Global scale error assessments of soil moisture estimates from microwave-based active and passive satellites and land surface models over forest and mixed irrigated/dryland agriculture regions

Over the past four decades, satellite systems and land surface models have been used to estimate global-scale surface soil moisture (SSM). However, in areas such as densely vegetated and irrigated regions, obtaining accurate SSM remains challenging. Before using satellite and model-based SSM estimates over these areas, we should understand the accuracy and error characteristics of various SSM products. Thus, this study aimed to compare the error characteristics of global-scale SSM over vegetated and irrigated areas as obtained from active and passive satellites and model-based data: Advanced Scatterometer (ASCAT), Soil Moisture and Ocean Salinity (SMOS), Advanced Microwave Scanning Radiometer 2 (AMSR2), Soil Moisture Active Passive (SMAP), European Centre for Medium-Range Weather Forecasts Reanalysis 5 (ERA5), and Global Land Data Assimilation System (GLDAS). We employed triple collocation analysis (TCA) and caluclated conventional error metrics from in-situ SSM measurements. We also considered all possible triplets from 6 different products and showed the viability of considering the standard deviation of TCA-based numbers in producing robust results. Over forested areas, it was expected that model-based SSM data might provide more accurate SSM estimates than satellites due to the intrinsic limitations of microwave-based systems. Alternately, over irrigated regions, observation-based SSM data were expected to be more accurate than model-based products because land surface models (LSMs) cannot capture irrigation signals caused by human activities. Contrary to these expectations, satellite-based SSM estimates from ASCAT, SMAP, and SMOS showed fewer errors than ERA5 and GLDAS SSM products over vegetated conditions. Furthermore, over irrigated areas, ASCAT, SMOS, and SMAP outperformed other SSM products; however, model-based data from ERA5 and GLDAS outperformed AMSR2. Our results emphasize that, over irrgated areas, considering satellite-based SSM data as alternatives to model-based SSM data sometimes produces misleading results; and considering model-based data as alternatives to satellite-based SSM data in forested areas can also sometimes be misleading. In addition, we discovered that no products showed much degradation in TCA-based errors under different vegetated conditions, while different irrigation conditions impacted both satellite and model-based SSM data sets. The present research demonstrates that limitations in satellite and modeled SSM data can be overcome in many areas through the synergistic use of satellite and model-based SSM products, excluding areas where satellite-based data are masked out. In fact, when four satellite and model data sets are used selectively, the probability of obtaining SSM with stronger signal than noise can be close to 100%.