4.3. Sampling SWC in semi- and arid regions
Sampling strategies may need to be adjusted to determine accurate areal SWC at catchment-wide scale (Grayson et al., 1998), and the interval of sampling is a critical factor which will determine the total number of samples and the cost of the experiment. According to Manfreda et al. (2006), NRS is a control factor in estimation of SWC variance, especially for soils with shallow-rooted vegetation; this is because the number of field samples must be large enough to obtain true values which are distributed close to the average for the study (A et al., 2019; Brocca et al., 2007; Brocca et al., 2012). To obtain the minimum NRS which can ensure measurement accuracy, we randomly resampled the records of planted forest with an incremental step method (bootstrap method in Fig. 6a), with the first derivative of the fitting curve in Fig. 6b ranging from ±1 to ±0.58, and the number of required samples ranging from 5 to 9 (marked with L1 and U1) and 20 to 23 (marked with L2 and U2); the results suggest that ~20 sampling points with a sampling space of 5 to 10 meters in the planted forest could yield a representative SWC value in arid regions (Entekhabi et al., 1994; Vinnikov et al., 1999). However, this experimental method may not be appropriate for a large area especially in regions with frequent rainfall (Brocca et al., 2012).