Applicability of cosmic-ray neutron sensing for measuring soil water content to heterogeneous landscapes under subtropical hydroclimatic conditions

Abstract

Cosmic-ray neutron sensing (CRNS) is one of the few techniques that can directly acquire information about the temporal dynamics of soil water content (SWC) at a hectometer scale. The performance of CRNS depends on the climate, soil and conditions of the underlying surface. The region of red soil in China is characterized by heterogeneous land uses (combination of upland and paddy fields), distinct dry-wet periods and low contents of soil organic carbon. The applicability of CRNS has not yet been evaluated under the subtropical hydroclimatic conditions of this region. We therefore compared spatially averaged SWC measured using CRNS with point-scale data measured using time domain reflectometry (TDR) at five locations and oven-dried soil samples collected at 27 locations across the CRNS footprint in an agricultural ecosystem of the red-soil region between March 2019 and February 2020. The footprint of CRNS varied with time. The horizontal radius of CRNS varied between 175 m in July and 215 m in October, with a mean of 194 m, and the measuring depth varied between 16 cm in March and 23 cm in August, with a mean of 19 cm. The relative contribution of rainfed cropland, citrus land and paddy fields were 86.7, 10.7 and 2.6%, respectively. The differences of the relative contribution among land uses were determined by the distance to the sensor and the area, which were less affected by the types of land use. The neutron count rate (N0, counts/h) was weakly temporally variable, with a mean coefficient of variation of 3%. The influence of its variation on the accuracy of CRNS, however, should not be ignored. Both the timing and frequency of sampling affected the calculation of N0. Occasional sampling in the dry season is a recommended strategy to acquire an accurate N0 of a region. CRNS performed well for determining the dynamics of spatially averaged SWC in heterogeneous landscapes. More specifically, the accuracy of CRNS was comparable with those using the oven-drying method, with R2, root mean square error and Nash-Sutcliffe model efficiency coefficient of 0.86, 0.026 cm3/cm3 and 0.81, respectively. This study evaluated the applicability of CRNS in a region with complex climatic, hydrological and underlying-surface conditions. These results are useful for developing a reasonable sampling strategy for acquiring an accurate N0 and calibrating the CRNS. Future studies, however, should focus on the application of the CRNS technique to resolve the eco-environmental issues of the region.