Reappraisal of SMAP inversion algorithms for soil moisture and vegetation optical depth

Lun Gao, Ardeshir Ebtehaj, Mario Julian Chaubell, Morteza Sadeghi, Xiaojun Li, Jean Pierre Wigneron

Research output: Contribution to journalArticlepeer-review

30 Scopus citations

Abstract

NASA's Soil Moisture Active Passive (SMAP) satellite mission has been providing high-quality global estimates of soil moisture (SM) and vegetation optical depth (VOD) using L-band radiometry since 2015. To date, a variety of retrieval algorithms as well as surface roughness and scattering albedo have been developed. However, a comprehensive evaluation of different algorithms with the new surface parameters across diverse biomes, climates, and terrain slopes is lacking. To narrow down this knowledge gap, here we examine the performance of various existing algorithms, including V-pol Single Channel Algorithms (SCA-V), H-pol Single Channel Algorithms (SCA[sbnd]H), classic DCA, extended DCA (E-DCA), regularized DCA (RDCA), land parameter retrieval model (LPRM), multi-temporal DCA (MT-DCA), constrained multi-channel algorithm (CMCA), and spatially constrained multi-channel algorithm (S-CMCA). The SM estimates are evaluated against in-situ measurements from the International Soil Moisture Network (ISMN) while VOD estimates are compared with the two-band enhanced vegetation index (EVI2), tree height, and aboveground biomass. This study has led to several important findings: (1) The overall bias, root mean square error (RMSE), and unbiased RMSE (ubRMSE) of SM estimates from different algorithms generally increase with vegetation density while their temporal correlations with in-situ measurements decrease as the terrain slope increases. (2) The divergence between different SM estimates is relatively larger over forested areas than non-forested areas. (3) In terms of temporal correlation with in-situ measurements, the SCA-V and RDCA outperform other algorithms over most land cover types and climates. (4) SCA-H typically underestimates SM more compared to other algorithms across sparsely vegetated areas and most climates. (5) The ubRMSE values demonstrate that all algorithms have close performance when EVI2 is less than 0.3; however, the performance of classic DCA decays notably when EVI2 exceeds 0.3. (6) VOD retrievals from RDCA exhibit improved spatial correlations with EVI2, tree height, and aboveground biomass across the globe compared to other algorithms. Overall, RDCA exhibits a good compromise between the high performance of SM and VOD.

Original languageEnglish (US)
Article number112627
JournalRemote Sensing of Environment
Volume264
DOIs
StatePublished - Oct 1 2021

Bibliographical note

Funding Information:
This research is conducted under the support of the Future Investigators in NASA Earth and Space Science (FINESST, 80NSSC19K1333) and the NASA's Terrestrial Hydrology Program (THP, 80NSSC18K1528). The first author also appreciates early discussion with Dr. Dara Entekhabi at Massachusetts Institute of Technology, Dr. Steven K. Chan at the NASA Jet Propulsion Laboratory, and Dr. Alexandra Konings at Stanford University.

Funding Information:
This research is conducted under the support of the F uture Investigators in NASA Earth and Space Science (FINESST, 80NSSC19K1333) and the NASA's Terrestrial Hydrology Program (THP, 80NSSC18K1528). The first author also appreciates early discussion with Dr. Dara Entekhabi at Massachusetts Institute of Technology, Dr. Steven K. Chan at the NASA Jet Propulsion Laboratory, and Dr. Alexandra Konings at Stanford University.

Publisher Copyright:
© 2021 Elsevier Inc.

Keywords

  • L-band radiometry
  • SMAP
  • Soil moisture
  • Vegetation optical depth

Fingerprint

Dive into the research topics of 'Reappraisal of SMAP inversion algorithms for soil moisture and vegetation optical depth'. Together they form a unique fingerprint.

Cite this