Article

Modeling grassland spring onset across the Western United States using climate variables and MODIS-derived phenology metrics

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

Vegetation phenology strongly controls photosynthetic activity and ecosystem function and is essential for monitoring the response of vegetation to climate change and variability. Terrestrial ecosystem models require robust phenology models to understand and simulate the relationship between ecosystems and a changing climate. While current phenology models are able to capture inter-annual variation in the timing of vegetation spring onset, their spatiotemporal performances are not well understood. Using green-up dates derived from MODIS, we test 9 phenological models that predict the timing of grassland spring onset via commonly available climatological variables. Model evaluation using satellite observations suggests that Modified Growing-Degree Day (MGDD) models and Accumulated Growing Season Index (AGSI) models achieve reasonable accuracy (RMSE < 20 days) after model calibration. Inclusion of a photoperiod trigger and varied critical forcing thresholds in the temperature-based phenology model improves model applicability at a regional scale. In addition, we observe that AGSI models outperform MGDD models by capturing inter-annual phenology variation in large semi-arid areas, likely due to the explicit consideration of water availability. Further validation based on flux tower sites shows good agreement between the modeled timing of spring onset and references derived from satellite observations and in-situ measurements. Our results confirm recent studies and indicate that there is a need to calibrate current phenology models to predict grassland spring onsets accurately across space and time. We demonstrate the feasibility of combining satellite observations and climatic datasets to develop and refine phenology models for characterizing the spatiotemporal patterns of grassland green-up variations.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... Besides, the responses of carbon fluxes, phenological and physiological indicators for GRA to local environmental factors are more complicated than for forest ecosystems. The SOS of GRA in moist environments was sensitive to variation of temperature, but the SOS of GRA in dry areas was limited by soil water content and often initiated by precipitation events (Xin et al., 2015). The relationships between spring carbon uptake anomalies and carbon uptake anomalies in other seasons were insignificant for GRA sites (Fu et al., 2017b), which indirectly implied that climate event and human disturbance effects on seasonal carbon uptake of GRA should not be overlooked and exceed the legacy effects of spring carbon uptake on later seasons. ...
... The large fluctuation in phenological and physiological indicators and seasonal GPP in GRA sites caused by climate and human disturbances may lead to unstable long-term trends in phenological and physiological indicators and seasonal GPP, which reduces the prediction ability of the models for GRA sites. Besides, a previous study has suggested that the spatiotemporal variation of vegetation phenology for GRA was difficult to be captured using satellite observations ( Xin et al., 2015). The C3 grasses starts growth early in the spring and may have a second growing peak when temperature cools down in the fall ( Wang et al., 2013), which also make it difficult to estimate the phenological and physiological indicators. ...
... The C3 grasses starts growth early in the spring and may have a second growing peak when temperature cools down in the fall ( Wang et al., 2013), which also make it difficult to estimate the phenological and physiological indicators. Given that GRA is a key component in terrestrial biomes, more attention should be paid to improve the models for estimating physiology, phenology and carbon fluxes of GRA ( Xin et al., 2015) and to further discover climate event and human disturbance effects on phenology and carbon dynamic of GRA. ...
... Therefore, many studies incorporated water availability sequentially or parallel to simulate the spring phenology of vegetation in northern grassland areas based GDD model. The results demonstrated a higher accuracy of water-incorporated models in simulating the spring phenology of grassland vegetation in most areas (Chen et al., 2015(Chen et al., , 2014Fan et al., 2020;Li and Zhou, 2012;Liu et al., 2013;Xin et al., 2015). However, most modeling studies on the spring phenology of grassland plants are implemented at site-species and regional scale. ...
... In this study, we found that temperature and precipitation-based models (TPP or TPS) are optimal models in simulating spring phenology of grassland vegetation for approximately 2/3 of the study region, which indicates a critical role of moisture in regulating spring phenology of grassland vegetation at the mid-latitudes of the Northern Hemisphere. Similar results have also been obtained by previous studies based on ground observations in the Inner Mongolia grasslands and remote sensing data in the North American grasslands (Chen et al., 2014;Xin et al., 2015). Moreover, the proportion of pixels with the temperature-precipitation sequential model as the optimal model is about twice of the temperature-precipitation parallel model, indicating that for most areas, precipitation affects spring vegetation growth first and temperature triggers spring phenology occurrence only when the moisture requirements are fulfilled. ...
... RMSEs of predicted green-up dates for same species at same locations in the Inner Mongolia grasslands based on GDD model are generally larger than those based on TPP and TPS models (Chen et al., 2014), which can indirectly prove the critical effect of moisture conditions on regulating spring phenology of herbaceous plants. In the North American grasslands, however, a much larger average RMSE (16.4 days) was detected by Xin et al. (2015), who compared the performance of multiple phenological models in simulating GSS extracted from remote sensing data. The better accuracy of our study may result from the relatively more accurate phenology metric retrieved from remote sensing data. ...
Article
Understanding spatial heterogeneity of grassland phenology responses to climate change is of crucial importance for revealing regional and species differences in ecosystem processes. In this study, three spring phenology models, namely, growing-degree-day model (GDD), temperature-precipitation parallel model (TPP) and temperature-precipitation sequential model (TPS) were employed to simulate the growing season start (GSS) retrieved from remote sensing data during 1981–2014 in mid-latitude (30°N-55°N) grasslands of the Northern Hemisphere. Results show that the average accuracies of predicted GSS based on TPP (root mean square errors (RMSE) = 9.9 days) and TPS (RMSE = 9.7 days) models are slightly higher than that based on GDD model (RMSE = 10.1 days) overall. Meanwhile, TPP/TPS model also exhibits a stronger capacity to simulate interannual variation of GSS (correlation coefficient (R) = 0.38/0.41 on average) than the GDD model (R = 0.3 on average). A revised Akaike information criterion (AICc_R) by including correlation coefficient between predicted and retrieved GSS was designed for the optimal model selection. The optimal models based on AICc_R present a stronger power in capturing the temporal pattern of spring phenology. GDD, TPP, and TPS model account for 32.7%, 20.7%, and 46.6% of the whole study region, respectively. Sub-regionally, TPS model dominates the temperate grasslands (66.1%), whereas GDD model dominates the cool semidesert grasslands (58.1%). Regions occupied by GDD model as the optimal model are generally cooler and wetter during February to May than those taken up by TPP and TPS model. Further analysis indicates higher heat and water requirements are needed in the most of warmer places. Overall, this study emphasizes the important role of thermal-moisture background in controlling the spatial pattern of spring phenology of northern grasslands in response to climate change. Precipitation is very important for triggering spring phenology in the temperate grasslands but not in the cool semidesert grasslands.
... In contrast, land surface phenology derived from satellite data provides exhaustive spatial distributions, which can be used to establish much more robust phenology models to illustrate climate impacts on phenology variation on the QTP [21,35]. However, current detections of land surface phenology vary greatly, with different methods of processing the noisy time series of satellite observations [36,37]. ...
... This could overcome the drawbacks in commonly used process-based phenology models established using extremely limited in situ observations. The process-based model using LSGO has shown advantages over a large region [35,50], such as the model for the woody leaf unfolding date in the deciduous broadleaf forest region of China [50]. ...
Article
Full-text available
As a sensitive indicator for climate change, the spring phenology of alpine grassland on the Qinghai–Tibet Plateau (QTP) has received extensive concern over past decade. It has been demonstrated that temperature and precipitation/snowfall play an important role in driving the green-up in alpine grassland. However, the spatial differences in the temperature and snowfall driven mechanism of alpine grassland green-up onset are still not clear. This manuscript establishes a set of process-based models to investigate the climate variables driving spring phenology and their spatial differences. Specifically, using 500 m three-day composite MODIS NDVI datasets from 2000 to 2015, we first estimated the land surface green-up onset (LSGO) of alpine grassland in the QTP. Further, combining with daily air temperature and precipitation datasets from 2000 to 2015, we built up process-based models for LSGO in 86 meteorological stations in the QTP. The optimum models of the stations separating climate drivers spatially suggest that LSGO in grassland is: (1) controlled by temperature in the north, west and south of the QTP, where the precipitation during late winter and spring is less than 20 mm; (2) driven by the combination of temperature and precipitation in the middle, east and southwest regions with higher precipitation and (3) more likely controlled by both temperature and precipitation in snowfall dominant regions, since the snow-melting process has negative effects on the air temperature. The result dictates that snowfall and rainfall should be concerned separately in the improvement of the spring phenology model of the alpine grassland ecosystem.
... Satellite based NDVI phenological analysis has been widely utilized to find both spatial and temporal trends in important climate indices such as start of season (Hüttich et al., 2009;Jonsson and Eklundh, 2002;Xin et al., 2015;Zhanga et al., 2003). Conventional methods for observing growth cycles over a season with NDVI data involve first quality checking, filtering and gap-filling the data, then fitting a complex time-series function. ...
... There are many indicators of grass growth that could have been chosen such as sward height, biomass, quality, phenological stage or cover, many of which are reflected in NDVI measurements (Ali et al., 2016;Butterfield and Malmström, 2009). A recent study of grassland phenology across continental USA found that Growing Degree Days were the best match to MODIS derived indices (Xin et al., 2015). However this study and others have found that the fragmentation of landscape and the poor spatial and temporal quality of phenological ground truth limit the ability to truly test these satellite based indices of phenology (Reed et al., 2009) and linking ground phenology indicators to satellite derived indicators generally is tricky so data are often presented only for comparison, rather than validation (Rodriguez-Galiano et al., 2015). ...
Article
Seasonal progress anomaly mapping uses satellite data to identify the current stage of growth of vegetation compared to the normal stage of growth at that time and place. Usually applied to crops, the potential for such a service in intensive grasslands is developed here. The research identifies the need for monitoring progress of grass growth in spring for effective herd and paddock management in the context of increasing seasonal weather variability. Using 12 Years of NASA MODIS satellite data and 12 years of ground climate station data in Ireland, NDVI was modeled against time as a proxy for grass growth. This model is the reference for estimating current seasonal progress of grass growth against a ten year average. The model is developed to estimate Seasonal Progress Anomalies in the Time domain (SPAT), giving a result in terms of days behind and days ahead of the norm instead of percentage difference which is the current reporting method for this type of service. As a comparison SPAT estimates for 2012 and 2013 are compared to ground based estimates of seasonal progress anomaly from 30 climate stations and show a correlation coefficient of 0.897 and RMSE of 15days. SPAT maps for these two seasons for the whole of Ireland are generated every 16 days with a spatial resolution of 250 m and two are given as examples. The method can successfully map current grass growth trends compared to the average and present this information to the farmer in simple everyday language. The usefulness of a producing a service based on this approach to aid farmers’ planning in the context of increasing weather volatility is discussed.
... Although we use spatio-temporal fusion techniques to enhance the resolution of the vegetation index series, it is still difficult to accurately catch the rapid changes of vegetation over a short period using these fusion techniques. The existing spatio-temporal fusion techniques usually have an error of about 20d when performing vegetation SOS extraction [48]. Taking this issue into account, we fitted the data using the D-L algorithm. ...
... Additionally, the absolute value of the deviation of the inversion using f 1 ustar f m was the largest at 18d, and the absolute value of the deviation of 50% of the sampling points was concentrated in 8.5d-27.9d. The RMSE of M 1 and M 8 data were 11.3d and 14.3d respectively, which shown that remote sensing identification of vegetation SOS in Xilinhot based on MODIS data is feasible [48]. ...
Article
Full-text available
Estimating the Start of Growing Season (SOS) of grassland on the global scale is an important scientific issue since it can reflect the response of the terrestrial ecosystem to environmental changes and determine the start time of grazing. However, most remote sensing data has coarse- temporal and spatial resolution, resulting in low accuracy of SOS retrieval based on remote sensing methods. In recent years, much research has focused on multi-source data fusion technology to improve the spatio-temporal resolution of remote sensing information, and to provide a feasible path for high-accuracy remote sensing inversion of SOS. Nevertheless, there is still a lack of quantitative evaluation for the accuracy of these data fusion methods in SOS estimation. Therefore, in this study, the SOS estimation accuracy is quantitatively evaluated based on the spatio-temporal fusion daily datasets through the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) and other models in Xilinhot City, Inner Mongolia, China. The results show that: (1) the accuracy of SOS estimation based on spatio-temporal fusion daily datasets has been slightly improved, the average Root Mean Square Error (RMSE) of SOS based on 8d composite datasets is 11.1d, and the best is 9.7d (fstarfm8); (2) the estimation accuracy based on 8d composite datasets (RMSE¯ = 11.1d) is better than daily fusion datasets (RMSE¯ = 18.2d); (3) the lack of the Landsat data during the SOS would decrease the quality of the fusion datasets, which ultimately reduces the accuracy of the SOS estimation. The RMSE¯ of SOS based on all three models increases by 11.1d, and the STARFM is least affected, just increases 2.7d. The results highlight the potential of the spatio-temporal data fusion method in high-accuracy grassland SOS estimation. It also shows that the dataset fused by the STARFM algorithm and composed for 8 days is better for SOS estimation.
... We chose the image chip centered at 42.535 • N and 72.185 • W to guarantee that the ground measurements (i.e., Data Archive HF-003) are located within the image chip and thus can be used to validate the LSP metrics derived from remote sensing data. In addition, choosing image chips instead of pixels can minimize the pixel-to-pixel difference, and therefore has been adopted in several angle effect studies [50,51]. ...
... The reconstructed NDVI/EVI time series were then used to derive LSP metrics, including the start of season (SOS) and the end of season (EOS) [87], with SOS and EOS referring to the point at which the curve intersected a proportion of the seasonal amplitude measured from the left minimum and from the right minimum, respectively [51,[88][89][90][91][92][93]. ...
