Uncertainty of the Reference Pixel Average Soil Moisture Observed at SMAP Core Validation
Abstract
Despite efforts to maximize the sampling coverage and improve the upscaling functions for core validation sites (CVS) established by the NASA Soil Moisture Active/Passive (SMAP) mission, spatial averages of point-scale, in situ soil moisture observations cannot perfectly represent the true spatial soil moisture average within SMAP reference pixels. As a result, some residual level of underlying uncertainty must be tolerated in validation of satellite-based soil moisture retrievals using ground data. Using data collected for 33-km SMAP_L2SMP_E reference pixels from permanent and temporary networks at 15 SMAP CVS, this study examines uncertainties in the core site average soil moisture (CSASM)- which serves as the true footprint-scale average soil moisture for SMAP calibration/validation efforts - due to limitations in spatial and temporal sampling and discusses their implications to the SMAP validation metrics. A conservative estimate of the spatial uncertainty of CSASM is found to be large relative to CSASM itself, which translates into large uncertainty bounds on SMAP bias estimates obtained as differences against CSASM. At most CVS, SMAP bias is found to fall within 95% confidence interval of zero-bias, indicating these biases are statistically insignificant (i.e. easily attributed to spatial sampling limitations within each CVS). In addition, observations from temporary networks at four CVS suggest the spatial variability is under-sampled at many CVS with the existing permanent network and the spatial uncertainties (and thus bias uncertainty bounds) are likely underestimated. The persistent presence of large bias uncertainty bars is linked to the temporal stability of spatial soil moisture patterns. Likewise, the unbiased-RMSE (ubRMSE) of CSASM (versus unknown true mean soil moisture over the SMAP reference pixel) is estimated via two distinct error analysis methods: classical sampling theory and triple collocation. Despite their differences, the CSASM ubRMSE estimated from both methods are relatively similar, with an average of 16% higher via TC. Both methods suggest that the ubRMSE of CSASM at most SMAP CVS falls within 0.01 and 0.02 m3/m3.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.H51W1640C
- Keywords:
-
- 1833 Hydroclimatology;
- HYDROLOGYDE: 1843 Land/atmosphere interactions;
- HYDROLOGYDE: 1855 Remote sensing;
- HYDROLOGYDE: 1866 Soil moisture;
- HYDROLOGY