Maqu network for validation of satellite-derived soil moisture products

Abstract

Soil moisture monitoring of the Tibetan Plateau is of primary importance for understanding land–atmosphere interactions of this region and their effects on the climate of eastern and South-East Asia. Operational satellite-derived soil moisture products, such as those obtained from AMSR-E data by VUA–NASA and ASCAT data by TU-Wien, as well as that to become available in the near future (such as data from SMOS and SMAP), can provide the information required, but their accuracy for this region needs to be evaluated before further application. For this reason, a soil moisture and temperature monitoring network was set up in the water source region of the Yellow River, in the north-eastern region of the Tibetan Plateau (Maqu county). It consists of 20 stations distributed, according to a stratified sampling, over an area of approximately 40km×80km.This study describes the Maqu network and presents the first set of data measured from July 2008 to December 2009, which shows the capability of the network to monitor the spatial and temporal soil moisture variability of the area with a high degree of accuracy. Temporal stability analyses revealed that the soil moisture spatial patterns are not always stable in time. The sites that show the highest and the most variable bias with respect to the average are located in regions with extreme soil properties, covering relatively small areas.The weighted spatial average of measured soil moisture was successfully used as ground reference for the validation of the AMSR-E soil moisture products and ASCAT soil wetness index products. For the monsoon season, overall good agreement was found between in situ time series and AMSR-E products, with a linear fit between the two datasets close to the 1:1 line and a standard error of the regression lower than 0.05. The agreement between ASCAT and in situ data was affected by several large variations of the former corresponding to little changes of the latter, thus the standard error of the regression was higher than 0.07.