Details

Autor(en) / Beteiligte

• Gruber, Alexander
• Reichle, Rolf H.

Titel

Uncertainty Estimation for SMAP Level-1 Brightness Temperature Assimilation at Different Timescales

Ist Teil von

• IEEE journal of selected topics in applied earth observations and remote sensing, 2022-01, Vol.15, p.9127-9145

Ort / Verlag

Goddard Space Flight Center: IEEE

Erscheinungsjahr

2022

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• Soil Moisture Active Passive (SMAP) mission brightness temperature (<inline-formula><tex-math notation="LaTeX">T_{b}</tex-math></inline-formula>) observations are assimilated into NASA's Catchment Land Surface Model using an ensemble Kalman filter to update simulations of surface and root-zone soil moisture. Different time-series components of the <inline-formula><tex-math notation="LaTeX">T_{b}</tex-math></inline-formula> observations are assimilated, including anomalies, interannual variations, and high-frequency variations. To optimize the weights that the data assimilation (DA) puts on the observations, the ratio between the uncertainties of modeled and observed <inline-formula><tex-math notation="LaTeX">T_{b}</tex-math></inline-formula> is approximated using modeled and observed soil moisture uncertainties estimated using triple collocation analysis. In a benchmark experiment, <inline-formula><tex-math notation="LaTeX">T_{b}</tex-math></inline-formula> observations are assimilated using a spatially constant 4-K observation uncertainty, as in the operational SMAP Level-4 algorithm. All the DA experiments exhibit notable skill improvements in most regions. Improvements are largest for the interannual variations in the simulations of both surface and root-zone soil moisture (mean improvements in terms of Pearson correlation (-) are 0.08 and 0.06, respectively). Anomaly simulations improve similarly (0.07), and improvements in the high-frequency variations are only observed for surface soil moisture simulations (0.06). No notable difference in skill-neither improvement nor deterioration-is observed between the experiments that use optimized observation uncertainty parameters and the 4-K benchmark experiment. This may be explained by the presence of large observation operator errors, which are analytically shown to have the potential to render postupdate uncertainty insensitive to inaccuracies in estimates of the Kalman gain. These results have important implications for the design of soil moisture DA systems, in particular for parameterizing model and observation uncertainties.