Validation of satellite obtained soil moisture with in-situ soil moisture sensor data and volumetric soil moisture samples in Bago, Myanmar.

Abstract

Soil moisture is a small component, but essential variable in the hydrological cycle. Over the past years remote sensing techniques developed substantially. VanderSat developed a downscaling technique to deliver satellite soil moisture sampled at 100 m resolution, daily and worldwide. The aim of this study is to do a quality control on this product by comparing with in-situ soil moisture sensor data and volumetric soil moisture measurements collected in the field. In June 2017, 5 locations were selected in Bago, Myanmar to monitor in-situ soil moisture content in the unsaturated zone. In-situ soil moisture is measured with Decagon GS1 ruggedized soil moisture sensors. At each location 5 sensors are present: one at surface level, two at 10cm and two at 20cm depth. The sensors are connected to an EM50 analogue data logger. Since installation of the sensors a measurement is collected every half an hour. This yields in a time series of over one year. Every location has different soil and vegetation conditions. During an 8-week fieldwork in June and July 2018, 342 volumetric soil moisture measurements were collected. Volumetric soil moisture measurements are sampled at every sensor location on the same depth with the use of Kopecky rings. After quality control on dry bulk density, 329 samples were used for validation with in-situ soil moisture sensor data at the same date and time. The best correlation between volumetric soil moisture measurements and in-situ soil moisture sensor data is found at nest 1 and the worst at nest 2. This is because at nest 2 the soil is saturated and at nest 1 the soil is very suitable to collect reliable volumetric soil moisture samples. Root mean square errors range between 0.01 and 0.29 m3/m3. From the 25 sensors, 10 exceed the threshold value of 0.05 m3/m3 used for validation of SMAP satellites (Colliander et al., 2017). Time series of over one year in-situ soil moisture sensor data from every nest are validated with VanderSat soil moisture X and L-band. The best correlations for the X and L-band are found at nest 5, probably because this nest has the highest representation of the surrounding environment. Root mean square errors range between 0.07 and 0.16 m3/m3 for X-band, and 0.06 to 0.23 m3/m3 for L-band. The X and L-band both exceed the threshold value of 0.05 m3/m3 used for validation of SMAP satellites (Colliander et al., 2017). From the comparison of volumetric soil moisture measurements with in-situ soil moisture sensor data and the VanderSat soil moisture product during the monsoon period can be concluded that the X-band displays the best correlation and the L-band overestimates soil moisture content.