Improving Soil Moisture Estimation by Identification of NDVI Thresholds Optimization: An Application to the Chinese Loess Plateau

Abstract

Accuracy soil moisture estimation at a relevant spatiotemporal scale is scarce but beneficial for understanding ecohydrological processes and improving weather forecasting and climate models, particularly in arid and semi-arid regions like the Chinese Loess Plateau (CLP). This study proposed Criterion 2, a new method to improve relative soil moisture (RSM) estimation by identification of normalized difference vegetation index (NDVI) thresholds optimization based on our previously proposed iteration procedure of Criterion 1. Apparent thermal inertia (ATI) and temperature vegetation dryness index (TVDI) were applied to subregional RSM retrieval for the CLP throughout 2017. Three optimal NDVI thresholds (NDVI0 was used for computing TVDI, and both NDVIATI and NDVITVDI for dividing the entire CLP) were firstly identified with the best validation results (R¯) of subregions for 8-day periods. Then, we compared the selected optimal NDVI thresholds and estimated RSM with each criterion. Results show that NDVI thresholds were optimized to robust RSM estimation with Criterion 2, which characterized RSM variability better. The estimated RSM with Criterion 2 showed increased accuracy (maximum R¯ of 0.82 ± 0.007 for Criterion 2 and of 0.75 ± 0.008 for Criterion 1) and spatiotemporal coverage (45 and 38 periods (8-day) of RSM maps and the total RSM area of 939.52 × 104 km2 and 667.44 × 104 km2 with Criterion 2 and Criterion 1, respectively) than with Criterion 1. Moreover, the additional NDVI thresholds we applied was another strategy to acquire wider coverage of RSM estimation. The improved RSM estimation with Criterion 2 could provide a basis for forecasting drought and precision irrigation management.

Modeling Soil Moisture from Multisource Data by Stepwise Multilinear Regression: An Application to the Chinese Loess Plateau

“…The relative soil moisture (RSM) represents the percentage of SM that accounts for the moisture storage capacity and was used to describe the SM levels in the present study. The monthly, seasonal, and annual RSM maps of the CLP in 2017 (previously published [32]) were used and regarded as the reference RSM for modeling in this study. The previously published RSM was generated via 8-day RSM maps.…”

“…The previously published RSM was generated via 8-day RSM maps. The overall 8-day RSM was combined at a 500 m resolution by corresponding subregional RSM, which was produced with three groups of selected optimal NDVI thresholds using MODIS-derived ATI (apparent thermal inertia) and TVDI (temperature vegetation dryness index), and the average of ATI and TVDI against 20 cm depth in situ RSM observations [32]. Here, many studies pointed out TVDI and ATI could adequately reflect the changes of RSM at a 20 cm depth [11,[80][81][82].…”

“…To assess the performance of the models over the CLP, the Pearson's r and the mean absolute error (MAE) of the reference RSM against in situ RSM observations were calculated (at the monthly, seasonal, and annual scales). The r and MAE varied from 0.47 in October to 0.68 in January and from 3.02% in March to 4.97% in October, respectively, on the monthly scale (Table A1) [32]. Since the r of reference RSM and in situ RSM observations had a moderate correlation coefficient (r = 0.73 for the annual scale) [85], the reference RSM was assumed to be the actual RSM with larger spatial coverage than those in situ RSM observations and the reference RSM kept a better trend with the in situ observed RSM at the station scale.…”

“…Although Yuan et al improved SM estimation concerning the spatiotemporal coverage for the CLP using MODIS-derived apparent thermal inertia (ATI) and Temperature Vegetation Dryness Index (TVDI), certain pixels did not have SM values. This led to the display of incomplete SM maps at the monthly scale [32].…”

“…However, from the perspective of statistical analysis, the problem is more challenging if the sample size is not sufficient to apply the SMLR model [72,73]. Thus, the sparse in situ SM observations (e.g., only 49 available annual SM observations) are not suitable to be regarded as the reference SM for the entire CLP (640,000 km 2 ) [32]. The incomplete reference SM maps from our previously published, to some extent, were still not available for certain applications.…”