Article
Full-text available
Vegetation indices are widely used to derive land surface phenology (LSP). However, due to inconsistent illumination geometries, reflectance varies with solar zenith angles (SZA), which in turn affects the vegetation indices, and thus the derived LSP. To examine the SZA effect on LSP, the MODIS bidirectional reflectance distribution function (BRDF) product and a BRDF model were employed to derive LSPs under several constant SZAs (i.e., 0°, 15°, 30°, 45°, and 60°) in the Harvard Forest, Massachusetts, USA. The LSPs derived under varying SZAs from the MODIS nadir BRDF-adjusted reflectance (NBAR) and MODIS vegetation index products were used as baselines. The results show that with increasing SZA, NDVI increases but EVI decreases. The magnitude of SZA-induced NDVI/EVI changes suggests that EVI is more sensitive to varying SZAs than NDVI. NDVI and EVI are comparable in deriving the start of season (SOS), but EVI is more accurate when deriving the end of season (EOS). Specifically, NDVI/EVI-derived SOSs are relatively close to those derived from ground measurements, with an absolute mean difference of 8.01 days for NDVI-derived SOSs and 9.07 days for EVI-derived SOSs over ten years. However, a considerable lag exists for EOSs derived from vegetation indices, especially from the NDVI time series, with an absolute mean difference of 14.67 days relative to that derived from ground measurements. The SOSs derived from NDVI time series are generally earlier, while those from EVI time series are delayed. In contrast, the EOSs derived from NDVI time series are delayed; those derived from the simulated EVI time series under a fixed illumination geometry are also delayed, but those derived from the products with varying illumination geometries (i.e., MODIS NBAR product and MODIS vegetation index product) are advanced. LSPs derived from varying illumination geometries could lead to a difference spanning from a few days to a month in this case study, which highlights the importance of normalizing the illumination geometry when deriving LSP from NDVI/EVI time series.
... Jolly et al. (2005) developed a generalized growing season phenology model from photoperiod, minimum temperatures, and vapor pressure deficit to address phenological timing at model sites. Xin et al. (2015) expanded this work by modeling growing season start dates in United States (US) grasslands through multiple methods, including variants of the Jolly et al. (2005) growing season index. Their models achieved reasonable accuracy in predicting growing season onset across this large study area. ...
... Xin et al. (2015) also modeled start of season timing in western U.S. grasslands using eMODIS phenology data, and their best models outperformed ours, with RMSE of 16.4 days. This indicates the potential to improve our model by incorporating either a modified growing degree day or accumulated growing season index term, as they used (Xin et al. 2015). However, Xin et al. did not address end of season timing, whereas we employed a unified model of growing season start and end, which may constrain model performance. ...
Article
Full-text available
Climate and vegetation phenology are closely linked, and climate change is already impacting phenology in many systems. These impacts are expected to progress in the future. We sought to forecast future shifts in rangeland growing season timing due to climate change, and interpret their importance for land management and ecosystem function. We trained a model on remotely sensed land surface phenology and climate data collected from 2001 to 2014 in temperate United States rangelands. We used this model to forecast annual growing season start dates, end dates, and season length through 2099 among six general circulation models and under RCP 4.5 and 8.5 scenarios. Growing season start was projected to shift earlier throughout our study area. In 2090–2099, start of season advanced by an average of 10 (RCP 4.5) to 17 (RCP 8.5) days. End of season also advanced by 12 (RCP 4.5) to 24 (RCP 8.5) days, but with greater heterogeneity. Start and end of season change mainly offset one another, so growing season length changes were lesser (2 days in RCP 4.5, and 7 in RCP 8.5). Some mountainous areas experienced both earlier start of season and later end of season, lengthening their growing season. Earlier phenology in rangelands would force adaptation in grazing and impact ecosystem function. Mountainous areas with earlier start and later end of season may become more viable for grazing, but most areas may experience slightly shortened growing seasons. Autumn phenology warrants greater research, and our finding of earlier autumn senescence contradicts some prior research.
... The phenology of the prairie grasslands is known to be sensitive to seasonal and interannual changes in climate variables, including temperature, precipitation, and drought (Lesica and Kittelson 2010;Reed 2006;Cui, Martz, and Guo 2017;Yuan, Wang, and Mitchell 2014). Satellite data with different temporal resolutions, including 7-, 10-, and 15-day AVHRR NDVI (Li and Guo 2012;Bradley et al. 2007;Reed et al. 1994), 1-, 5-, 8-, and16-day MODIS VIs (Dye et al. 2016;Xin et al. 2015;Moon et al. 2019;Cui et al. 2019), and 3-day Visible Infrared Imaging Radiometer Suite (VIIRS) EVI2 (Zhang et al. 2018;Moon et al. 2019) have been used to detect changes in LSP metrics on North American grasslands. The effect of temporal resolution of satellite data on the accuracy of LSP metrics of Canadian prairie grasslands remains unclear for two reasons. ...
Article
The temporal resolution of vegetation indices (VIs) determines the details of seasonal variation in vegetation dynamics observed by remote sensing, but little has been known about how the temporal resolution of VIs affects the retrieval of land surface phenology (LSP) of grasslands. This study evaluated the impact of temporal resolution of MODIS NDVI, EVI, and per-pixel green chromatic coordinate (GCCpp) on the quality and accuracy of the estimated LSP metrics of prairie grasslands. The near-surface PheonoCam phenology data for grasslands centered over Lethbridge PhenoCam grassland site were used as the validation datasets due to the lack of in situ observations for grasslands in the Prairie Ecozone. MODIS Nadir Bidirectional Reflectance Distribution Function (BRDF)-Adjusted Reflectance (NBAR) data from 2001 to 2017 were used to compute the time series of daily reference and to simulate 2–32 day MODIS VIs. The daily reference and simulated multi-day time series were fitted with the double logistic model, and the LSP metrics were then retrieved from the modeled daily time series separately. Comparison within satellite-based estimates showed no significant difference in the phenological metrics derived from daily reference and multi-day VIs resampled at a time step less than 18 days. Moreover, a significant decline in the ability of multi-day VIs to predict detailed temporal dynamics of daily reference VIs was revealed as the temporal resolution increased. Besides, there were a variety of trends for the onset of phenological transitions as the temporal resolution of VIs changed from 1 to 32 days. Comparison with PhenoCam phenology data presented small and insignificant differences in the mean bias error (MBE) and the mean absolute error (MAE) of grassland phenological metrics derived from daily, 8-, 10-, 14-, and 16-day MODIS VIs. Overall, this study suggested that the MODIS VIs resampled at a time step less than 18 days are favorable for the detection of grassland phenological transitions and detailed seasonal dynamics in the Prairie Ecozone.
... Even in areas with high vegetation cover and plants with distinct signals, phenology may be impossible to detect some years due little to no plant productivity from inadequate precipitation. These limitations cause drylands to be excluded in many large scale LSP analyses [14][15][16][17][18] Most studies focusing on drylands evaluate aggregate or peak annual VI as opposed to distinct seasonal transitions [13,19], and even then can occasionally have inconclusive results [20,21]. ...
Preprint
Full-text available
Land surface phenology, the tracking of seasonal productivity via satellite remote sensing, enables global scale tracking of ecosystem processes, but its utility is limited in some areas. In dryland ecosystems low vegetation cover can cause the growing season vegetation index (VI) to be indistinguishable from the dormant season VI, making phenology extraction impossible. Here, using simulated data and multi-temporal UAV imagery of a desert shrubland, we explore the feasibility of detecting LSP with respect to fractional vegetation cover, plant functional types, and VI uncertainty. We found that plants with distinct VI signals, such as deciduous shrubs with a high leaf area index, require at least 30-40\% fractional cover on the landscape to consistently detect pixel level phenology with satellite remote sensing. Evergreen plants, which have lower VI amplitude between dormant and growing seasons, require considerably higher cover and can have undetectable phenology even with 100\% vegetation cover. We also found that even with adequate cover, biases in phenological metrics can still exceed 20 days, and can never be 100\% accurate due to VI uncertainty from shadows, sensor view angle, and atmospheric interference. Many dryland areas do not have detectable LSP with the current suite of satellite based sensors. Our results showed the feasibility of dryland LSP studies using high-resolution UAV imagery, and highlighted important scale effects due to within canopy VI variation. Future sensors with sub-meter resolution will allow for identification of individual plants and are the best path forward for studying large scale phenological trends in drylands.
... Previous studies normally used satellite-based remote sensing methods to explore phenological variations at regional or global scales (Wu, Hou, Peng, Gonsamo, & Xu, 2016;Xin, Broich, Zhu, & Gong, 2015). However, satellite-derived phenology, also known as land surface phenology, is based on derivative methods and reflects the state of community development, which is essentially different from species-level phenology (Liang, Schwartz, & Fei, 2011;Tao, Huang, & Wang, 2020). ...
Article
Full-text available
Autumn phenology variation in temperate grassland is a direct indicator of land degradation, and a phenology‐based model has been used to simulate human‐induced land degradation. It is important to reveal how environmental factors influence autumn phenology of herbaceous plants. In this study, we examined the effects of temperature, photoperiod, and soil moisture on the leaf withering date (LWD) of four herbaceous species (Xanthium sibiricum, Plantago asiatica, Iris lactea, and Taraxacum mongolicum) at 15 sites in the Inner Mongolian steppe from 1981 to 2012. Then, we developed two process‐based models coupling the effects of the three factors based on an existing cold degree–days (CDD)‐based model for predicting the LWD of the four species. The new models showed better accuracy than CDD model during validation. The model structure and parameters suggested that soil moisture likely influenced LWD nonlinearly and presented a multiplicative interaction with temperature and photoperiod. In addition, photoperiod had divergent effects on the accumulation of cold temperature for different species and sites. For X. sibiricum, the photoperiod only determined the start of cold temperature accumulation. Considering the carryover effect of spring phenology did not improve model performance. Our study also showed that local adaptation of plants was commonly ignored when using same parameters in model calibration with observation data from multiple sites; however, model calibrations using site‐specific data were less stable, and the parameters could not be extrapolated across space. It will be necessary to establish a generalized regional model driven by both climatic and genotypic factors in future research.
... Therefore, we suggest that this estimator should be used to obtain a consistent and 396 unbiased estimation of density if the population is entirely stochastically distributed (randomness) and should be 397 avoided in the uniformly or contagiously distributed populations. Though, the bias can be corrected by an 398 exponential function (Xin et al., 2015). Basiri et al., (2018)also stated that the PCQ is a precise estimator of 399 population density but on the contrary to Silva et al. (2017), they believe that the PCQ show moderately poor 400 precision in random and clumped patterns, which, is not consistent with our results. ...
Article
Full-text available
Choosing appropriate estimator that provides an accurate and precise prediction of plant population's density is vital specifically when different density intensities and distribution patterns are concerned. Therefore, the efficiency of plotless plant density estimators for the various spatial patterns found in nature have been examined using a simulated population based on an observed population of Astragalus microcephalus in a semi-arid environment. We first surveyed the density of A. microcephalus in the field to have an estimation of the real density of the species (control method). Then a simulation scheme in three density intensities (low (mean−SD), moderate (equal to mean) and high (mean + SD)) and three distribution patterns (random, regular and aggregated) was drawn. Seven distance-based plant estimators were applied to evaluate their efficiency in the three density intensities and also distribution patterns within eight 40 × 100 m sampling units of the simulated scheme (repeats). The predictive precision and accuracy of the estimators in various density intensities and distribution patters were evaluated using the ideal point error-index and comparing the estimators predicted values with the controls (real densities). Angle Order (AO) and Third Closest Individual (TCI) in regular, TCI and Point Centered Quarter (PCQ) in random and AO in aggregate distribution pattern was the best plotless density estimators of plant populations. Overall, TCI, AO and PCQ were the most accurate and precise estimators of density among the seventh studied estimators in different density intensities and distribution patterns. Using these two estimators is recommended to achieve an unbiased estimation of plant population's density.
... In this study, we mainly focused on temperature-controlled ecosystems, which cover most regions in the CONUS. Due to the fact that spring vegetation phenology in the western CONUS with much higher interannual variation could be driven by the seasonality of rainfall or snow fall 83,94 , phenological interannual variations were not investigated in this study. It should be noted that detecting phenological events is very challenging in semiarid regions, where seasonal vegetation dynamics are complex, so that most currently available satellite-derived phenology products exclude phenology detections in the western CONUS 95 . ...
Article
Full-text available
Warming climate and its impact on vegetation phenological trends have been widely investigated. However, interannual variability in temperature is considerably large in recent decades, which is expected to trigger an increasing trend of variation in vegetation phenology. To explore the interannual phenological variation across the contiguous United States (CONUS), we first detected the onset of vegetation greenup using the time series of the daily two-band Enhanced Vegetation Index (EVI2) observed from the AVHRR Long-Term Data Record (1982–1999) and the MODIS Climate Modeling Grid (2000–2016). We then calculated the interannual variation in greenup onset during four decadal periods: 1982–1989, 1990–1999, 2000–2009 and 2010–2016. Further, the trend of interannual variation in greenup onset from 1982 to 2016 was analyzed at pixel and state levels. Extreme phenological events were also determined using a greenup onset anomaly for each pixel. Similar approaches were applied to spring temperatures to detect extreme years and to the temporal trend of interannual variation to explain the phenological variation. The results revealed that 62% of pixels show an increasing interannual variation in greenup onset, and in 44% of pixels, this variation could be explained by the temperature. Although extreme phenology occurred locally in different years, three nationwide extreme phenological years were distinguished. The extreme warm spring that occurred in 2012 resulted in the occurrence of greenup onset as much as 20 days earlier than normal in large parts of the CONUS. In contrast, greenup onset was much later (up to 30 days) in 1983 and 1996 due to cool spring temperatures. These findings suggest that interannual variation in spring phenology could be much stronger in the future in response to climate variation, which could have more significant impacts on terrestrial ecosystems than the regular long-term phenological trend.
