Abstract
Soil moisture plays an important role in the global water cycle and has an important impact on energy fluxes at the land surface. It also defines the initial and boundary condition of terrestrial hydrological processes, including infiltration, runoff, and evapotranspiration. Therefore, accurate estimation of soil moisture pattern is of critical importance. Satellite-based soil moisture can be obtained with well-defined temporal and spatial resolutions and with global coverage. However, they only provide surface soil moisture at the upper few centimeters of the soil column. Soil moisture simulation models can produce estimates of soil moisture profile up to several meters of depth in different time steps. However, uncertainty in model parameters (e.g., unknown initial soil moisture profile) and meteorological forcing can substantially alter the accuracy of the model estimates. In this article, the potential of using surface soil moisture measurements to retrieve the initial soil moisture profile will be explored in a synthetic study, using two proposed reduced-order variational data assimilation (VDA) techniques and a simple 1-D soil moisture model. The accuracy and feasibility of the proposed approaches are confirmed by comparing the initial soil moisture profiles estimated using the proposed reduced-order VDA techniques versus the full-adjoint VDA technique. Results illustrated that the reduced-order VDA techniques can estimate initial soil moisture profile from near surface soil moisture observations with the comparable level of accuracy as full-adjoint VDA. The effectiveness of the reduced-order VDA in retrieving the initial soil moisture profile is further demonstrated by assimilating surface soil moisture into HYDRUS-1D, mimicking real-world errors.
Highlights
S PATIALLY distributed soil moisture profiles are required for land surface and land-atmosphere interaction studies, improving agricultural productivity, assessing drought and flood conditions, estimating groundwater supplies, and landslide prediction
The results presented in this experiment demonstrate the efficiency and effectiveness of the proposed reduced-order variational data assimilation (VDA) methods (RA-VDA and ModelReduced VDA (MR-VDA)) in estimating initial soil moisture profile and improving the accuracy of time series of soil moisture profile by assimilating surface soil moisture observation into the highly nonlinear HYDRUS 1D model
In the ModelReduced VDA (MR-VDA), the adjoint model is approximated based on the reduced-order model, and the entire optimization is performed in a reduced space
Summary
S PATIALLY distributed soil moisture profiles are required for land surface and land-atmosphere interaction studies, improving agricultural productivity, assessing drought and flood conditions, estimating groundwater supplies, and landslide prediction. Measurement and estimation of soil moisture profile and simulation of its pattern play a key role in hydrological research. Non-point soil moisture measurement techniques exist from local to global scale by using geophysical methods like ground penetrating radar [4], cosmic ray probes [5], ground-based radiometry [6] electromagnetic methods [7], airborne SAR polarimetry [8], airborne L-band radiometry [9], and spaceborne sensors like AQUA/ AMSR-E [10], ERS-Scat [11], ASCAT [12], WindSat [13] , ENVISAT/ ASAR [14] , ALOS/PALSAR [15] , SMOS/MIRAS [16], and SMAP [17], [18]. Soil Moisture simulation models, generally called soil vegetation atmosphere transfer (SVAT) models, can produce estimates of soil moisture profile to several meters of depth on different (e.g. hourly, daily, monthly) time steps. Initial soil moisture profile plays a key role in infiltration, and the time evolution of the profile of soil moisture
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
