Search

The effect of machine learning and other enhancements on statistical–dynamical forecasts of soil moisture (0–10 and 0–100 cm) and a reference evapotranspiration fraction [evaporative stress index (ESI)] on subseasonal time scales (15–28 days) are explored. The predictors include the current and past land surface conditions and dynamical model hindcasts from the Subseasonal to Seasonal Prediction project (S2S). When the methods are enhanced with machine learning and other improvements, the increases in skill are almost exclusively coming from predictors drawn from observations of current and past land surface states. This suggests that operational S2S flash drought forecasts should focus on optimizing use of information on current conditions rather than on integrating dynamically based forecasts, given the current state of knowledge. Nonlinear machine learning methods lead to improved skill over linear methods for soil moisture but not for ESI. Improvements for both soil moisture and ESI are realized by increasing the sample size by including surrounding grid points in training and increasing the number of predictors. In addition, all the improvements in the soil moisture forecasts predominantly impact soil moistening rather than soil drying—i.e., prediction of conditions moving away from drought rather than into drought—especially when the initial soil state is drier than normal. The physical reasons for the nonlinear machine learning improvements are also explored. Significance Statement: Rapidly intensifying droughts pose extra challenges for predictability. Here, dynamical forecast model output is combined with nonlinear machine learning methods to improve forecasts of rapid changes in soil moisture and the evaporative stress index (ESI). [ABSTRACT FROM AUTHOR]

In the Czech Republic, soil moisture content during the growing season has been decreasing over the past six decades, and drought events have become significantly more frequent. In 2003, 2015, 2018 and 2019, drought affected almost the entire country, with droughts in 2000, 2004, 2007, 2012, 2014 and 2017 having smaller extents but still severe intensities in some regions. The current methods of visiting cadastral areas (approximately 13,000) to allocate compensation funds for the crop yield losses caused by drought or aggregating the losses to district areas (approximately 1000 km 2 ) based on proxy data are both inappropriate. The former due to the required time and resources, the later due to low resolution, which leads to many falsely negative and falsely positive results. Therefore, the study presents a new method to combine ground survey, remotely sensed and model data for determining crop yield losses. The study shows that it is possible to estimate them at the cadastral area level in the Czech Republic and attribute those losses to drought. This can be done with remotely sensed vegetation, water stress and soil moisture conditions with modeled soil moisture anomalies coupled with near-real-time feedback from reporters and with crop status surveys. The newly developed approach allowed the achievement of a proportion of falsely positive errors of less than 10% (e.g., oat 2%, 8%; spring barley 4%, 3%; sugar beets 2%, 21%; and winter wheat 2%, 6% in years 2017, resp. 2018) and allowed for cutting the loss assessment time from eight months in 2017 to eight weeks in 2018. [ABSTRACT FROM AUTHOR]

This study examines the evolution of soil moisture, evapotranspiration, vegetation, and atmospheric conditions during an unusual flash drought–flash recovery sequence that occurred across the south-central United States during 2015. This event was characterized by a period of rapid drought intensification (flash drought) during late summer that was terminated by heavy rainfall at the end of October that eliminated the extreme drought conditions over a 2-week period (flash recovery). A detailed analysis was performed using time series of environmental variables derived from meteorological, remote sensing, and land surface modeling datasets. Though the analysis revealed a similar progression of cascading effects in each region, characteristics of the flash drought such as its onset time, rate of intensification, and vegetation impacts differed between regions due to variations in the antecedent conditions and the atmospheric anomalies during its growth. Overall, flash drought signals initially appeared in the near-surface soil moisture, followed closely by reductions in evapotranspiration. Total column soil moisture deficits took longer to develop, especially in the western part of the region where heavy rainfall during the spring and early summer led to large moisture surpluses. Large differences were noted in how land surface models in the North American Land Data Assimilation System depicted soil moisture evolution during the flash drought; however, the models were more similar in their assessment of conditions during the flash recovery period. This study illustrates the need to use multiple datasets to track the evolution and impacts of rapidly evolving flash drought and flash recovery events. [ABSTRACT FROM AUTHOR]

The evolution of a flash drought event, characterized by a period of rapid drought intensification, is assessed using standard drought monitoring datasets and on-the-ground reports obtained via a written survey of agricultural stakeholders after the flash drought occurred. The flash drought impacted agricultural production across a five-state region centered on the Black Hills of South Dakota during the summer of 2016. The survey asked producers to estimate when certain drought impacts, ranging from decreased soil moisture to plant stress and diminished water resources, first occurred on their land. The geographic distribution and timing of the survey responses were compared to the U.S. Drought Monitor and to datasets depicting anomalies in evapotranspiration, precipitation, and soil moisture. Overall, the survey responses showed that this event was a multifaceted drought that caused a variety of impacts across the region. Comparisons of the survey reports to the drought monitoring datasets revealed that the topsoil moisture dataset provided the earliest warning of drought development, but at the expense of a high false alarm rate. Anomalies in evapotranspiration were closely aligned to the survey reports of plant stress and also provided a more focused depiction of where the worst drought conditions were located. This study provides evidence that qualitative reports of drought impacts obtained via written surveys provide valuable information that can be used to assess the accuracy of high-resolution drought monitoring datasets. [ABSTRACT FROM AUTHOR]