... Given the complex potential interactions affecting spring phenology under climate change, further research is clearly required, for example, through experimental control of environmental drivers to improve understanding of the process (De . In addition, compared with tree and forest phenology, grassland phenology is addressed in only a few modelling studies, and the models are generally less developed than those for trees (Xin et al., 2015). Our study might provide an avenue to improve model performance by consideration of the shifting dominant factor in changing climatic conditions. ...
Article
Aim Vegetation phenology is highly sensitive to climate change. The timing of spring phenology in temperate grasslands is regulated primarily by temperature and precipitation. The aim of this study was to determine whether the primary factor regulating vegetation phenology has changed under ongoing climate change and the underlying mechanisms. Location Temperate semi‐dry grasslands in China. Time period 1982–2015. Major taxa studied Temperate grassland. Methods We extracted start of season (SOS) dates using five standard methods from satellite‐derived normalized difference vegetation index (NDVI) data and determined the primary factor regulating spring phenology using partial correlation analysis. Results The SOS date did not change significantly during the entire 1982–2015 study period in these semi‐dry grasslands, but interannual variability increased significantly from the first subperiod [1982–1998, 8.8 ± 1.1 days (mean ± SE)] to the second subperiod (1999–2015, 10.3 ± 1.1 days). Interestingly, we found that the primary factor regulating SOS shifted from precipitation during 1982–1998 to temperature during 1999–2015. Specifically, we found that during the first period, the SOS in 67.5% of the study area was determined by precipitation (mean partial correlation coefficient, r = 0.58 ± 0.16), but during the second period the main regulating factor in 75.0% of the study area was temperature (r = 0.61 ± 0.14). Main conclusions The change in the primary driver of spring phenology was attributed mainly to significant increases in preseason precipitation. Our study highlights that the response of spring phenology to climatic factors might change under ongoing climate change. This shift should be addressed in phenology models to simulate grassland phenology better, in addition to its impact on carbon and water cycles in future climate conditions.
... Further, the time series of VI represented by a set of functions [39], linear regression [40], Markov model [41], and curve-fitting functions. Sigmoid function has been exploited by [42,43] and achieved better results due to its robustness and ease to derive phenological features for the characterization of vegetation variability [44]. Although above-mentioned methods of temporal feature extraction offer many alternatives and flexibilities in deployment to assess vegetation dynamics, in practice, there are some important factors such as manually designed model and feature extraction, intra-class variability, uncertain atmospheric conditions, empirical seasonal patterns, which make the selection of such methods more difficult. ...
Article
Full-text available
Understanding the use of current land cover, along with monitoring change over time, is vital for agronomists and agricultural agencies responsible for land management. The increasing spatial and temporal resolution of globally available satellite images, such as provided by Sentinel-2, creates new possibilities for researchers to use freely available multi-spectral optical images, with decametric spatial resolution and more frequent revisits for remote sensing applications such as land cover and crop classification (LC&CC), agricultural monitoring and management, environment monitoring. Existing solutions dedicated to cropland mapping can be categorized based on per-pixel based and object-based. However, it is still challenging when more classes of agricultural crops are considered at a massive scale. In this paper, a novel and optimal deep learning model for pixel-based LC&CC is developed and implemented based on Recurrent Neural Networks (RNN) in combination with Convolutional Neural Networks (CNN) using multi-temporal sentinel-2 imagery of central north part of Italy, which has diverse agricultural system dominated by economic crop types. The proposed methodology is capable of automated feature extraction by learning time correlation of multiple images, which reduces manual feature engineering and modeling crop phenological stages. Fifteen classes, including major agricultural crops, were considered in this study. We also tested other widely used traditional machine learning algorithms for comparison such as support vector machine SVM, random forest (RF), Kernal SVM, and gradient boosting machine, also called XGBoost. The overall accuracy achieved by our proposed Pixel R-CNN was 96.5%, which showed considerable improvements in comparison with existing mainstream methods. This study showed that Pixel R-CNN based model offers a highly accurate way to assess and employ time-series data for multi-temporal classification tasks.
... In many studies, the onset of spring was usually analysed and modelled by satellite data, phenological Fabian 1999, D'Odorico et al 2002) and meteorological observations. For example, Xin et al (2015) tested 9 phenological models to predict the timing of grassland spring onset and found that most of them produce biased errors and would misrepresent the timing of vegetation spring onsets. Model development is always constrained by the lack of long-term records from ground phenological observation. ...
Article
Full-text available
The onset of spring is crucial for planning agricultural practices and has influences on animal activities as well. It is no doubt beneficial to properly forecast spring onset in advance. In this work, we present a new method based on harmonic analysis of daily mean temperature to predict the date of spring onset in the next year. The algorithm of the proposed method considers the memory of the seasonal cycle and can be easily conducted using the local past records. This study is based on gridded observational surface air temperature (SAT) in Europe. The SAT data are first decomposed into harmonics by wavelet transform and each harmonic’s time-dependent amplitude and phase are extracted. Then the time evolution for amplitude and phase are modelled by an AR(2) process, where we employ a new method to fit its coefficients from the data. This provides a prediction of the time dependent amplitudes and phases in the next year, from which we compose the future seasonal cycle and identify the prediction of spring onset by threshold crossing. We compare our model with a classical climatological seasonal cycle forecast. While this benchmark forecast leads to only very small variations of the predicted onset date, our method yields predictions which vary by about 15 days from year to year. We verify the correctness of our predictions by a correlation measure and the root mean squared errors.
... 2c,d), the correlation between SOSMCD12Q2 and SOS was significant (r = 0.74, p < 0.01), although the root mean square error (RMSE) of SOS was smaller than that of SOSMCD12Q2. Compared with our SOS, SOSMCD12Q2 showed an overall earlier trend that generally shifted its peak forward, consistent with previous studies (Vintrou et al., 2014;Wang et al., 2017;Xin et al., 2015). Similarly, there was a good correlation between EOSMCD12Q2 and EOS (r = 0.68, p < 0.01), with an RMSE for the latter of 4.03 d. ...
Article
Most studies on vegetation phenology along the urban–rural gradient (URG) have focused on inland cities, with a comparative lack of research on coastal cities despite their different climatic background. We used the normalized difference vegetation index (NDVI), land surface temperature (LST), and land cover data to determine spatiotemporal patterns in vegetation phenology with respect to LST along the URG in China’s coastal Dalian sub-province, with a focus on the main city of Dalian and three sub-cities (Pulandian, Wafangdian, and Zhuanghe). Our results were well-correlated with MODIS Land Cover Dynamics Product (MCD12Q2) reference data and matched patterns found in previous studies, indicating that the amplitude method of TIMESAT for obtaining vegetation phenology is practical. Start of growing season (SOS) and end of growing season (EOS) of urban areas were earlier and later than rural areas, respectively. The four urban areas had dissimilar vegetation types and urbanization levels leading to different changes in SOS and EOS along the URG; the average △SOS (the difference in SOS along the URG) and △EOS (the difference in EOS along the URG) of the main and sub-cities were 7.4 and 5.0 d, respectively. Changes in LST along the URG exhibited a non-linear relationship, with the maximum usually appearing 6–8 km from the urban areas. There was a strong linear relationship between vegetation phenology and LST along the URG. The winter–spring and yearly LSTs were negatively correlated with SOS, with both having roughly similar effects. The fall and yearly LSTs had significantly positive correlations with EOS, with the latter having a stronger effect. This study will be helpful for understanding climatic changes arising from urbanization in coastal areas and improving the management and productivity of the ecological environment.
... In recent years, aerospace and unmanned aerial vehicle remote sensing technology has rapidly developed , thus providing regional and even global remote sensing image data. With longer time series, broad coverage, high spatial resolution and easier access, this remote sensing data has been increasingly used in the study of vegetation and phenological evolution (Song et al., 2015;Xin et al., 2015). Consequently, this study used long time series satellite data to investigate the vegetation dynamics of the Loess Plateau. ...
Article
Purpose The ecological environment of the Loess Plateau, China, is extremely fragile under the context of global warming. Over the past two decades, the vegetation of the Loess Plateau has undergone great changes. This paper aims to clarify the response mechanisms of vegetation to climate change, to provide support for the restoration and environmental treatment of vegetation on the Loess Plateau. Design/methodology/approach The Savitsky–Golay (S-G) filtering algorithm was used to reconstruct time series of moderate resolution imaging spectroradiometer (MODIS) 13A2 data. Combined with trend analysis and partial correlation analysis, the influence of climate change on the phenology and enhanced vegetation index (EVI) during the growing season was described. Findings The S-G filtering algorithm is suitable for EVI reconstruction of the Loess Plateau. The date of start of growing season was found to gradually later along the Southeast–Northwest direction, whereas the date of the end of the growing season showed the opposite pattern and the length of the growing season gradually shortened. Vegetation EVI values decreased gradually from Southeast to Northwest. Vegetation changed significantly and showed clear differentiation according to different topographic factors. Vegetation correlated positively with precipitation from April to July and with temperature from August to November. Originality/value This study provides technical support for ecological environmental assessment, restoration of regional vegetation coverage and environmental governance of the Loess Plateau over the past two decades. It also provides theoretical support for the prediction model of vegetation phenology changes based on remote sensing data.
... As the Community Land Model accounts for as much as 17 plant functional types in modeling the land surface processes but only three vegetation types (i.e., evergreen, deciduous, and stressed-deciduous vegetation) in modeling phenology (Oleson et al., 2013), the phenology module requires substantial improvement. Recent studies attempted to use both remote sensing and large-scale meteorological data to calibrate and evaluate multiple GDD-based phenology models on modeling the timing of key phenophases across vegetation types Xin et al., 2015). ...
Article
Full-text available
Abstract Vegetation leaf phenology, often reflected by the dynamics in leaf area index (LAI), influences a variety of land surface processes. Robust models of vegetation phenology are pivot components in both land surface models and dynamic global vegetation models but remain challenging in terms of the model accuracy. This study develops a semiprognostic phenology model that is suitable for simulating time series of vegetation LAI. This method establishes a linear relationship between the steady‐state LAI (i.e., the LAI when the environment conditions remain unchanging) and gross primary productivity, meaning that the LAI an unchanging environment can carry is proportional to the photosynthetic products produced by plant leaves and implements with a simple light use efficiency algorithm of MOD17 to form a closed set of equations. We derive an analytical solution based on the Lambert W function to the closed equations and then apply a simple restricted growth process model to simulate the time series of actual LAI. The results modeled using global climate data demonstrate that the model is able to capture both the spatial pattern and intra‐annual and interannual variation of LAI derived from the satellite‐based product on a global scale. The results modeled using the flux tower data suggest that the developed model is able to explain over 70% variation in daily LAI for each plant functional type except evergreen broadleaf forest. The developed semiprognostic approach provides a simple solution to modeling the spatiotemporal variation in vegetation LAI across plant functional types on the global scale.
... However, precipitation becomes a key factor of the growth of plants in desert systems. This conforms the conclusion of Xin et al. (2015) that water has a stronger influence on green vegetation in arid and semiarid regions. In addition, elevation variability is strongly linked to phenological and climatic trends. ...
Article
Vegetation is highly sensitive to climate changes in arid regions. The relationship between vegetation and climate changes can be effectively characterized by vegetation phenology. However, few studies have examined the vegetation phenology and productivity changes in arid Central Asia (ACA). The vegetation phenological information of ACA was extracted using MODIS NDVI (Normalized Difference Vegetation Index) data, and the dynamics of vegetation phenological changes under spatiotemporal variations were quantitatively assessed. Moreover, the impacts of climate change on vegetation phenology and net primary productivity were analyzed by combining meteorological data with that of MODIS NPP (Net Primary Productivity) during the same period. The results demonstrated that the start of the season (SOS) of vegetation in the study was concentrated from mid-February to mid-April, while the end of the season (EOS) was concentrated from early October to mid-December. The length of growing season (LOS) ranged from 6 to 10 months. The SOS of vegetation was gradually postponed at a rate of 0.16 d·year⁻¹. The EOS advanced at a rate of 0.69 d·year⁻¹. The LOS was gradually shortened at a rate of 0.89 d·year⁻¹. For each per 1000 m increase in elevation, the SOS of vegetation was postponed by 12.40 d; the EOS advanced by 0.40 d, and the LOS was shortened by 11.70 d. For the impacts of climate changes on vegetation phenology and NPP, the SOS of vegetation phenology negatively correlated with temperature but positively correlated with precipitation and NPP. The EOS and LOS positively correlated with temperature but negatively with precipitation and NPP. Results indicated that the SOS was not moved ahead but was delayed, while the EOS advanced rather than being postponed under climate change. These results can offer new insights on the phenological response to climate change in arid regions and on non-systematic changes in phenology under global warming.
... Historically, most of statistical and mechanistic phenology models were developed for tree species, rather than non-woody species (Chuine et al. 2013). Despite huge progress in phenological modeling in the recent decades, the potential of modeling non-woody and non-agricultural plants with estimation of cross-scalar phenology was still underestimated or applied sporadically Xin et al. 2015). The approach presented in this paper shows the moderate-to-high potential of using machine learning models to fill temporal and spatial gaps in ground-based observations as well as forecasting selected phenological phases by means of remotely sensed and meteorologically based products. ...
