On the application of RADARSAT-2 IEM retrievals for determining inter-field soil moisture distributions a priori to vegetation emergence

English Abstract:

Near-daily soil moisture products are available from satellite remote sensing, however, data are generally limited to relatively coarse spatial footprints (i.e. 10s of km). Knowledge of the sub-footprint spatial distribution can assist in downscaling coarse soil moisture measurements to a level more suitable to supporting agricultural producers. Field-scale soil moisture retrievals are available from C-band synthetic aperture radar (SAR) backscatter modelling, however, these are limited to pre-vegetation/post-harvest periods due to a saturation of C-band microwaves over vegetation canopies. The objective of this study is to assess how these field-scale SAR retrievals may be used to generate knowledge of a sub-footprint soil moisture scaling structure a priori to vegetation emergence. Inter-field soil moisture dynamics are assessed for 33 study fields within a 15 km Soil Moisture Ocean Salinity L.2 grid cell over April 25 to July 19, 2012. Fields were cropped for corn, canola, soybeans and spring wheat. RADARSAT-2 data are acquired over the study fields on 7 dates between April 25 and May 27. Soil moisture is mapped in each field using the multipolarization IEM approach (available in RADARSAT-2 toolkit v.9.5.1) and the field-means are averaged to generate an up-scaled soil moisture estimate. Following vegetation growth the same fields were sampled in-situ on 15 dates between June 06-July 19 as part of SMAPVEX12. Results show reasonable correspondence between the up-scaled IEM soil moisture and SMOS L.2 soil moisture, with a r-value of 0.72, RMSE of 0.08 m³m^-3 and mean difference of 0.06 m³m^-3 (i.e. SMOS dry bias). These results are consistent with a 2-year SMOS L.2 validation over this domain using in-situ network data. However, inter-field distributions from RADARSAT-2 IEM modelling over the pre-vegetation period show some significant differences between dates and also to patterns observed from in-situ sampling during the vegetation growth cycle. It is hypothesized that the frequent land management activities over the pre-vegetation period influence the inter-field soil moisture variability and acts to limit temporally stable patterns. Similarly, unique soil moisture variability controls may be present during crop growth cycles highlighting the need for field scale crop mapping.