Detecting soil moisture drydowns in remote sensed products and exploring their potential for land surface model parameter estimation studies

NINA RAOULT; SÖRENSSON ANNA AMELIA; SALVIA, M. MERCEDES; RUSCICA ROMINA

Nueva Orleans

Encuentro; AGU - Remote Sensing, Modeling, and Data Assimilation of the Terrestrial Water Cycle; 2021

Soil moisture (SM) drydowns occur during periods without water inputs. These temporal features of the SM time series can be fit using a time scale τ, which gives information on the integrated soil response to climate, land cover, and soil hydraulic conditions during a dry period. As such, it is important that drydowns are well represented in our land surface models (LSMs). First, we tested using τ as a data assimilation metric to constrain parameters in LSMs over a handful of in situ sites. Next, the large number of satellite products available providing surface soil moisture estimates meant that we could start extending our study of these features globally. However, we found that in products based on individual sensors, the detection of fast drydowns was hindered by the product´s sampling frequency. To overcome this issue, we started considering merged satellite products like the European Space Agency Climate Change Initiative SM combined dataset (C_CCI). When merging different retrievals to create the C_CCI, data were rescaled using a land surface model, preserving the original dynamics of the retrievals but imposing the absolute values and dynamic range of the model on the product. To see whether this rescaling affects our ability to detect SM drydowns in C_CCI, we tested two detection methods: i) based on SM dynamics (Mdyn), ii) using precipitation data (Mprec). Since Mprec requires an additional product compatible with C_CCI, where rainfall events coincide with SM dynamics, Mdyn is preferable. However, we find that the smaller SM range imposed by the model on C_CCI hinders drydown detection. Only detection is affected by the rescaling, not the drydown time scales themselves.