Article
Full-text available
Changes in the timing of plant phenological phases are important proxies in contemporary climate research. However, most of the commonly used traditional phenological observations do not give any coherent spatial information. While consistent spatial data can be obtained from airborne sensors and preprocessed gridded meteorological data, not many studies robustly benefit from these data sources. Therefore, the main aim of this study is to create and evaluate different statistical models for reconstructing, predicting, and improving quality of phenological phases monitoring with the use of satellite and meteorological products. A quality-controlled dataset of the 13 BBCH plant phenophases in Poland was collected for the period 2007–2014. For each phenophase, statistical models were built using the most commonly applied regression-based machine learning techniques, such as multiple linear regression, lasso, principal component regression, generalized boosted models, and random forest. The quality of the models was estimated using a k-fold cross-validation. The obtained results showed varying potential for coupling meteorological derived indices with remote sensing products in terms of phenological modeling; however, application of both data sources improves models’ accuracy from 0.6 to 4.6 day in terms of obtained RMSE. It is shown that a robust prediction of early phenological phases is mostly related to meteorological indices, whereas for autumn phenophases, there is a stronger information signal provided by satellite-derived vegetation metrics. Choosing a specific set of predictors and applying a robust preprocessing procedures is more important for final results than the selection of a particular statistical model. The average RMSE for the best models of all phenophases is 6.3, while the individual RMSE vary seasonally from 3.5 to 10 days. Models give reliable proxy for ground observations with RMSE below 5 days for early spring and late spring phenophases. For other phenophases, RMSE are higher and rise up to 9–10 days in the case of the earliest spring phenophases.
... Remote sensing-based techniques are able to successfully monitor large-scale ecosystem changes and have been frequently applied in recent years [18,73], but there are some uncertainties along with the data management in this study. GIMMS datasets possess long-time-series observation records for land surface changes as a benefit, but their rough spatial resolution (0.5 degree, about 8 km) is too large to match the field investigation because vegetation phenology varies with vegetation type and soil texture, even in an 8 km-pixel area. ...
Article
Full-text available
Both vegetation phenology and net primary productivity (NPP) are crucial topics under the background of global change, but the relationships between them are far from clear. In this study, we quantified the spatial-temporal vegetation start (SOS), end (EOS), and length (LOS) of the growing season and NPP for the temperate grasslands of China based on a 34-year time-series (1982–2015) normalized difference vegetation index (NDVI) derived from global inventory modeling and mapping studies (GIMMS) and meteorological data. Then, we demonstrated the relationships between NPP and phenology dynamics. The results showed that more than half of the grasslands experienced significant changes in their phenology and NPP. The rates of their changes exhibited spatial heterogeneity, but their phenological changes could be roughly divided into three different clustered trend regions, while NPP presented a polarized pattern that increased in the south and decreased in the north. Different trend zones’ analyses revealed that phenology trends accelerated after 1997, which was a turning point. Prolonged LOS did not necessarily increase the current year’s NPP. SOS correlated with the NPP most closely during the same year compared to EOS and LOS. Delayed SOS contributed to increasing the summer NPP, and vice versa. Thus, SOS could be a predictor for current year grass growth. In view of this result, we suggest that future studies should further explore the mechanisms of SOS and plant growth.
... The phenology of the North American grasslands is known to be sensitive to seasonal and interannual changes in climatic variables, such as drought, temperature, and precipitation (Lesica and Kittelson 2010;Reed 2006;Cui et al. 2017). A variety of satellitederived datasets, including AVHRR NDVI (Li and Guo 2012;Bradley et al. 2007;Reed et al. 1994) and MODIS VIs (Dye et al. 2016;Xin et al. 2015), have been used to detect LSP changes in North American grasslands in response to climate change. However, validation of the accuracy of satellite LSP metrics remains uncertain. ...
Article
Ground validation of satellite-based vegetation phenology has been challenging because ground phenology data are sparsely distributed and mostly observed from limited numbers of plant species at discrete phenophases. The recently developed PhenoCam network has measured continuous growth of vegetation canopy greenness that can be used to validate satellite-based vegetation phenology across a variety of plant functional types. In this study, we used PhenoCam green chromatic coordinate (GCC) in North America to evaluate grassland phenology derived from three types of MODIS vegetation indices: the normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), and a per-pixel GCC (GCCpp) which was computed to describe the average vegetation color at the pixel level. The start of greenness (SOG), end of greenness (EOG), and length of greenness (LOG), and the dates for detailed seasonal dynamics for each site-year were compared. Our results indicate that MODIS VIs can be used to predict phenological metrics and seasonal dynamics in grassland greenness measured from PhenoCam GCC. More importantly, we quantified the difference between SOG, EOG, and LOG and seasonality estimated from satellite and near-surface remote sensing and discovered that GCCpp may be more suitable than NDVI and EVI at estimating dynamics in grassland greenness during senescence.
... For ENF and MF with relatively stable greenness, EVI failed to accurately estimate SOS and EOS, and LST became the more important variable for modeling SOS and EOS. The phenological indicators for ENF and GRA sites were more difficult to estimate than those for DBF sites, which was consistent with the results reported by Xin et al. (2015) and Wu et al. (2017). ...
Article
Vegetation phenology such as the start (SOS) and end (EOS) of the growing season, physiology (represented by seasonal maximum capacity of carbon uptake, GPPmax), and gross primary production (GPP) are sensitive indicators for monitoring ecosystem response to environmental change. However, uncertainty and disagreement between models limit the use phenology metrics and GPP derived from remote sensing data. Statistical models for estimating phenology and physiology were constructed based on key predictor variables derived from enhanced vegetation index (EVI) and land surface temperature (LST) data. Then, a statistical model that integrated remote sensing-based phenology and physiology (RS-SMIPP) data was constructed to estimate seasonal and annual GPP. These models were calibrated and validated with GPP observations from 512 site-years of FLUXNET data covering four plant functional types (PFTs) in the northern hemisphere: deciduous broadleaf forest, evergreen needle-leaf forest, mixed forest, and grassland. Our results showed that phenology and physiology were accurately estimated with relative root mean squared error (RMSEr) <20%, and the errors varied among the PFTs. Spring EVI was an important factor in explaining variation of GPPmax. The RS-SMIPP model outperformed the MOD17 algorithm in accurately estimating seasonal and annual GPP and reduced RMSEr from 25.34%–43.44% to 9.53%–26.19% for annual GPP of the different PFTs. These findings demonstrate that remote sensing-based phenological and physiological indicators could be used to explain the variations of seasonal and annual GPP, and provide an efficient way for improving GPP estimations at a global scale.
... Various researchers have used time series remote sensing data to study the impact of climate change on vegetation dynamics at regional and global scales using AVHRR, MODIS, SPOT images (Moulin et al. 1997;Zhang et al. 2001;Maignan et al. 2008;Zoffoli et al. 2008;Stöckli et al. 2011;Bala et al. 2013;Verhegghen et al. 2014;Forkel et al. 2015;Wang et al. 2016;Kumar et al. 2018;Bagchi et al. 2017). Umpteen number of studies have been carried out in the world to understand the grassland vegetation dynamics including those in Australia (Ma et al. 2013), USA (Pau and Still 2014;Xin et al. 2015), Africa (Guan et al. 2014), Europe (Fontana et al. 2008), Asia (Xia et al. 2014;Hou et al. 2014;Tasumi et al. 2014;Sha et al. 2016;Qamer et al. 2016). Only a few studies have focused on the grasslands located in different parts of India such as alpine grasslands of Western and Eastern Himalayan regions (Lal et al. 1991), Banni Grasslands of Gujarat (Jadhav et al. 1993), Shola Grasslands of Western Ghats (Babu et al. 1997), semi-arid Savanna Grasslands (Vanak et al. 2015), and these studies are limited to mapping the extent of this landscape using remote sensing technology. ...
Article
The present study has analysed grassland phenology: start of greening (SOG), end of greening (EOG) and length of greening (LOG), and their rate of change in the western Himalaya in India (Himachal Pradesh) using MODIS NDVI time series data (2001–2015). These metrics were inspected at different stratification levels: state, elevation, climatic zones and bio-geographic provinces. Delayed SOG was observed over 44.87% (P \ 0.1), and delayed EOG over 63.3% (P \ 0.1) of grassland grids. LOG was shortened in 24.37% (P < 0.1) and extended in 58.04% (P < 0.1) of the grids. At the state level,when statistically significant pixels (SSP) and all the pixels (AP) are used (given as SSP:AP), SOG is delayed by 20.27:6.28 days year-15, while EOG is delayed by 38.02:14.97 days year-15 and LOG is extended by 35.07:8.70 year-15 days. Extended LOG is observed over the temperate and cold arid regions, and shortened LOG is observed over sub-alpine and alpine regions. Variations in SOG and EOG are not uniform across different climatic and bio-geographic regions.However, in the sub-alpine and alpine zones, SOG and EOG followed elevation gradients, i.e. late SOG with early EOG over higher elevations, and early SOG with late EOG over lower elevations. Our study has revealed an interesting pattern of translational phenology (i.e. late SOG and late EOG) of grasslands which hints towards shifting winter period. Overall, itis observed that variations in timing of snowfall and snow cover extent are the reasons for inter-annual variations in the grassland phenology. The Weblink: https://rdcu.be/btrVL
... Even in areas with high vegetation cover and plants with distinct signals, phenology may be impossible to detect some years due to little to no plant productivity from inadequate precipitation. These limitations cause drylands to be excluded in many large-scale LSP analyses [14][15][16][17][18], while other analysis include them without any regard for their detectability and resulting bias [19,20]. Most studies focusing on drylands evaluate aggregate or peak annual VI as opposed to distinct seasonal transitions [13,21] and even then can occasionally have inconclusive results [22,23]. ...
Article
Land surface phenology (LSP) enables global-scale tracking of ecosystem processes, but its utility is limited in drylands due to low vegetation cover and resulting low annual amplitudes of vegetation indices (VIs). Due to the importance of drylands for biodiversity, food security, and the carbon cycle, it is necessary to understand the limitations in measuring dryland dynamics. Here, using simulated data and multitemporal unmanned aerial vehicle (UAV) imagery of a desert shrubland, we explore the feasibility of detecting LSP with respect to fractional vegetation cover, plant functional types, VI uncertainty, and two different detection algorithms. Using simulated data, we found that plants with distinct VI signals, such as deciduous shrubs, can require up to 60% fractional cover to consistently detect LSP. Evergreen plants, with lower seasonal VI amplitude, require considerably higher cover and can have undetectable phenology even with 100% vegetation cover. Our evaluation of two algorithms showed that neither performed the best in all cases. Even with adequate cover, biases in phenological metrics can still exceed 20 days and can never be 100% accurate due to VI uncertainty from shadows, sensor view angle, and atmospheric interference. We showed how high-resolution UAV imagery enables LSP studies in drylands and highlighted important scale effects driven by within-canopy VI variation. With high-resolution imagery, the open canopies of drylands are beneficial as they allow for straightforward identification of individual plants, enabling the tracking of phenology at the individual level. Drylands thus have the potential to become an exemplary environment for future LSP research.
... By tracking the pixel-based EVI over time, the phenological metric -PGSwas used to discriminate the grass functional type pixel by pixel. Xin et al. (2015) reported that in histograms for spring onset anomalies of grasslands in the western United States, the frequency of pixel numbers was normally distributed. In our study, hypothetically, the PGS histogram of our study area per C 3 or C 4 season follows (or resembles) a Gaussian distribution (also known as normal distribution) as well. ...
Article
Species composition is a key determinant of grassland ecosystem function and resilience. Climate change is predicted to alter the distribution of cool season (C3) and warm season (C4) grasses, however, the lack of spatial distributions and temporal variations of grass functional type information severely limits our understanding of climate impacts on grasslands. This study classified C3 and C4 grasses per pixel according to the peak of growing season generated from Enhanced Vegetation Index time series. From 2003 to 2017, the C3-C4 composition of Australian rain-fed grasslands and pastures was mapped at 500 m resolution on an annual basis across a wide geographical range (10°S – 45°S), and revealed extreme inter-annual fluctuations. Over the 15-year period, the satellite-derived ratio of C4 to C3 grasses significantly increased (p < 0.05), indicating a long-term shift in community composition that was confirmed with 182,911 Atlas of Living Australia ground observations. The most pronounced changes occurred in mid-latitude transitional areas where C3 and C4 grasses co-dominate. Our climate analysis indicated the inter-annual fluctuations of C4/C3 grass ratios were significantly associated (p < 0.05) with warm/cool season rainfall ratios, and not with temperature or annual rainfall. This suggests that an increase in the warm/cool season rainfall ratio favors C4 grasses and a decrease in the warm/cool season rainfall ratio favors C3 grasses. Our findings reveal spatially-detailed dynamics of grasslands and demonstrate large-scale grassland compositional changes over 15 years. The grass composition maps should help improve ecological forecasting of grass distributions and enable researches on grassland ecosystem responses to climate change that are relevant to both adaptation of rangeland agricultural and fire management practices. Our study should also help predict grass distribution under future climate conditions, and assist in the accurate modelling of global water, carbon, and energy exchanges between the land surface and atmosphere.
... Among the curve-fitting functions, the logistic/sigmoid function proposed by Badhwar (1984) has gained popularity for its robustness and convenience to derive phenological features in characterizing vegetation dynamics and growing cycles (X. Zhang et al., 2003, Fisher et al., 2006, Beck et al., 2006, Fisher and Mustard, 2007, Dannenberg et al., 2015, Xin et al., 2015, Gonsamo and Chen, 2016. ...
Article
This study aims to develop a deep learning based classification framework for remotely sensed time series. The experiment was carried out in Yolo County, California, which has a very diverse irrigated agricultural system dominated by economic crops. For the challenging task of classifying summer crops using Landsat Enhanced Vegetation Index (EVI) time series, two types of deep learning models were designed: one is based on Long Short-Term Memory (LSTM), and the other is based on one-dimensional convolutional (Conv1D) layers. Three widely-used classifiers were also tested for comparison, including a gradient boosting machine called XGBoost, Random Forest, and Support Vector Machine. Although LSTM is widely used for sequential data representation, in this study its accuracy (82.41%) and F1 score (0.67) were the lowest among all the classifiers. Among non-deep-learning classifiers, XGBoost achieved the best result with 84.17% accuracy and an F1 score of 0.69. The highest accuracy (85.54%) and F1 score (0.73) were achieved by the Conv1D-based model, which mainly consists of a stack of Conv1D layers and an inception module. The behavior of the Conv1D-based model was inspected by visualizing the activation on different layers. The model employs EVI time series by examining shapes at various scales in a hierarchical manner. Lower Conv1D layers of the optimized model capture small scale temporal variations, while upper layers focus on overall seasonal patterns. Conv1D layers were used as an embedded multi-level feature extractor in the classification model which automatically extracts features from input time series during training. The automated feature extraction reduces the dependency on manual feature engineering and pre-defined equations of crop growing cycles. This study shows that the Conv1D-based deep learning framework provides an effective and efficient method of time series representation in multi-temporal classification tasks.