Observations of land surface temperature (LST) are crucial for the monitoring of surface energy fluxes from satellite. Methods that require high temporal resolution LST observations (e.g., from geostationary orbit) can be difficult to apply globally because several geostationary sensors are required to attain near-global coverage (60°N to 60°S). While these LST observations are available from polar-orbiting sensors, providing global coverage at higher spatial resolutions, the temporal sampling (twice daily observations) can pose significant limitations. For example, the Atmosphere Land Exchange Inverse (ALEXI) surface energy balance model, used for monitoring evapotranspiration and drought, requires an observation of the morning change in LST-a quantity not directly observable from polar-orbiting sensors. Therefore, we have developed and evaluated a data-mining approach to estimate the midmorning rise in LST from a single sensor (two observations per day) of LST from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor on the Aqua platform. In general, the data-mining approach produced estimates with low relative error (5 to 10%) and statistically significant correlations when compared against geostationary observations. This approach will facilitate global, near-real-time applications of ALEXI at higher spatial and temporal coverage from a single sensor than currently achievable with current geostationary data sets. [ABSTRACT FROM AUTHOR]

The U.S. Drought Monitor (USDM) classifies drought into five discrete dryness/drought categories based on expert synthesis of numerous data sources. In this study, an empirical methodology is presented for creating a nondiscrete USDM index that simultaneously 1) represents the dryness/wetness value on a continuum and 2) is most consistent with the time scales and processes of the actual USDM. A continuous USDM representation will facilitate USDM forecasting methods, which will benefit from knowledge of where, within a discrete drought class, the current drought state most probably lies. The continuous USDM is developed such that the actual discrete USDM can be reconstructed by discretizing the continuous USDM based on the 30th, 20th, 10th, 5th, and 2nd percentiles-corresponding with USDM definitions for the D4- D0 drought classes. Anomalies in precipitation, soil moisture, and evapotranspiration over a range of different time scales are used as predictors to estimate the continuous USDM. The methodology is fundamentally probabilistic, meaning that the probability density function (PDF) of the continuous USDM is estimated and therefore the degree of uncertainty in the fit is properly characterized. Goodness-of-fit metrics and direct comparisons between the actual and predicted USDM analyses during different seasons and years indicate that this objective drought classification method is well correlated with the current USDM analyses. In Part II, this continuous USDM index will be used to improve intraseasonal USDM intensification forecasts because it is capable of distinguishing between USDM states that are either far from or near to the next-higher drought category. [ABSTRACT FROM AUTHOR]

Probabilistic forecasts of U.S. Drought Monitor (USDM) intensification over 2-, 4-, and 8-week time periods are developed based on recent anomalies in precipitation, evapotranspiration, and soil moisture. These statistical forecasts are computed using logistic regression with cross validation. While recent precipitation, evapotranspiration, and soil moisture do provide skillful forecasts, it is found that additional information on the current state of the USDM adds significant skill to the forecasts. The USDM state information takes the form of a metric that quantifies the 'distance' from the next-higher drought category using a nondiscrete estimate of the current USDM state. This adds skill because USDM states that are close to the next-higher drought category are more likely to intensify than states that are farther from this threshold. The method shows skill over most of the United States but is most skillful over the north-central United States, where the cross-validated Brier skill score averages 0.20 for both 2- and 4-week forecasts. The 8-week forecasts are less skillful in most locations. The 2- and 4-week probabilities have very good reliability. The 8-week probabilities, on the other hand, are noticeably overconfident. For individual drought events, the method shows the most skill when forecasting high-amplitude flash droughts and when large regions of the United States are experiencing intensifying drought. [ABSTRACT FROM AUTHOR]

