Accurately constraining surface soil moisture/evapotranspiration coupling using ground observations and satellite remote sensing

Abstract:

The one-way coupling between soil water availability and surface energy fluxes between the land surface and the boundary layer represents a critical dynamic component of the integrated bedrock-to-boundary layer system. Recent cross-comparisons between long-term ground-based observations of surface soil moisture (SM) and evapotranspiration (ET) have suggested that land surface models generally overestimate the strength of this one-way coupling. However, these assessments neglect the impact of random error in ground-based SM and ET measurements. Such independent error will tend to spuriously degrade their sampled cross-correlation. Therefore, the apparent over-coupling of LSMs may be due to independent errors in SM and ET observations and not reflective of a systematic shortcoming in LSMs.

In order to examine this issue in greater detail, we propose a new triple collocation-based (TC) technique to estimate the true Spearman rank correlation coefficient between daily surface soil moisture and ET. Specifically, we apply a TC-based technique to concurrent SM and ET datasets obtained from: 1) an LSM, 2) ground-based observations, and 3) remotely sensing. The approach yields a Spearman rank correlation estimate that is unbiased in the presence of (potentially-significant) random error in any member of the SM and ET triplet. As such, the approach provides a highly-relevant example of exploiting long-term remote sensing and in situ datasets to improve process-level understanding within the bedrock-to-boundary layer system.

Application of the technique to North American Land Data Assimilation System-2 (NLDAS-2) models over Ameriflux sites between 2002 and 2014 (utilizing satellite-based ALEXI ET and AMSR-E/LPRM surface soil moisture products) suggests that the apparent over-coupling of LSM soil moisture/ET predictions at Ameriflux sites can largely be attributed to independent errors in ground-based soil moisture and ET observations at each site. Generally, NLDAS-2 models tend to either accurately represent (or even slightly under-predict) the true one-way SM/ET coupling strength within arid and semi-arid regions. The consequences of this result for the coupled modeling of the bedrock-to-boundary layer system will be discussed.