... Furthermore, it is interesting that the growth peak of NDVI in the whole watershed occurs in spring and autumn rather than summer, from March to May, NDVI growth reached its first peak (p < 0.05) i.e., the re-greening stage of plants, during the same period, where precipitation showed a significant increasing trend (p < 0.05), and the temperature from January to March showed a significant increasing trend (p < 0.05), which indicates that moisture and temperature in early spring are the dominant climatic factors in plant re-greening period in the study area. The accumulation of temperature in the early stage and the increase of available water is beneficial to breaking dormancy, stimulating cell division, improving photosynthetic efficiency and carbohydrate accumulation, and providing sufficient energy for plant growth; this is consistent with the results of previous studies in arid regions [43,44]. Instead, during the critical period of plants growth and development, May to August, NDVI tends to decrease throughout the watershed, this period, there was a significant downward trend in temperature (p < 0.05), precipitation showed a significant increase (p < 0.05), SPEI changed from drought to wetness; this indicates that summer drought and low temperatures limit plant growth. ...
Article
Full-text available
The response of plants to climate change has become a topical issue. However, there is no consensus on the synergistic processes of the canopy and trunk growth within different vegetation types, or on the consistency of the response of the canopy and trunk to climate change. This paper is based on Normalized Difference Vegetation Index (NDVI), tree-ring width index (TRW) and climate data from the Irtysh River basin, a sensitive area for climate change in Central Asia. Spatial statistical methods and correlation analysis were used to analyze the spatial and temporal trends of plants and climate, and to reveal the differences in the canopy and trunk response mechanisms to climate within different vegetation types. The results show a warming and humidifying trend between 1982 and 2015 in the study area, and NDVI and TRW increases in different vegetation type zones. On an interannual scale, temperature is the main driver of the canopy growth in alpine areas and precipitation is the main limiting factor for the canopy growth in lower altitude valley and desert areas. The degree of response of the trunk to climatic factors decreases with increasing altitude, and TRW is significantly correlated with mean annual temperature, precipitation and SPEI in desert areas. On a monthly scale, the earlier and longer growing season due to the accumulation of temperature and precipitation in the early spring and late autumn periods contributes to two highly significant trends of increase in the canopy from March to May and August to October. Climatic conditions during the growing season are the main limiting factor for the growth of the trunk, but there is considerable variation in the driving of the trunk in different vegetation type zones. The canopy growth is mainly influenced by climatic factors in the current month, while there is a 1–2-month lag effect in the response of the trunk to climatic factors. In addition, the synergy between the canopy and the trunk is gradually weakened with increasing altitude (correlation coefficient is 0.371 in alpine areas, 0.413 in valley areas and 0.583 in desert areas). These findings help to enrich the understanding of the response mechanisms to climate change in different vegetation type zones and provide a scientific basis for the development of climate change response measures in Central Asia.
... The GDD model assumes that plant leaf onset begins when daily mean temperatures accumulated from a fixed date reach a critical threshold. Studies have identified that various environmental factors other than temperature could affect plant phenology to certain degrees (Polgar and Primack, 2011), and therefore, efforts have been made to improve the GDD model by adding different influential factors, such as photoperiod, soil temperature, humidity, and soil moisture (Melaas et al., 2013;Chuine et al., 1999;Xin et al., 2015a;Yang et al., 2012). Land surface models like the Community 30 ...
Article
Full-text available
Terrestrial plants play a key role in regulating the exchange of energy and materials between the land surface and the atmosphere. Robust terrestrial biosphere models that simulate both time series of leaf dynamics and canopy photosynthesis are required to understand the vegetation-climate interactions. This study proposes a time stepping scheme to simulate leaf area index (LAI), phenology, and gross primary production (GPP) simultaneously via only climate variables based on an ecological assumption that plants allocate leaf biomass till an environment could sustain to maximize photosynthetic reproduction. The method establishes a linear function between the steady-state LAI and the corresponding GPP, which is used to track the suitability of environmental conditions for plant photosynthesis, and applies the MOD17 algorithm to form simultaneous equations together, which can be solved numerically. To account for the time lag in plant responses of leaf allocation to environment variation, a time stepping scheme is developed to simulate the LAI time series based on the solved steady-state LAI. The simulated LAI time series is then used to derive the timing of key phenophases and simulate canopy GPP with the MOD17 algorithm. The developed method is applied to deciduous broadleaf forests in eastern United States and has found to perform well on simulating canopy LAI and GPP at the site scale as evaluated using both flux tower and satellite data. The method could also capture the spatiotemporal variation of vegetation LAI and phenology across eastern United States as compared with satellite observations. The developed time-stepping scheme provides a simplified and improved version of our previous modeling approach and forms a potential basis for regional to global applications in future studies.
... Advanced Very High Resolution Radiometer (AVHRR; 1981+)-NDVI data have been extensively used to retrieve vegetation phenology at landscape scales (Barichivich et al., 2013;Cong et al., 2013;Jeong et al., 2011;Maignan et al., 2008;Shen et al., 2014;Wang et al., 2011;White et al., 2009). After 1999, Moderate Resolution Imaging Spectroradiometer (MODIS; 2000+)-NDVI data with finer spatial resolution and better data quality are broadly employed to extract vegetation phenology at different scales (Delbart et al., 2015;Xin et al., 2015;Yang et al., 2012). Therefore, combining NDVI data from AVHRR and MODIS is an alternative way to obtain long-term and high-quality remote sensing data for land surface monitoring. ...
Article
Full-text available
Revealing grassland growing season spatial patterns and their climatic controls is crucial for estimating the spatial heterogeneity of grassland productivity and carbon sequestration. In this study, we first used satellite-derived normalized difference vegetation index data and a double logistic function to extract the start (SOS), end (EOS), and length of the growing season in midlatitude grasslands of the Northern Hemisphere during 1981–2014. Then, we verified the accuracy of satellite-derived SOS and EOS using ground-observed phenological records and gross primary production data at some locations. Moreover, we analyzed the spatial patterns of growing season indicators and their climatic controls. Results show that both SOS and EOS appear first in cool semidesert grasslands (CG), then in temperate grasslands (TG) and alpine grasslands (AG), and finally in warm semidesert grasslands (WG). A delaying tendency of SOS and EOS from north to south was identified in TG of North America. In contrast, an advancing tendency of SOS and EOS from north to south was detected in CG of Central and Western Asia. Further analysis indicates that a spatial opposite effect of spring temperature and precipitation triggers SOS in TG, whereas a spatial synergy effect of spring temperature and precipitation triggers SOS in CG of Asia, WG, and AG. Meanwhile, a spatial synergy effect of autumn temperature and precipitation triggers EOS for TG of North America and AG, whereas a spatial opposite effect of autumn temperature and precipitation determines EOS for CG.
... Adequate water supply usually promoted early plant growth in arid and semi-arid regions under suitable high-temperature conditions [32]. In water-deficient regions, climate warming exacerbated drought stress and delayed plant growth [17,58]. Therefore, precipitation pattern appears to be a dominant climatic driver of interannual variation of plant phenology in arid and semi-arid regions [20,24], with temperature triggering the SOS only when the water supply is adequate [56]. ...
Article
Full-text available
Plant phenological variations depend largely on temperature, but they cannot be explained by temperature alone in arid and semi-arid regions. To reveal the response mechanisms of grassland phenology to climate change, the effects of temperature, moisture and light at the start (SOS), peak (POS) and end (EOS) of the growing season for Stipa krylovii (S. krylovii) in Inner Mongolian grassland was analysed from 1985–2018 with partial least squares (PLS) regression. The results showed that the SOS was significantly delayed at a rate of 5.4 d/10a (change over 10 years), while POS and EOS were insignificantly advanced, which were inconsistent with the existing understanding that climate warming advances the SOS and delays the EOS. The vapor pressure deficit (VPD) in July, maximum air temperature (Tmax) in September of the previous year, diurnal temperature range (DTR) from mid-February to mid-March, and Tmax from late March to mid-April of the current year were the critical factors and periods triggering the SOS, which contributed to 68.5% of the variation in the SOS. Additionally, the minimum air temperature (Tmin) occurred from mid-December to late December, and precipitation (PRE) occurred from mid-June to late July for POS, which could explain 52.1% of POS variations. In addition, Tmax from late August to early September influenced the EOS with an explanation of 49.3%. The results indicated that the phenological variations in S. krylovii were the result of the combined effects of climatic conditions from the previous year and the current year. Additionally, an increase in the preseason DTR delayed the SOS, and excessive summer precipitation induced an earlier POS, while warming in early autumn induced an earlier EOS, reflecting the adaptation mechanism of the perennial dense-cluster herbaceous plants in semi-arid regions to climate change. These findings could enrich the understanding of plant phenology in response to climate change.
... Further, time series of VI represented by a set of functions [Galford(2008)], linear regression [Funk(2009)], Markov model [Siachalou(2015)] and curve-fitting functions. Sigmoid function has been exploited by [Xin(2015), Xin(2016)], and achieved better results due to its robustness and ease to derive phenological features for the characterization of vegetation variability [Dannenberg(2015)]. Although above-mentioned methods of temporal feature extraction offer many alternatives and flexibilities in deployment to assess vegetation dynamics, in practice, there are some important factors such as manually designed model and feature extraction, intra-class variability, uncertain atmospheric conditions, empirical seasonal patterns, which make the selection of such methods more difficult. ...
Preprint
Full-text available
The increasing spatial and temporal resolution of globally available satellite images, such as provided by Sentinel-2, creates new possibilities for researchers to use freely available multi-spectral optical images, with decametric spatial resolution and more frequent revisits for remote sensing applications such as land cover and crop classification (LC&CC), agricultural monitoring and management, environment monitoring. Existing solutions dedicated to cropland mapping can be categorized based on per-pixel based and object-based. However, it is still challenging when more classes of agricultural crops are considered at a massive scale. In this paper, a novel and optimal deep learning model for pixel-based LC&CC is developed and implemented based on Recurrent Neural Networks (RNN) in combination with Convolutional Neural Networks (CNN) using multi-temporal sentinel-2 imagery of central north part of Italy, which has diverse agricultural system dominated by economic crop types. The proposed methodology is capable of automated feature extraction by learning time correlation of multiple images, which reduces manual feature engineering and modeling crop phenological stages. Fifteen classes, including major agricultural crops, were considered in this study. We also tested other widely used traditional machine learning algorithms for comparison such as support vector machine SVM, random forest (RF), Kernal SVM, and gradient boosting machine, also called XGBoost. The overall accuracy achieved by our proposed Pixel R-CNN was 96.5%, which showed considerable improvements in comparison with existing mainstream methods. This study showed that Pixel R-CNN based model offers a highly accurate way to assess and employ time-series data for multi-temporal classification tasks.
Article
Plant phenology is one of the main controls of variation in net ecosystem productivity (NEP). Accurate representation of vegetation phenology is important for predicting ecosystem carbon budget. Although using satellite observation to determine vegetation phenology is becoming a mature option, there are still uncertainties in its application at site scales. Our purpose is to establish a more robust phenological index to accurately predict carbon uptake phenology, which detailed results can complement the shortcomings of MODIS NDVI-derived phenology. Here we used a growing season index (GSI) phenology model to simulate carbon flux phenology (CFP) including the start of carbon uptake (CUstart), the end of carbon uptake (CUend) and the length of carbon uptake period (CUP) in an alpine meadow ecosystem on the Tibetan Plateau and to compare the results with those modeled from MODIS NDVI. We also further analyzed the main environmental factors in controlling CFP. The results indicated that the GSI model made substantially more precise prediction for CUstart, CUend and CUP (with higher correlation R²>0.90) than that of the MODIS derived phenology. The GSI model was also superior to NDVI in predicting both seasonal and annual variations of net ecosystem productivity (NEP). Moreover, CUP played an important role in regulating ecosystem carbon balance in the study site because NEP was significantly positive correlated with the period of annual carbon uptake. NEP would increase by 1.63 g C m⁻² year⁻¹ if one CUP-day was extended. Further, CUP was influenced by variation in CUstart. Previously overlooked water variability (soil water content and VPD) played a significant role in controlling CUP and CUstart. In addition, temperature could enhance water stress to delay CUstart and shorten CUP. It is indicated that decrease in carbon uptake could be induced by accelerative water stress in the face of global warming in the alpine meadow. These results suggest that CFP is more sensitive to not only temperature but also water condition, and a combination of soil water and temperature could be a useful way to enhance the estimation of CFP in future ecosystem model.
Chapter
Satellite remote sensing observations have the potential for efficient and reliable mapping of spatial soil moisture distributions. However, soil moisture retrievals from active/passive microwave remote sensing data are typically complex due to inherent difficulty in characterizing interactions among land surface parameters that contribute to the retrieval process. Therefore, adequate physical mathematical descriptions of microwave backscatter interaction with parameters such as land cover, vegetation density, and soil characteristics are not readily available. In such condition, non-parametric models could be used as a possible alternative for better understanding the impact of variables in the retrieval process and relating it in the absence of exact formulation. The following chapter will provide a conceptual framework for SMPM: A non-parametric approach over an agriculture land.
Chapter
In soil-landscape parameters mapping, the implementation of geomatics-GIS, GPS, remote sensing, and DEM, suggests new alternatives. Different approaches have been applied for retrieval of soil-landscape parameters. In recent years, machine learning algorithms have received increasing attention for digital mapping.
