Analysis of field-sampled, in-situ network, and PALS airborne soil moisture observations over SMAPVEX12

Abstract:

The Soil Moisture Active Passive Validation Experiment in 2012 (SMAPVEX12) was conducted over an agricultural domain in southern Manitoba, Canada. The purpose of the campaign was to develop ground and airborne datasets for pre-launch validation of SMAP satellite soil moisture retrieval algorithms. Three key soil moisture datasets were collected in support of the campaign objectives: 1) intensive field sampling over (up to) 55 agricultural fields on 17 sampling days; 2) a continuously operated temporary in-situ network (> 30 stations) distributed over the domain; and 3) L-band microwave data from NASA’s Passive Active L-band Sensor (PALS) onboard a Twin-Otter aircraft. This presentation addresses whether dense temporary in-situ networks can supplant intensive field-sampling during pre-/post-launch validation campaigns. SMAPVEX12 datasets are examined at the field and aircraft pixel (~800 m) scale, and at the domain scale. Preliminary results demonstrate that, at the field-scale, there is generally limited agreement between a single station and sampled data over its field. Over the duration of the campaign, the majority of temporary soil moisture stations have > 0.04 m³m^-3 RMSE with sampled field data, suggesting that a single station has limited representativeness of an agricultural field. Furthermore, the in-situ stations and field-sampled data are compared with PALS generated soil moisture to assess differences in daily RMSE. For wet-periods, both ground datasets provide a comparable RMSE for the PALS estimate. Although for dry-periods, the difference in RMSE between the ground datasets becomes more significant (> 0.04 m³m^-3). This is because the field-sampled data exhibit a sharper dry-down than the in-situ station measurements. However, at the domain scale there is strong agreement between the soil moisture datasets. Additional results describe the sources of variability affecting these soil moisture datasets and the statistical number of stations needed to represent the SMAPVEX12 domain. This research is of importance for the efficient allocation of ground resources during remote sensing validation campaigns for soil moisture.