Enhancing Spatial Resolution of SMAP Soil Moisture Products through Spatial Downscaling over a Large Watershed: A Case Study for the Susquehanna River Basin in the Northeastern United States

Abstract

Soil moisture (SM) with a high spatial resolution plays a paramount role in many local and regional hydrological and agricultural applications. The advent of L-band passive microwave satellites allowed for it to be possible to measure near-surface SM at a global scale compared to in situ measurements. However, their use is often limited because of their coarse spatial resolution. Aiming to address this limitation, random forest (RF) models are adopted to downscale the SMAP level-3 (L3SMP, 36 km) and SMAP enhanced (L3SMP_E, 9 km) SM to 1 km. A suite of predictors derived from the Sentinel-1 C-band SAR and MODIS is used in the downscaling process. The RF models are separately trained and verified at both spatial scales (i.e., 36 and 9 km) considering two experiments: 1) using predictors derived from the MODIS and Sentinel-1 along with other predictors such as elevation and brightness temperature and 2) using all predictors of the first experiment except for the Sentinel-1 predictors. Only dates when the Sentinel-1 images were available are considered for the comparison of the two experiments. The comparison of the results of the two experiments indicates that the removal of Sentinel-1 predictors from the second experiment only reduces the R value from 0.84 to 0.83 and from 0.91 to 0.86 for 36 and 9 km spatial scales, respectively. Among the predictors used in the downscaling, the brightness temperature in VV polarization is identified as the most important predictor, followed by NDVI, surface albedo and API. On the contrary, the Sentinel-1 predictors play a less important role with no marked contribution in enhancing the predictive accuracy of RF models. In general, the two experiments have limitation, such as a small sample size for the training of the RF model because of the scarcity of Sentinel-1 images (i.e., revisit time of 12 days). Therefore, based on this limitation, a third experiment is proposed, in which the Sentinel-1 predictors are not considered at all in the training of the RF models. The results of the third experiment show a good agreement between the downscaled L3SMP and L3SMP_E SM, and in situ SM measurements at both spatial scales. In addition, the temporal availability of the downscaled SM increased. Moreover, the downscaled SM from both SMAP products presented greater spatial detail while preserving the spatial patterns found in their original products. The use of the two SMAP SM products as background fields for the downscaling process does not show marked differences. Overall, this study demonstrates encouraging results in the downscaling of SMAP SM products over humid climate with warm summers dominated by vegetation.

Watershed scale soil moisture estimation model using machine learning and remote sensing in a data-scarce context

“…). This behavior is unexpected; typically, models are expected to perform better during dry periods, as demonstrated in numerous previous studies (Wakigari & Leconte, 2022). One possible explanation, reinforced by subsequent results, is that soil moisture distribution becomes more complex during dry periods as well as the inclusion of DEM derivatives in our work that have been demonstrated to be strong predictive factors of soil moisture at basin scale, especially in wet season (Raduła et al, 2018).…”

“…Unlike previous studies such as Abbaszadeh et al (2019) and Wakigari & Leconte (2022) that parameterized random forests using all available data for an entire study period, this study used approximately 1300 SMAP pixels within the study area. While this approach captures dynamic relationships between soil moisture and covariates, this strategy reduces the sample size used for training the daily models; posing the risk that the models may not have enough observations to effectively capture the relationships between covariates and soil moisture of the SMAP product.…”

Spatial Downscaling and Gap-Filling of SMAP Soil Moisture to High Resolution Using MODIS Surface Variables and Machine Learning Approaches over ShanDian River Basin, China

“…This might be because of the randomization and robustness of the RF algorithm that help the model avoid overfitting when even thousands of variables are given simultaneously. Moreover, the spatial aggregation of high-resolution predictors, such as VIs, WIs, Albedo, and LST, had a smoothing impact on the extreme values, resulting in the training of RF and ANN models with minimal extremes, as previously indicated by Wakigari et al [77]. However, this was not special to our work since existing downscaling approaches depend on calibration at a coarse spatial resolution as the initial step, making the aggregation of high spatial-resolution predictors inevitable.…”

“…RF is easy to work with and adaptable, and it is less sensitive to changes in hyperparameters than other models. It has been shown in previous studies that the RF model is capable of representing a workable SM downscaling model, and it has also been proved to be successful in complicated nonlinear fitting [34,49,50,53,76,77].…”

Karakteristik Estuari di Muara Angke pada Musim Timur

“…Berdasarkan profil kontur salinitas vertikal, bahwa salinitas di lapisan permukaan bertambah relatif kecil ke lapisan tengah dan begitu juga salinitas di lapisan dasar berkurang relatif kecil ke lapisan tengah sedangkan di lapisan tengah sendiri nilai salinitasnya berada diantara nilai salinitas di permukaan dan di dasar yang menandakan di lapisan tengah ini terjadi percampuran massa air tawar dan air asin. Sama seperti yang dilakukan Wakigari & Leconte (2022) di Sungai Susquehanna di Amerika Serikat bahwa estuari tercampur sebagian memiliki karakteristik utama yaitu nilai salinitasnya akan bervariasi di sepanjang muara maupun terhadap kedalaman dan nilai salinitas dari lapisan atas ke lapisan bawah memiliki nilai yang hampir seragam.…”