Article
Phenology is an important indicator of the biological responses to climate change, and previous studies have reported divergent phenological responses of grasslands in Inner Mongolia from the analyses of in-situ measurements and satellite observations; thus, robust phenology models are required to understand and simulate the relationship between ecosystems and climate change. However, phenology models for natural grasslands in which heating accumulation, chilling requirements, precipitation effects and other climate factors are introduced at the same time are relatively deficient; as a result, the accuracy of the herbaceous plant phenology model and its universality across different species and regions are restricted. We proposed an improved phenology model (unified model combined with precipitation driving, UMPD) that reflects the combined effect of temperature and precipitation at the forcing unit level. Green-up evaluation based on satellite observations shows that the UMPD phenology model achieves better accuracy than previous models (RMSE < 10 days), and the simulated spatial distribution of green-up date change is more consistent with the ground observation, which means that the inclusion of the precipitation driving effect in the temperature-based phenology model improves model applicability at a regional scale and under complex climatic conditions. Based on the UMPD model and its parameters, spatial and temporal changes in green-up dates for Leymus chinensis steppe in Inner Mongolia from 1962 to 2017 were simulated, and the results showed an overall advancing trend with obvious spatial differences and complex stage characteristics. We demonstrate the feasibility of combining temperature and precipitation driving effects to develop and refine phenology models, and the improved model provides wide potential for characterizing the spatiotemporal patterns of grassland green-up date variations.
Article
Aims Variations in vegetation spring phenology are widely attributed to temperature in temperate and cold regions. However, temperature effect on phenology remains elusive in cold and arid/semiarid ecosystems because soil water condition also plays an important role in mediating phenology. Methods We used growing degree day (GDD) model and growing season index (GSI) model, coupling minimum temperature (Tmin) with soil moisture (SM) to explore the influence of heat requirement and hydroclimatic interaction on the start of carbon uptake period (SCUP) and net ecosystem productivity (NEP) in two alpine meadows with different precipitation regimes on the Qinghai-Tibet Plateau (QTP). One is the water-limited alpine steppe-meadow, and the other is the temperature-limited alpine shrub-meadow. Important Findings We observed two clear patterns linking GDD and GSI to SCUP: SCUP was similarly sensitive to variations in preseason GDD and GSI in the humid alpine shrub-meadow, while SCUP was more sensitive to the variability in preseason GSI than GDD in the semiarid alpine steppe-meadow. The divergent patterns indicated a balance of the limiting climatic factors between temperature and water availability. In the humid meadow, higher temperature sensitivity of SCUP could maximize thermal benefit without drought stress, as evidenced by the stronger linear correlation coefficient (R2) and Akaike’s information criterion (AIC) between observed SCUPs and those of simulated by GDD model. However, greater water sensitivity of SCUP could maximize the benefit of water in semiarid steppe-meadow, which is indicated by the stronger R2 and AIC between observed SCUPs and those of simulated by GSI model. Additionally, although SCUPs were determined by GDD in the alpine shrub-meadow ecosystem, NEP was both controlled by accumulative GSI in two alpine meadows. Our study highlights the impacts of hydroclimatic interaction on spring carbon flux phenology and vegetation productivity in the humid and semiarid alpine ecosystems. The results also suggest that water, together with temperature should be included in the models of phenology and carbon budget for alpine ecosystems in semiarid regions. These findings have important implications for improving vegetation phenology models, thus advancing our understanding of the interplay between vegetation phenology, productivity and climate change in future.
Article
Remote sensing has become an important technique for monitoring vegetation phenology. The quality of remote-sensing images and derived products is key to successful extraction of vegetation phenophases. There is a need to develop quality management methods to evaluate the data uncertainty and assist the removal of the noises. This paper developed a shape quality assurance score threshold (SQAT) method which accounts for the trend in satellite-derived vegetation index associated with the process of vegetation growth. The proposed method was tested on six widely used methods for extracting vegetation phenophases. Results showed that the SQAT method can effectively identify noises in the vegetation index time series and improve the accuracies of estimated start of season (SOS) and end of season (EOS) of the six methods. After removal of identified noises, the Pearson correlation coefficient (r) averagely increased by 8% for SOS, and 11% for EOS. Regression analyses of vegetation phenophases between the PhenoCam observations and MCD12Q2 product showed that the proposed method performs better than the QA score of MCD12Q2 for quality management. This paper provides promising method for quality management; it has the potential to reduce the uncertainty of the vegetation index time series that can support studies of vegetation phenology monitoring.
Article
Full-text available
Accurate estimation of the timing of intensive spring leaf growth initiation at mid and high latitudes is crucial for improving the predictive capacity of biogeochemical and Earth system models. In this study, we focus on the modeling of climatological onset of spring leaf growth in Central Europe and use three spring phenology models driven by three meteorological datasets. The MODIS-adjusted NDVI3g dataset was used as a reference for the period between 1982 and 2010, enabling us to study the long-term mean leaf onset timing and its interannual variability (IAV). The performance of all phenology model–meteorology database combinations was evaluated with one another, and against the reference dataset. We found that none of the constructed model–database combinations could reproduce the observed start of season (SOS) climatology within the study region. The models typically overestimated IAV of the leaf onset, where spatial median SOS dates were best simulated by the models based on heat accumulation. When aggregated for the whole study area, the complex, bioclimatic index-based model driven by the CarpatClim database could capture the observed overall SOS trend. Our results indicate that the simulated timing of leaf onset primarily depends on the choice of model structure, with a secondary contribution from the choice of the driving meteorological dataset.
Article
Full-text available
Aim Vegetation phenology that characters the periodic life cycles of plants is indicative of the interactions between the biosphere and the atmosphere. Robust modelling of vegetation phenology metrics that correspond to canopy development events is essential to our understanding of how plants and ecosystems respond to a changing climate. Given considerable uncertainties associated with vegetation phenology modelling using numerical models, we explore the deep learning approach to predicting the timing of global vegetation phenology metrics. Location Global. Time period 2001–2015. Major taxa studied Deciduous vegetation (DV), stressed deciduous vegetation (SDV), evergreen vegetation (EV). Methods We developed a one-dimensional convolutional neural network regression (1D-CNNR) model with 10 hierarchical structures to model global vegetation phenology using meteorological variables as inputs. The developed deep learning model was evaluated using satellite-derived phenology metrics (i.e., green-up, maturity, senescence, and dormancy) and compared with the terrestrial ecosystem model Biome-BGC (BioGeochemical Cycles). Results Our experimental results show that the 1D-CNNR model well captures both the spatial pattern and inter-annual variation of satellite-derived multiyear vegetation phenology metrics on a global scale. The median root-mean-square errors (RMSEs) and standard deviations between phenology metrics derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) data and predicted by the 1D-CNNR model on a global scale from 2001 to 2015 are 4.1 ± 5.9, 4.2 ± 12.1, 3.0 ± 6.8, and 3.4 ± 4.3 days for green-up, maturation, senescence, and dormancy, respectively, for the DV type; 13.3 ± 29.6, 8.4 ± 29.1, 8.1 ± 21.3, and 9.1 ± 21.6 days for green-up, maturation, senescence, and dormancy, respectively, for the SDV type; and 13.9 ± 17.4, 17.7 ± 34.6, 18.8 ± 42.9, and 12.1 ± 17.7 days for green-up, maturation, senescence, and dormancy, respectively, for the EV type. Main conclusions This research demonstrates that the 1D-CNNR model has the potential for large-scale modelling of vegetation phenology. Results from the deep learning model suggest that there is room to improve numerical vegetation phenology models for use in land surface models.
Article
We investigate the temporal dynamics of shifts in phenological responses of a range of key stages of the growing season in New Zealand’s three indigenous grassland types over the last 16 years (2001–2016). A near-daily Normalized Difference Vegetation Index (NDVI) time series from MODerate Resolution Imaging Spectroradiometer (MODIS) was used to extract five annual growth phenology indices, namely the Start, End, Length, Peak and Peak NDVI of a growing season. The start of the growing season advanced (i.e. happened earlier) by a median of 7.2, 6.0 and 8.8 days per decade in Alpine, Tall Tussock and Low Producing grassland, whereas the end of the season advanced by a median of 4.5, 0.4 and 0.4 days in the three types respectively. The length of growing season was extended by 3.2, 5.2 and 7.1 days per decade in these three grassland types. Over 86% of the investigated grassland areas showed an advancing (earlier) start of the growing season, and 74% of Alpine grassland showed a trend toward an earlier end of season. Over 63% of all grassland types showed an increase in growing season length. A trend toward earlier growing season peak and overall increasing NDVI in the three grassland types indicate a tendency for increasing vegetation vitality in grassland ecosystems in recent years. The start of growing season was correlated with atmospheric pressure (negatively) and precipitation (positively) changes in winter–spring months, while the timing of the season end is positively correlated with air temperature and solar radiation in summer–autumn months. Our study shows that different grassland types differ in magnitude – but not in direction – of their recent shifts in timing of key growing season stages with high-alpine grasslands showing the strongest response. This study highlights the usefulness of remote sensing for monitoring ecosystem-level phenological shifts over large areas and long time periods.
Article
Full-text available
The timing of vegetation spring onset is largely influenced by climate factors, making it sensitive to climate variation. Robust models that predict vegetation spring onset via the climate forcing data are needed in the land surface models for understanding the impacts of climate change on vegetation processes. In this study, we apply and assess both numerical phenology models and the machine learning models on predicting the timing of spring onset for different vegetation types, including deciduous vegetation, evergreen vegetation and stressed deciduous vegetation. We perform model calibration for numerical phenology models and machine learning models using both in-situ observations of spring onset dates from National Phenology Network in the United States and satellite-derived green-up dates in the Northern Hemisphere. Our experiment showed better performance of numerical models calibrated by ground phenology observations. Among all the numerical phenology models, the models developed based on Growing Season Index perform well on predicting the spring onsets of deciduous vegetation and stressed deciduous vegetation across the Northern Hemisphere. Machine learning models if trained appropriately could also capture the spatial variation of satellite-derived spring onset dates. Our study highlights the need of improvements on numerical phenology models for their uses in the land surface models. We also illustrate the benchmarking role of the machine learning models on predicting vegetation spring onsets via climate variables.
Article
Full-text available
Pixels, blocks (i.e., grouping of pixels), and polygons are the fundamental choices for use as assessment units for validating per-pixel image classification. Previous research conducted by the authors of this paper focused on the analysis of the impact of positional accuracy when using a single pixel for thematic accuracy assessment. The research described here provided a similar analysis, but the blocks of contiguous pixels were chosen as the assessment unit for thematic validation. The goal of this analysis was to assess the impact of positional errors on the thematic assessment. Factors including the size of a block, labeling threshold, landscape characteristics, spatial scale, and classification schemes were also considered. The results demonstrated that using blocks as an assessment unit reduced the thematic errors caused by positional errors to under 10% for most global land-cover mapping projects and most remote-sensing applications achieving a half-pixel registration. The larger the block size, the more the positional error was reduced. However, there are practical limitations to the size of the block. More classes in a classification scheme and higher heterogeneity increased the positional effect. The choice of labeling threshold depends on the spatial scale and landscape characteristics to balance the number of abandoned units and positional impact. This research suggests using the block of pixels as an assessment unit in the thematic accuracy assessment in future applications.
Article
Different species within the same community may exhibit distinct phenological responses to climate change, so it is necessary to study species differences in the green-up date among abundant species within a wide area, and a suitable phenology model should be introduced to explain the associated climate-driven mechanism. Although various models have been developed, very few studies have aimed to compare their efficiency and robustness, and the relative contributions of climate driving factors have not been sufficiently examined. We analyzed phenology data for 12 species across 17 stations in Inner Mongolia and found that essential spatiotemporal and interspecies differences existed in the green-up date. Five process-based models were established for each species and their performance was comprehensively evaluated. The two-phase models (sequential model, parallel model, unified model and unified model combined with precipitation driving) generally performed better than the one-phase model (thermal time model), and the model considering precipitation performed the best, which indicates that it is necessary to introduce the chilling effect and precipitation driving effect to improve the model accuracy in arid environments. We proposed a method to estimate the contribution rates of various climate driving factors, and significant differences in the relative demand for the various climate driving factors among different species were clearly revealed. The results indicated that for natural vegetation in Inner Mongolia, the need for the chilling and temperature driving is relatively high, and the precipitation driving is very important for herbaceous vegetation, which leads to considerable spatial and interspecies differences in green-up date. We demonstrated the feasibility of quantitatively evaluating the contributions of different climate driving factors with a process-based model, and the contradiction in phenological changes among different studies may eventually be clarified.
Article
Owing to climate change and frequent extreme weather events, changes in spring flowering phenology have been observed in temperate forests. The flowering time response to climate change is divergent among species and is difficult to predict due to the complexity of flowering mechanisms. To compare the effects of spring warming, winter chilling, and day length on spring flowering time, we evaluated eight process-based models (two types of forcing models, two types of chilling-forcing models, and four models with the effect of day length added to the aforementioned four models). We used flowering data of seven temperate species (Cornus officinalis, Rhododendron mucronulatum, Forsythia koreana, Prunus yedoensis, Rhododendron yedoense var. poukhanense, Rhododendron schlippenbachii, and Robinia pseudoacacia) observed in nine different arboretums in South Korea over 9 years. Generally, the forcing model performed better than the sequential chilling-forcing model, regardless of the species. The performance gap between the models was reduced when day length term was included in model, but the chilling-forcing model did not outperform the forcing model. The effect of day length on flowering time differed depending on the species. Prunus yedoensis, which had a particularly low warming sensitivity compared to other species, was more dependent on day length than other species. On the other hand, day length had little effect on the flowering time of Robinia pseudoacacia and Cornus officinalis, mostly found in the early successional stage. These findings imply that the effect of chilling on flowering time would be minor for the seven species inhabiting the warm-temperate forest, and the effect of day length on flowering time was species-specific and dependent on species' temperature (warming) sensitivity and life strategy. In the future warm climate, the flowering time of day length sensitive species would not advance significantly, which may result in a phenological mismatch and endanger their life.