Summary Shortwave vegetation index (VI) and leaf area index (LAI) remote sensing products yield inconsistent depictions of biophysical response to drought and pluvial events that have occurred in Brazil over the past decade. Conflicting reports of severity of drought impacts on vegetation health and functioning have been attributed to cloud and aerosol contamination of shortwave reflectance composites, particularly over the rainforested regions of the Amazon basin which are subject to prolonged periods of cloud cover and episodes of intense biomass burning. This study compares timeseries of satellite-derived maps of LAI from the Moderate Resolution Imaging Spectroradiometer (MODIS) and precipitation from the Tropical Rainfall Mapping Mission (TRMM) with a diagnostic Evaporative Stress Index (ESI) retrieved using thermal infrared remote sensing over South America for the period 2003–2013. This period includes several severe droughts and floods that occurred both over the Amazon and over unforested savanna and agricultural areas in Brazil. Cross-correlations between absolute values and standardized anomalies in monthly LAI and precipitation composites as well as the actual-to-reference evapotranspiration (ET) ratio used in the ESI were computed for representative forested and agricultural regions. The correlation analyses reveal strong apparent anticorrelation between MODIS LAI and TRMM precipitation anomalies over the Amazon, but better coupling over regions vegetated with shorter grass and crop canopies. The ESI was more consistently correlated with precipitation patterns over both landcover types. Temporal comparisons between ESI and TRMM anomalies suggest longer moisture buffering timescales in the deeper rooted rainforest systems. Diagnostic thermal-based retrievals of ET and ET anomalies, such as used in the ESI, provide independent information on the impacts of extreme hydrologic events on vegetation health in comparison with VI and precipitation-based drought indicators, and used in concert may provide a more reliable evaluation of natural and managed ecosystem response to variable climate regimes. [ABSTRACT FROM AUTHOR]

Abstract: Drought assessment is a complex undertaking, requiring monitoring of deficiencies in multiple components of the hydrologic budget. Precipitation anomalies reflect variability in water supply to the land surface, while soil moisture, groundwater and surface water anomalies reflect deficiencies in moisture storage. In contrast, evapotranspiration (ET) anomalies provide unique yet complementary information, reflecting variations in actual water use by crops – a useful diagnostic of vegetation health. Here we describe a remotely sensed Evaporative Stress Index (ESI) based on anomalies in actual-to-reference ET ratio. Actual ET is retrieved from thermal remote sensing data using a diagnostic soil-plant-atmosphere modeling system forced by measurements of morning land-surface temperature (LST) rise from geostationary satellites. In comparison with vegetation indices, LST is a relatively fast-response variable, with the potential for providing early warning of crop stress reflected in increasing canopy temperatures. Spatiotemporal patterns in ESI have been compared with patterns in the U.S. Drought Monitor and in standard precipitation-based indices, demonstrating reasonable agreement. However, because ESI does not use precipitation as an input, it provides an independent assessment of evolving drought conditions, and is more portable to data-sparse parts of the world lacking dense rain-gauge and Doppler radar networks. Integrating LST information from geostationary and polar orbiting systems through data fusion, the ESI has unique potential for sensing moisture stress at field scale, with potential benefits to yield estimation and loss compensation efforts. The ESI is routinely produced over the continental U.S. using data from the Geostationary Operational Environmental Satellites, with expansion to North and South America underway. In addition drought and ET monitoring applications are being developed over Africa and Europe using land-surface products from the Meteosat Second Generation (MSG) platform. [Copyright &y& Elsevier]

Abstract: Freshwater resources are becoming increasingly limited in many parts of the world, and decision makers are demanding new tools for monitoring water availability and rates of consumption. Remotely sensed thermal-infrared imagery collected by Landsat provides estimates of land-surface temperature that allow mapping of evapotranspiration (ET) at the spatial scales at which water is being used. This paper explores the utility of moderate-resolution thermal satellite imagery in water resource management. General modeling techniques for using land-surface temperature in mapping the surface energy balance are described, including methods developed to safeguard ET estimates from expected errors in the remote sensing inputs. Examples are provided of how remotely sensed maps of ET derived from Landsat thermal imagery are being used operationally by water managers today: in monitoring water rights, negotiating interstate compacts, estimating water-use by invasive species, and in determining allocations for agriculture, urban use, and endangered species protection. Other applications include monitoring drought and food insecurity, and evaluation of large-scale land-surface and climate models. To better address user requirements for high-resolution, time-continuous ET data, novel techniques have been developed to improve the spatial resolution of Landsat thermal-band imagery and temporal resolution between Landsat overpasses by fusing information from other wavebands and satellites. Finally, a strategy for future modification to the Landsat program is suggested, improving our ability to track changes in water use due to changing climate and growing population. The long archive of global, moderate resolution TIR imagery collected by the Landsat series is unmatched by any other satellite program, and will continue to be an invaluable asset to better management of our earth''s water resources. [Copyright &y& Elsevier]

