A robust gap-filling approach for European Space Agency Climate Change Initiative (ESA CCI) soil moisture integrating satellite observations, model-driven knowledge, and spatiotemporal machine learning

[摘要] Spatiotemporally continuous soil moisture (SM) data areincreasingly in demand for ecological and hydrological research. Satelliteremote sensing has potential for mapping SM, but the continuity ofsatellite-derived SM is hampered by data gaps resulting from inadequate satellite coverage, snow cover, frozen soil, radio-frequency interference, and so on. Therefore, we propose a new gap-filling approach to reconstructdaily SM time series using the European Space Agency Climate Change Initiative (ESA CCI). The developed approach integrates satellite observations,model-driven knowledge, and a machine learning algorithm that leverages bothspatial and temporal domains. Taking SM in China as an example, thereconstructed SM showed high accuracy when validated against multiple setsof in situ measurements, with a root mean square error (RMSE) and a mean absolute error (MAE) of 0.09–0.14 and0.07–0.13 cm 3  cm −3 ,respectively. Further evaluation with a 10-fold cross-validation revealed median values of the coefficient of determination (R 2 ), RMSE, and MAEof 0.56, 0.025, and 0.019 cm 3  cm −3 , respectively.The reconstructive performance was noticeably reduced both when excludingone explanatory variable and keeping the other variables unchanged and when removing the spatiotemporal domain strategy or the residual calibrationprocedure. In comparison with gap-filled SM data based on asatellite-derived diurnal temperature range (DTR), the gap-filled SM datafrom bias-corrected model-derived DTRs exhibited relatively lower accuracybut higher spatial coverage. Application of our gap-filling approach tolong-term SM datasets (2005–2015) produced a promising result ( R 2 =0.72 ). A more accurate trend was achieved relative to that of the originalCCI SM when assessed with in situ measurements (i.e., 0.49 versus 0.28,respectively, in terms of R 2 ). Our findings indicate the feasibility ofintegrating satellite observations, model-driven knowledge, andspatiotemporal machine learning to fill gaps in short- and long-term SM timeseries, thereby providing a potential avenue for applications to similarstudies.

[学科分类] 妇产科学