Article
Understanding climatic effects on cropland water use efficiency at different elevations is imperative for managing agricultural water and production in response to ongoing climate change in climate-sensitive areas with complex topography, such as southwestern China. We investigated climatic effects on cropland water use efficiency in southwestern China at each 100-m elevation bin during 2001–2017. The maximum water use efficiency was 1.71 gC kg−1 H2O for the 1900–1999 m elevation bin under the growing season temperature and precipitation of 14.58±0.32°C and 965.40±136.45 mm, respectively. The water use efficiency slopes were dominated by the evapotranspiration slopes at elevations below 1999 m but were controlled by the gross primary productivity slopes at elevations above 2000 m. This difference was caused by the substantial responses of evaporation to climate change at lower elevations and the increased climatic sensitivity of gross primary productivity at higher elevations. In comparison to those at other elevations, croplands at lower elevations were more vulnerable to extreme drought because of the dominant role fluctuating evapotranspiration played in water use efficiency. The findings will improve cropland water management in the study area.
Article
Full-text available
The primary goal of thematic accuracy assessment is to measure the quality of land cover products and it has become an essential component in global or regional land cover mapping. However, there are many uncertainties introduced in the validation process which could propagate into the derived accuracy measures and therefore impact the decisions made with these maps. Choosing the appropriate reference data sample unit is one of the most important decisions in this process. The majority of researchers have used a single pixel as the assessment unit for thematic accuracy assessment, while others have claimed that a single pixel is not appropriate. The research reported here shows the results of a simulation analysis from the perspective of positional errors. Factors including landscape characteristics, the classification scheme, the spatial scale, and the labeling threshold were also examined. The thematic errors caused by positional errors were analyzed using the current level of geo-registration accuracy achieved by several global land cover mapping projects. The primary results demonstrate that using a single-pixel as an assessment unit introduces a significant amount of thematic error. In addition, the coarser the spatial scale, the greater the impact on positional errors as most pixels in the image become mixed. A classification scheme with more classes and a more heterogeneous landscape increased the positional effect. Using a higher labeling threshold decreased the positional impact but greatly increased the number of abandoned units in the sample. This research showed that remote sensing applications should not employ a single-pixel as an assessment unit in the thematic accuracy assessment.
Article
Full-text available
Water availability is one of the most important environmental controls on vegetation phenology, especially in semi-arid regions. It is often represented in terms of soil moisture in small-scale studies, whereas it tends to be represented by precipitation in large-scale (e.g., regional) studies. Clearly, soil moisture is the more appropriate indicator for root water uptake and vegetation growth/phenology. Its potential advantage and applicability needs to be demonstrated at regional scales. The paper presents a data-based regional study of the effectiveness of novel water and temperature-based indices to predict spring vegetation green-up dates based on the Normalized Difference Vegetation Index (NDVI) observations in the grasslands of Inner Mongolia, China. The macro-scale hydrological model, VIC (Variable Infiltration Capacity), is employed to generate a soil moisture database across the region. In addition to a standard index based on temperature, two potential hydrology-based indices for prediction of spring onset dates are defined, based on the simulated soil moisture data as well as on observed precipitation data. Results indicate that the correspondence between the NDVI-derived green-up onset date and the soil-moisture-derived potential onset date exhibits a significantly better correlation as a function of increasing aridity compared to that based on precipitation. In this way the soil-moisture-based index is demonstrated to be superior to the precipitation-based index in terms of capturing grassland spring phenology. The results also showed that both of the hydrological (water-based) indices were superior to the thermal (temperature-based) index in determining the patterns of grass green-up in the Inner Mongolia region, indicating water availability to be the dominant control on average. The understanding about the relative controls on grassland phenology and the effectiveness of alternative indices to capture these controls are important for future studies of vegetation phenology change under climate change.
Article
Full-text available
Land surface phenological cycles of vegetation greening and browning are influenced by variability in climatic forcing. Quantitative spatial information on phenological cycles and their variability is important for agricultural applications, wildfire fuel accumulation, land management, land surface modeling, and climate change studies. Most phenology studies have focused on temperature-driven Northern Hemisphere systems, where phenology shows annually recurring patterns. However, precipitation-driven non-annual phenology of arid and semi-arid systems (i.e., drylands) received much less attention, despite the fact that they cover more than 30% of the global land surface. Here, we focused on Australia, a continent with one of the most variable rainfall climates in the world and vast areas of dryland systems, where a detailed phenological investigation and a characterization of the relationship between phenology and climate variability are missing. To fill this knowledge gap, we developed an algorithm to characterize phenological cycles, and analyzed geographic and climate-driven variability in phenology from 2000 to 2013, which included extreme drought and wet years. We linked derived phenological metrics to rainfall and the Southern Oscillation Index (SOI). We conducted a continent-wide investigation and a more detailed investigation over the Murray–Darling Basin (MDB), the primary agricultural area and largest river catchment of Australia. Results showed high inter- and intra-annual variability in phenological cycles across Australia. The peak of phenological cycles occurred not only during the austral summer, but also at any time of the year, and their timing varied by more than a month in the interior of the continent. The magnitude of the phenological cycle peak and the integrated greenness were most significantly correlated with monthly SOI within the preceding 12 months. Correlation patterns occurred primarily over northeastern Australia and within the MDB, predominantly over natural land cover and particularly in floodplain and wetland areas. Integrated greenness of the phenological cycles (surrogate of vegetation productivity) showed positive anomalies of more than 2 standard deviations over most of eastern Australia in 2009–2010, which coincided with the transition from the El Niño-induced decadal droughts to flooding caused by La Niña.
Article
Full-text available
By the end of this century, mean annual temperatures in the Northeastern United States are expected to warm by 3–5 °C, which will have significant impacts on the structure and function of temperate forests in this region. To improve understanding of these impacts, we exploited two recent climate anomalies to explore how the springtime phenology of Northeastern temperate deciduous forests will respond to future climate warming. Specifically, springtime temperatures in 2010 and 2012 were the warmest on record in the Northeastern United States, with temperatures that were roughly equivalent to the lower end of warming scenarios that are projected for this region decades from now. Climate conditions in these two years therefore provide a unique empirical basis, that complements model-based studies, for improving understanding of how northeastern temperate forest phenology will change in the future. To perform our investigation, we analyzed near surface air temperatures from the United States Historical Climatology Network, time series of satellite-derived vegetation indices from NASA's Moderate Resolution Imaging Spectroradiometer, and in situ phenological observations. Our study region encompassed the northern third of the eastern temperate forest ecoregion, extending from Pennsylvania to Canada. Springtime temperatures in 2010 and 2012 were nearly 3 °C warmer than long-term average temperatures from 1971–2000 over the region, leading to median anomalies of more than 100 growing degree days. In response, satellite and ground observations show that leaf emergence occurred up to two weeks earlier than normal, but with significant sensitivity to the specific timing of thermal forcing. These results are important for two reasons. First, they provide an empirical demonstration of the sensitivity of springtime phenology in northeastern temperate forests to future climate change that supports and complements model-based predictions. Second, our results show that subtle differences in the character of thermal forcing can substantially alter the timing of leaf emergence and canopy development. By explicitly comparing and contrasting the timing of thermal forcing and leaf phenology in 2010 and 2012, we show that even though temperatures were warmer in 2012 than in 2010, the nature and timing of thermal forcing in 2010 lead to leaf emergence that was almost a week earlier than 2012.
Article
Full-text available
A key question for climate change adaptation is whether existing cropping systems can become less sensitive to climate variations. We use a field-level data set on maize and soybean yields in the central United States for 1995 through 2012 to examine changes in drought sensitivity. Although yields have increased in absolute value under all levels of stress for both crops, the sensitivity of maize yields to drought stress associated with high vapor pressure deficits has increased. The greater sensitivity has occurred despite cultivar improvements and increased carbon dioxide and reflects the agronomic trend toward higher sowing densities. The results suggest that agronomic changes tend to translate improved drought tolerance of plants to higher average yields but not to decreasing drought sensitivity of yields at the field scale.
Article
Full-text available
The dual-source Shuttleworth-Wallace model has been widely used to estimate and partition crop evapotranspiration (λET). Canopy stomatal conductance (Gsc), an essential parameter of the model, is often calculated by scaling up leaf stomatal conductance, considering the canopy as one single leaf in a so-called "big-leaf" model. However, Gsc can be overestimated or underestimated depending on leaf area index level in the big-leaf model, due to a non-linear stomatal response to light. A dual-leaf model, scaling up Gsc from leaf to canopy, was developed in this study. The non-linear stomata-light relationship was incorporated by dividing the canopy into sunlit and shaded fractions and calculating each fraction separately according to absorbed irradiances. The model includes: (1) the absorbed irradiance, determined by separately integrating the sunlit and shaded leaves with consideration of both beam and diffuse radiation; (2) leaf area for the sunlit and shaded fractions; and (3) a leaf conductance model that accounts for the response of stomata to PAR, vapor pressure deficit and available soil water. In contrast to the significant errors of Gsc in the big-leaf model, the predicted Gsc using the dual-leaf model had a high degree of data-model agreement; the slope of the linear regression between daytime predictions and measurements was 1.01 (R2 = 0.98), with RMSE of 0.6120 mm s-1 for four clear-sky days in different growth stages. The estimates of half-hourly λET using the dual-source dual-leaf model (DSDL) agreed well with measurements and the error was within 5% during two growing seasons of maize with differing hydrometeorological and management strategies. Moreover, the estimates of soil evaporation using the DSDL model closely matched actual measurements. Our results indicate that the DSDL model can produce more accurate estimation of Gsc and λET, compared to the big-leaf model, and thus is an effective alternative approach for estimating and partitioning λET.
Article
Full-text available
Until recently, advanced very high-resolution radiometer (AVHRR) observations were the only viable source of data for global land cover mapping. While many useful insights have been gained from analyses based on AVHRR data, the availability of moderate resolution imaging spectroradiometer (MODIS) data with greatly improved spectral, spatial, geometric, and radiometric attributes provides significant new opportunities and challenges for remote sensing-based land cover mapping research. In this paper, we describe the algorithms and databases being used to produce the MODIS global land cover product. This product provides maps of global land cover at 1-km spatial resolution using several classification systems, principally that of the IGBP. To generate these maps, a supervised classification methodology is used that exploits a global database of training sites interpreted from high-resolution imagery in association with ancillary data. In addition to the IGBP class at each pixel, the MODIS land cover product provides several other parameters including estimates for the classification confidence associated with the IGBP label, a prediction for the most likely alternative class, and class labels for several other classification schemes that are used by the global modeling community. Initial results based on 5 months of MODIS data are encouraging. At global scales, the distribution of vegetation and land cover types is qualitatively realistic. At regional scales, comparisons among heritage AVHRR products, Landsat TM data, and results from MODIS show that the algorithm is performing well. As a longer time series of data is added to the processing stream and the representation of global land cover in the site database is refined, the quality of the MODIS land cover product will improve accordingly.
Article
Full-text available
China has the largest afforested area in the world (∼62 million hectares in 2008), and these forests are carbon sinks. The climatic effect of these new forests depends on how radiant and turbulent energy fluxes over these plantations modify surface temperature. For instance, a lower albedo may cause warming, which negates the climatic benefits of carbon sequestration. Here, we used satellite measurements of land surface temperature (LST) from planted forests and adjacent grasslands or croplands in China to understand how afforestation affects LST. Afforestation is found to decrease daytime LST by about 1.1 ± 0.5 °C (mean ± 1 SD) and to increase nighttime LST by about 0.2 ± 0.5 °C, on average. The observed daytime cooling is a result of increased evapotranspiration. The nighttime warming is found to increase with latitude and decrease with average rainfall. Afforestation in dry regions therefore leads to net warming, as daytime cooling is offset by nighttime warming. Thus, it is necessary to carefully consider where to plant trees to realize potential climatic benefits in future afforestation projects.
Article
Full-text available
Procedures that may be used to evaluate the operational performance of a wide spectrum of geophysical models are introduced. Primarily using a complementary set of difference measures, both model acccuracy and precision can be meaningfully estimated, regardless of whether the model predicitons are manifested as scalars, directions, or vectors. It is additionally suggested that the reliability of the accuracy and precision measures can be determined from bootstrap estimates of confidence and significance. Recommended procedures are illustrated with a comparative evaluation of two models that estimate wind velocity over the South Atlantic Bight.
Article
Full-text available
This is the second part of a study on continental-scale water and energy flux analysis and validation conducted in phase 2 of the North American Land Data Assimilation System project (NLDAS-2). The first part concentrates on a model-by-model comparison of mean annual and monthly water fluxes, energy fluxes and state variables. In this second part, the focus is on the validation of simulated streamflow from four land surface models (Noah, Mosaic, Sacramento Soil Moisture Accounting (SAC-SMA), and Variable Infiltration Capacity (VIC) models) and their ensemble mean. Comparisons are made against 28-years (1 October 1979-30 September 2007) of United States Geological Survey observed streamflow for 961 small basins and 8 major basins over the conterminous United States (CONUS). Relative bias, anomaly correlation and Nash-Sutcliffe Efficiency (NSE) statistics at daily to annual time scales are used to assess model-simulated streamflow. The Noah (the Mosaic) model overestimates (underestimates) mean annual runoff and underestimates (overestimates) mean annual evapotranspiration. The SAC-SMA and VIC models simulate the mean annual runoff and evapotranspiration well when compared with the observations. The ensemble mean is closer to the mean annual observed streamflow for both the 961 small basins and the 8 major basins than is the mean from any individual model. All of the models, as well as the ensemble mean, have large daily, weekly, monthly, and annual streamflow anomaly correlations for most basins over the CONUS, implying strong simulation skill. However, the daily, weekly, and monthly NSE analysis results are not necessarily encouraging, in particular for daily streamflow. The Noah and Mosaic models are useful (NSE > 0.4) only for about 10% of the 961 small basins, the SAC-SMA and VIC models are useful for about 30% of the 961 small basins, and the ensemble mean is useful for about 42% of the 961 small basins. As the time scale increases, the NSE increases as expected. However, even for monthly streamflow, the ensemble mean is useful for only 75% of the 961 small basins.