Flash droughts are characterized by a period of rapid intensification over sub-seasonal time scales that culminates in the rapid emergence of new or worsening drought impacts. This study presents a new flash drought intensity index (FDII) that accounts for both the unusually rapid rate of drought intensification and its resultant severity. The FDII framework advances our ability to characterize flash drought because it provides a more complete measure of flash drought intensity than existing classification methods that only consider the rate of intensification. The FDII is computed using two terms measuring the maximum rate of intensification (FD_INT) and average drought severity (DRO_SEV). A climatological analysis using soil moisture data from the Noah land surface model from 1979–2017 revealed large regional and interannual variability in the spatial extent and intensity of soil moisture flash drought across the US. Overall, DRO_SEV is slightly larger over the western and central US where droughts tend to last longer and FD_INT is ~75% larger across the eastern US where soil moisture variability is greater. Comparison of the FD_INT and DRO_SEV terms showed that they are strongly correlated (r = 0.82 to 0.90) at regional scales, which indicates that the subsequent drought severity is closely related to the magnitude of the rapid intensification preceding it. Analysis of the 2012 US flash drought showed that the FDII depiction of severe drought conditions aligned more closely with regions containing poor crop conditions and large yield losses than that captured by the intensification rate component (FD_INT) alone. [ABSTRACT FROM AUTHOR]

Summary: An intercomparison among standard and remotely sensed drought indices was conducted using streamflow and soil moisture measurements collected in the Little River Experimental Watershed, Georgia, US, during the period from 2000 to 2008. All drought indices exhibited a linear, monotonic association with soil moisture, but there was a non-linear monotonic association between the drought indices and streamflow. Of the indices examined, the Evaporative Stress Index (ESI) showed reasonable performance with about 90% accuracy capturing moderate drought conditions and 80% accuracy capturing severe drought conditions in comparison to observed soil moisture and streamflow. While the ability of the ESI to capture shorter term droughts is equal or superior to the Palmer Drought Severity Index (PDSI) when characterizing droughts based on soil moisture and streamflow thresholds, the accuracy of the ESI was less efficient in the case of severe droughts. A drought index developed from the Advanced Microwave Scanning Radiometer (AMSR-E) soil moisture product showed reasonable correlations with the observed soil moisture and streamflow. However the ESI, Vegetation Health Index (VHI), and PDSI demonstrated greater skill in detecting drought in this study region. Multi-variable linear regression models revealed that the joint use of PDSI and appropriate remote sensing products improved predictions of observed hydrologic variables. Overall, the ESI was identified as a promising drought index for characterizing streamflow and soil moisture anomalies, particularly in regions where precipitation observations are unavailable, sparsely distributed, or biased with respect to regional averages. [ABSTRACT FROM AUTHOR]

Land degradation (LD) is one of the most catastrophic outcomes of long-lasting drought events and anthropogenic activities. Assessing climate and human-induced impacts on land can provide information for decision makers to mitigate the effects of these phenomena. The Northeastern region of Brazil (NEB) is the most populous dryland on the planet, making it a highly vulnerable ecosystem especially when considering the lingering drought that started in 2012. The present work consisted of detecting trends in biomass [leaf area index (LAI)] anomalies as indicators of LD in NEB. We also assessed how the loss of vegetation impacts the LD cycle, by measuring trends in albedo and evapotranspiration (ET). LAI, albedo and ET data were derived from MODIS sensors at 8-day temporal and 500 m spatial resolutions. For precipitation anomalies, we relied on CHIRPS-v2 10-day temporal at 5 km spatial resolution data. For detecting trends, we applied the Theil-Sen slope analysis on time series of MODIS LAI, albedo and ET images. Trend analysis was performed for the periods ranging from 2002–2012 (no severe droughts) to 2002–2016 (including the last drought). LAI trends were more pronounced and had a stronger signal than ET and albedo, therefore, LAI was our choice for mapping LD. The first analysis highlighted the human-induced LD prone areas whereas the last detected drought-induced LD prone areas. Considering only the trending areas, which was about 23.4% of the total, 4.5% of this area has undergone human-induced degradation whereas drought was responsible for 73%, although, not mutually exclusive. As reported in the literature and official data, grazing intensification might be a factor driving human-induced degradation. We noticed that the range of variation of LAI is narrow and even narrower for albedo, which demonstrates that land surface response is more influenced by soil reflectivity rather than the characteristic sparse vegetation coverage (LAI ranging from 0.04 to 0.4 in the Caatinga biome), which can barely alter albedo. Finally, the effects of LD on ET anomalies were assessed by Granger causality and impulse-response analyses as means to link land surface feature changes to the hydrological cycle. Albedo had a slightly weaker impulse than LAI on ET whereas precipitation played a major role. These relations are site-specific and, land surface features (biomass and albedo) showed to have a more substantial influence on ET in severely degraded areas. We concluded that drought led to trends indicating LD prone areas in NEB and the degradation cycle has positive feedback derived from ET reduction resulting in an increased net moisture deficit, although the latter statement has yet to be further investigated. The study warns of the desertification risk that NEB is facing and the need for the authorities to take action to mitigate degradation and drought effects on both traditionally surveyed (desertification nuclei) and newfound LD prone areas. We also highlight the limitation of confirming LD, as to date there is no post-drought data available and, lessons learned from the Sahel case make us cautious about claiming that an area is in fact degraded. [ABSTRACT FROM AUTHOR]