Location: Hydrology and Remote Sensing LaboratoryTitle: Field scale soil moisture retrieval using PALSAR-2 polarimetric decomposition and machine learning
|HUANG, X. - Applied Geosolutions, Llc|
|ZINITI, B. - Applied Geosolutions, Llc|
|WANG, J. - Western University|
|TORBICK, N. - Applied Geosolutions, Llc|
Submitted to: Agronomy
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/24/2020
Publication Date: 12/26/2020
Citation: Huang, X., Ziniti, B., Cosh, M.H., Reba, M.L., Wang, J., Torbick, N. 2020. Field scale soil moisture retrieval using PALSAR-2 polarimetric decomposition and machine learning . Agronomy. 11(1):35. https://doi.org/10.3390/agronomy11010035.
Interpretive Summary: Soil moisture monitoring to assess agricultural systems is of high value, but is limited in the visible remote sensing channels. Synthetic Aperture Radar offers a valuable means of monitoring through cloud layers to detect surface conditions, such as soil moisture. Using data acquired with the PALSAR-2 mission, different soil moisture algorithms were tested for their accuracy of estimating soil moisture over agricultural fields in Arkansas in 2019. The level of accuracy was encouraging for future algorithm development.
Technical Abstract: Soil moisture is a key indicator to assess cropland drought and irrigation status as well as forecast crop production. Comparing with the optical data which is obscured by the crop canopy cover, the Synthetic Aperture Radar (SAR) is an efficient tool to detect the underground surface soil moisture under the vegetation cover due to its strong penetration capability. This paper studies the soil moisture retrieval using the L-band polarimetric PALSAR-2 data acquired over the study region in the Arkansas state in United States. Both two-component modelbased decomposition (SAR data alone) and machine learning (SAR + Optical indices) methods are tested and compared in this paper. Validation using independent ground measurement shows that the both methods achieved the RMSE less than 10 [vol. %], while the machine learning methods outperforms the model-based decomposition achieving the RMSE of 7.57 [vol. %] and R2 of 0.62.