We have used NASA's Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA2) reanalysis data of aerosols and meteorology into a machine learning algorithm (MLA) to estimate surface PM2.5 concentration in Thailand. One year of hourly data from 51 ground monitoring stations in Thailand was spatiotemporally collocated with MERRA2 fields. The integrated data then used to train and validate a supervised MLA' random forest' to estimate hourly and daily PM2.5 concentrations. The MLA is cross-validated using a 10-fold random sampling approach. The trained MLA can estimate PM2.5 with close to zero mean bias across the country. The correlation coefficient of 0.95 with slope and intercept values of 0.95 and 0.88 are achieved between observed and estimated PM2.5. The MLA also shows underestimation at hourly scale under very clean conditions (PM2.5 < 10 µg m-3) and overestimation during high loading (PM2.5 > 80 µg m-3). The hourly data also demonstrate high skill in following the diurnal cycle during different seasons of the year. The daily mean PM2.5 (24-hour) values follow day-to-day variability very well, showing high value during winter months (November to February) and lower during other seasons. The trained MLA has the potential to reprocess the MERRA2 timeseries for the region, and the bias corrected data can be used in other applications such as long-term trend analysis and health exposure studies. The MLA can also be applied to GEOS forecasted fields to generate bias corrected air quality forecasts for the region.

In this study we analyzed two models commonly used in remote sensing-based root-zone soil moisture (SM) estimations: one utilizing the exponential decaying function and the other derived from the principle of maximum entropy (POME). We used both models to deduce root-zone (0–100 cm) SM conditions at 11 sites located in the southeastern USA for the period 2012–2017 and evaluated the strengths and weaknesses of each approach against ground observations. The results indicate that, temporally, at shallow depths (10 cm), both models performed similarly, with correlation coefficients (r) of 0.89 (POME) and 0.88 (exponential). However, with increasing depths, the models start to deviate: at 50 cm the POME resulted in r of 0.93 while the exponential filter (EF) model had r of 0.58. Similar trends were observed for unbiased root mean square error (ubRMSE) and bias. Vertical profile analysis suggests that, overall, the POME model had nearly 30% less ubRMSE compared to the EF model, indicating that the POME model was relatively better able to distribute the moisture content through the soil column.

The objective of this study is to evaluate the ability of soil physical characteristics (i.e., texture and moisture conditions) to better understand the breeding conditions of desert locust (DL). Though soil moisture and texture are well-known and necessary environmental conditions for DL breeding, in this study, we highlight the ability of model-derived soil moisture estimates to contribute towards broader desert locust monitoring activities. We focus on the recent DL upsurge in East Africa from October 2019 though June 2020, utilizing known locust observations from the United Nations Food and Agriculture Organization (FAO). We compare this information to results from the current literature and combine the two datasets to create “optimal thresholds” of breeding conditions. When considering the most optimal conditions (all thresholds met), the soil texture combined with modeled soil moisture content predicted the estimated DL egg-laying period 62.5% of the time. Accounting for the data errors and uncertainties, a 3 × 3 pixel buffer increased this to 85.2%. By including soil moisture, the areas of optimal egg laying conditions decreased from 33% to less than 20% on average.

Biophysical crop simulation models are normally forced with precipitation data recorded with either gauges or ground-based radar. However, ground-based recording networks are not available at spatial and temporal scales needed to drive the models at many critical places on earth. An alternative would be to employ satellite-based observations of either precipitation or soil moisture. Satellite observations of precipitation are currently not considered capable of forcing the models with sufficient accuracy for crop yield predictions. However, deduction of soil moisture from space-based platforms is in a more advanced state than are precipitation estimates so that these data may be capable of forcing the models with better accuracy. In this study, a mature two-source energy balance model, the Atmosphere Land Exchange Inverse (ALEXI) model, was used to deduce root zone soil moisture for an area of North Alabama, USA. The soil moisture estimates were used in turn to force the state-of-the-art Decision Support System for Agrotechnology Transfer (DSSAT) crop simulation model. The study area consisted of a mixture of rainfed and irrigated cornfields. The results indicate that the model forced with the ALEXI moisture estimates produced yield simulations that compared favorably with observed yields and with the rainfed model. The data appear to indicate that the ALEXI model did detect the soil moisture signal from the mixed rainfed/irrigation corn fields and this signal was of sufficient strength to produce adequate simulations of recorded yields over a 10 year period.