged Kendall’s τ improvements of 43% for the potential predictability and 23% for realtime predictions are attained with respect to standard models considering sea-surface temperature fields alone. A new subseasonal-to-seasonal predictive methodology for extreme rainfall events is proposed, based on probability forecasts of seasonal sequences of these weather types. The cross-validated realtime skill of the new probabilistic approach, as measured by the Hit Score and the Heidke Skill Score, is on the order of twice the associated with climatological values. The approach is designed to offer useful subseasonal-to-seasonal climate information to decision-makers interested not only in how many extreme events will happen in the season, but also in how, when and where those events will probably occur.
Ángel G Muñoz
added a project goal
+ To analyze cross-timescale interference between climate drivers acting at multiple time scales, and their impact on climate hazard predictability.
+ To define and apply a seamless diagnostic framework comparing modeled and observed weather type's statistics, spatial patterns and physical links to climate drivers across timescales.
+ To assess predictive skill of forecast systems at multiple timescales