Investigating Irrigation Impacts on Spatiotemporal Variability of Soil Moisture for Calibration of a Multifractal Downscaling Algorithm in Agricultural Areas

Heterogeneity in soil moisture (θ) fields remains a challenge for the calibration of statistically-based downscaling algorithms of coarse satellite estimates. In this study, a multifractal downscaling technique is applied using airborne θ observation at 1-km resolution from two campaigns in the National Airborne Field Experiments in 2005 and 2006 (NAFE05 and NAFE06). In these regions in Australia, the presence of irrigated areas introduces spatial heterogeneity in θ distribution, thus potentially affecting the calibration of the disaggregation algorithm. To investigate this effect, we first identify the potential irrigation pixels, using a two-steps procedure based on land cover information and available θ images. Next, we introduce three scenarios to quantify how irrigated pixels affect the scale invariance and multifractal properties of θ. In the first scenario (SC1), the analysis is performed on the original θ data; in the second (SC2) the irrigated pixels are removed from the images; in the third scenario (SC3), irrigated pixels are removed and interpolated based on θ values in neighboring pixels. By quantifying the linearity of the relations involved in the scale invariance analysis, we found that, in NAFE06, the presence of this properties is more evident on θ fields of SC3, indicating that irrigated pixels introduce spatial heterogeneity that affect the multifractal model calibration. In NAFE05, we did not find appreciable differences among the scenarios, due to the small size of the study area, low number of days, and reduced number and sparseness of the irrigated pixels. After conducting the scale invariance and multifractal analysis, we calibrate the downscaling model by identifying empirical relations between its parameters and coarse predictors. Model performances are finally evaluated by comparing synthetic disaggregated fields with aircraft and ground θ fields. This study is useful to advance the knowledge of statistical properties of θ spatial distribution, and to improve the use of satellite soil moisture products in agricultural regions.