Article
Full-text available
Results are presented from the second phase of the multiinstitution North American Land Data Assimilation System (NLDAS-2) research partnership. In NLDAS, the Noah, Variable Infiltration Capacity, Sacramento Soil Moisture Accounting, and Mosaic land surface models (LSMs) are executed over the conterminous U.S. (CONUS) in realtime and retrospective modes. These runs support the drought analysis, monitoring and forecasting activities of the National Integrated Drought Information System, as well as efforts to monitor large-scale floods. NLDAS-2 builds upon the framework of the first phase of NLDAS (NLDAS-1) by increasing the accuracy and consistency of the surface forcing data, upgrading the land surface model code and parameters, and extending the study from a 3-year (1997-1999) to a 30-year (1979-2008) time window. As the first of two parts, this paper details the configuration of NLDAS-2, describes the upgrades to the forcing, parameters, and code of the four LSMs, and explores overall model-to-model comparisons of land surface water and energy flux and state variables over the CONUS. Focusing on model output rather than on observations, this study seeks to highlight the similarities and differences between models, and to assess changes in output from that seen in NLDAS-1. The second part of the two-part article focuses on the validation of model-simulated streamflow and evaporation against observations. The results depict a higher level of agreement among the four models over much of the CONUS than was found in the first phase of NLDAS. This is due, in part, to recent improvements in the parameters, code, and forcing of the NLDAS-2 LSMs that were initiated following NLDAS-1. However, large inter-model differences still exist in the northeast, Lake Superior, and western mountainous regions of the CONUS, which are associated with cold season processes. In addition, variations in the representation of sub-surface hydrology in the four LSMs lead to large differences in modeled evaporation and subsurface runoff. These issues are important targets for future research by the land surface modeling community. Finally, improvement from NLDAS-1 to NLDAS-2 is summarized by comparing the streamflow measured from U.S. Geological Survey stream gauges with that simulated by four NLDAS models over 961 small basins.
Article
Full-text available
In temperate regions, the budburst date of deciduous trees is mainly regulated by temperature variation but the exact nature of the temperature dependence has been a matter of recent debate. One hypothesis is that budburst date depends purely on warm temperature; a competing hypothesis states that exposure to cold temperatures are also important. In this study, variability in budburst at a mixed deciduous forest is evaluated by using 15 years of budburst date data for 17 tree species at Harvard forest. We compared two widely-used budburst hypotheses through reversible jump Markov Chain Monte Carlo (RJMCMC). Then, the implications of uncertainty in our understanding of budburst were investigated using GFDL's LM3 model. For 15 of the 17 species, we found that more complicated models that included a chilling period were favored over a simpler model that did not include such dependence. Alternative ways of expressing the chilling period received nearly equal support from the observations. LM3 simulations with implementing several uncertainty factors in budburst variations (e.g., parameter, hypothesis, model structure, and model selection) showed that variations of budburst induced uncertainties in the timing of carbon uptake commencement up to 11 days and the magnitude of spring carbon uptake up to 1.03 g C m-2 day-1. The largest uncertainties are mainly attributed to differences of budburst parameter among species and of phenology model between dynamic vegetation models. Our results strongly suggest that exact assessment of variations in budburst is an important factor for reducing the uncertainties in terrestrial carbon cycle simulations.
Article
Full-text available
Phenology is critical to ecosystem carbon quantification, and yet has not been well modeled considering both aboveground and belowground environmental variables. This is especially true for alpine and pan-arctic regions where soil physical conditions play a significant role in determining the timing of phenology. Here we examine how the spatiotemporal pattern of satellite-derived phenology is related to soil physical conditions simulated with a soil physical model on the Tibetan Plateau for the period 1989–2008. Our results show that spatial patterns and temporal trends of phenology are parallel with the corresponding soil physical conditions for different study periods. On average, 1 °C increase in soil temperature advances the start of growing season (SOS) by 4.6 to 9.9 days among different vegetation types, and postpones the end of growing season (EOS) by 7.3 to 10.5 days. Soil wetting meditates such trends, especially in areas where warming effect is significant. Soil thermal thresholds for SOS and EOS, defined as the daily mean soil temperatures corresponding to the phenological metrics, are spatially clustered, and are closely correlated with mean seasonal temperatures in Spring and Autumn, respectively. This study highlights the importance and feasibility of incorporating spatially explicit soil temperature and moisture information, instead of air temperature and precipitation, into phenology models so as to improve carbon modeling. The method proposed and empirical relations established between phenology and soil physical conditions for Alpine ecosystems on the Tibetan plateau could also be applicable for other cold regions.
Article
Full-text available
Timely and accurate monitoring of forest disturbance is essential to help us understand how the Earth system is changing. MODIS (Moderate Resolution Imaging Spectroradiometer) imagery and subsequent MODIS products provide near-daily global coverage and have transformed the ways we study and monitor the Earth. To monitor forest disturbance, it is necessary to be able to compare observations of the same place from different times, but this is a challenging task using MODIS data as observations from different days have varying view angles and pixel sizes, and cover slightly different areas. In this paper, we propose a method to fuse MODIS and Landsat data in a way that allows for near real-time monitoring of forest disturbance. The method is based on using Landsat time-series images to predict the next MODIS image, which forms a stable basis for comparison with new MODIS acquisitions. The predicted MODIS images represent what the surface should look like assuming no disturbance, and the difference in the spectral signatures between predicted and observed MODIS images becomes the “signal” used for detecting forest disturbance. The method was able to detect subpixel forest disturbance with a producer's accuracy of 81% and a user's accuracy of 90%. Patches of forest disturbance as small as 5 to 7 ha in size were detected on a daily basis. The encouraging results indicate that the presented fusion method holds promise for improving monitoring of forest disturbance in near real-time.
Article
Full-text available
Climate, vegetation cover, and management create finescale heterogeneity in unirrigated agricultural regions, with important but not well-quantified consequences for spatial and temporal variations in surface CO2, water, and heat fluxes. Eddy covariance fluxes were measured in seven agricultural fields-comprising winter wheat, pasture, and sorghum - in the U.S. Southern Great Plains (SGP) during the 2001 - 03 growing seasons. Land cover was the dominant source of variation in surface fluxes, with 50%-100% differences between fields planted in winter - spring versus fields planted in summer. Interannual variation was driven mainly by precipitation, which varied more than twofold between years. Peak aboveground biomass and growing season net ecosystem exchange (NEE) of CO2 increased in rough proportion to precipitation. Based on a partitioning of gross fluxes with a regression model, ecosystem respiration increased linearly with gross primary production, but with an offset that increased near the time of seed production. Because the regression model was designed for well-watered periods, it successfully retrieved NEE and ecosystem parameters during the peak growing season and identified periods of moisture limitation during the summer. In summary, the effects of crop type, land management, and water limitation on carbon, water, and energy fluxes were large. Capturing the controlling factors in landscape-scale models will be necessary to estimate the ecological feedbacks to climate and other environmental impacts associated with changing human needs for agricultural production of food, fiber, and energy.
Article
Full-text available
The successful application of a conceptual rainfall-runoff (CRR) model depends on how well it is calibrated. Despite the popularity of CRR models, reports in the literature indicate that it is typically difficult, if not impossible, to obtain unique optimal values for their parameters using automatic calibration methods. Unless the best set of parameters associated with a given calibration data set can be found, it is difficult to determine how sensitive the parameter estimates (and hence the model forecasts) are to factors such as input and output data error, model error, quantity and quality of data, objective function used, and so on. Results are presented that establish clearly the nature of the multiple optima problem for the research CRR model SIXPAR. These results suggest that the CRR model optimization problem is more difficult than had been previously thought and that currently used local search procedures have a very low probability of successfully finding the optimal parameter sets. Next, the performance of three existing global search procedures are evaluated on the model SIXPAR. Finally, a powerful new global optimization procedure is presented, entitled the shuffled complex evolution (SCE-UA) method, which was able to consistently locate the global optimum of the SIXPAR model, and appears to be capable of efficiently and effectively solving the CRR model optimization problem.
Article
Full-text available
The two-layer variable infiltration capacity (VIC-2L) model is extended to incorporate a representation of subgrid variability in precipitation, using an analytical one- dimensional statistical dynamic representation for partial area coverage of precipitation. The analytical approach allows the effects of subgrid-scale spatial variability of precipitation on surface fluxes, runoff production, and soil moisture to be represented explicitly. With this method, spatially integrated representations of surface fluxes, runoff, and soil moisture due to subgrid-scale spatial variability in precipitation, infiltration, and vegetation cover are obtained. The results are compared with those obtained using an exhaustive pixel-based approach, and the results obtained by applying uniform precipitation over the precipitation- covered area. The precipitation coverage over a grid cell is shown to play a primary role in estimating the surface fluxes, runoff, and soil moisture. In general, the spatial distribution of precipitation within the precipitation-covered area plays a secondary role, in part because VIC-2L represents the subgrid spatial variability of soil properties. While the analytical approach gives good approximations to the pixel-based approach, and is superior to the uniform precipitation approach in general, the differences are not large.
Article
Full-text available
The relationship between vegetation and climate in the grassland and cropland of the northern US Great Plains was investigated with Normalized Difference Vegetation Index (NDVI) (1989–1993) images derived from the Advanced Very High Resolution Radiometer (AVHRR), and climate data from automated weather stations. The relationship was quantified using a spatial regression technique that adjusts for spatial autocorrelation inherent in these data. Conventional regression techniques used frequently in previous studies are not adequate, because they are based on the assumption of independent observations. Six climate variables during the growing season; precipitation, potential evapotranspiration, daily maximum and minimum air temperature, soil temperature, solar irradiation were regressed on NDVI derived from a 10-km weather station buffer. The regression model identified precipitation and potential evapotranspiration as the most significant climatic variables, indicating that the water balance is the most important factor controlling vegetation condition at an annual timescale. The model indicates that 46% and 24% of variation in NDVI is accounted for by climate in grassland and cropland, respectively, indicating that grassland vegetation has a more pronounced response to climate variation than cropland. Other factors contributing to NDVI variation include environmental factors (soil, groundwater and terrain), human manipulation of crops, and sensor variation.
Article
Full-text available
Climate, vegetation cover, and management create fine-scaleheterogeneity in unirrigated agricultural regions, with important but notwell-quantified consequences for spatial and temporal variations insurface CO2, water, and heat fluxes. We measured eddy covariance fluxesin seven agricultural fields--comprising winter wheat, pasture, andsorghum--in the U.S. Southern Great Plains (SGP) during the 2001-2003growing seasons. Land-cover was the dominant source of variation insurface fluxes, with 50-100 percent differences between fields planted inwinter-spring versus fields planted in summer. Interannual variation wasdriven mainly by precipitation, which varied more than two-fold betweenyears. Peak aboveground biomass and growing-season net ecosystem exchange(NEE) of CO2 increased in rough proportion to precipitation. Based on apartitioning of gross fluxes with a regression model, ecosystemrespiration increased linearly with gross primary production, but with anoffset that increased near the time of seed production. Because theregression model was designed for well-watered periods, it successfullyretrieved NEE and ecosystem parameters during the peak growing season,and identified periods of moisture limitation during the summer. Insummary, the effects of crop type, land management, and water limitationon carbon, water, and energy fluxes were large. Capturing the controllingfactors in landscape scale models will be necessary to estimate theecological feedbacks to climate and other environmental impactsassociated with changing human needs for agricultural production of food,fiber, and energy.
Article
Full-text available
Long-term global data sets of vegetation Leaf Area Index (LAI) and Fraction of Photosynthetically Active Radiation absorbed by vegetation (FPAR) are critical to monitoring global vegetation dynamics and for modeling exchanges of energy, mass and momentum between the land surface and planetary boundary layer. LAI and FPAR are also state variables in hydrological, ecological, biogeochemical and crop-yield models. The generation, evaluation and an example case study documenting the utility of 30-year long data sets of LAI and FPAR are described in this article. A neural network algorithm was first developed between the new improved third generation Global Inventory Modeling and Mapping Studies (GIMMS) Normalized Difference Vegetation Index (NDVI3g) and best-quality Terra Moderate Resolution Imaging Spectroradiometer (MODIS) LAI and FPAR products for the overlapping period 2000–2009. The trained neural network algorithm was then used to generate corresponding LAI3g and FPAR3g data sets with the following attributes: 15-day temporal frequency, 1/12 degree spatial resolution and temporal span of July 1981 to December 2011. The quality of these data sets for scientific research in other disciplines was assessed through (a) comparisons with field measurements scaled to the spatial resolution of the data products, (b) comparisons with broadly-used existing alternate satellite data-based products, (c) comparisons to plant growth limiting climatic variables in the northern latitudes and tropical regions, and (d) correlations of dominant modes of interannual variability with large-scale circulation anomalies such as the EI Niño-Southern Oscillation and Arctic Oscillation. These assessment efforts yielded results that attested to the suitability of these data sets for research use in other disciplines. The utility of these data sets is documented by comparing the seasonal profiles of LAI3g with profiles from 18 state-of-the-art Earth System Models: the models consistently overestimated the satellite-based estimates of leaf area and simulated delayed peak seasonal values in the northern latitudes, a result that is consistent with previous evaluations of similar models with ground-based data. The LAI3g and FPAR3g data sets can be obtained freely from the NASA Earth Exchange (NEX) website.
Article
Full-text available
As the Earth's third pole, the Tibetan Plateau has experienced a pronounced warming in the past decades. Recent studies reported that the start of the vegetation growing season (SOS) in the Plateau showed an advancing trend from 1982 t