作者:Peng, B (Peng, Bin)[ 1,2,3 ] ; Zhao, TJ (Zhao, Tianjie)[ 1 ] ; Shi, JC (Shi, Jiancheng)[ 1 ] ; Lu, H (Lu, Hui)[ 4,5 ] ; Mialon, A (Mialon, Arnaud)[ 6 ] ; Kerr, YH (Kerr, Yann H.)[ 6 ] ; Liang, X (Liang, Xu)[ 7 ] ; Guan, KY (Guan, Kaiyu)[ 2,3 ]

REMOTE SENSING OF ENVIRONMENT

卷: 199  页: 63-77

DOI: 10.1016/j.rse.2017.07.006

出版年: SEP 15 2017

摘要

Roughness effect parameterization is critical to accurately simulate brightness temperature (Tb) signals observed by a radiometer over bare soil surface. However, current roughness parameterization schemes usually suffer from severe error, which dominates the error budget in current Tb modeling over bare soil surface. In this study, uncertainty of soil roughness parameterization schemes is comprehensively assessed using data set collected during 2004 to 2006 at the Surface Monitoring Of the Soil Reservoir Experiment (SMOSREX) bare soil experimental site. To reduce uncertainty from sampling depth mismatch, the soil moisture profile with a 1 cm thickness from a calibrated Hydus-1D (HID) model is utilized to determine the optimal soil moisture inputs to soil emission model. Uncertainties of 15 literature-based roughness effect parameterization schemes developed for L-band Tb modeling are inter-compared. The "QM" model is further calibrated against multi-angle and dual polarization Tb observations at the SMOSREX bare soil site under different roughness conditions. Our results show that: (1) soil moisture sampling depth varies with soil moisture content and roughness condition. When soil is drier and rougher, the soil moisture sampling depth gets deeper. (2) The 15 roughness schemes generally perform better at vertical polarization than at horizontal polarization and better when soil surface is relative smooth than when soil surface gets rougher. The 15 roughness correction schemes have their own advantages and disadvantages with diverse error and bias characteristics. None of them has a superior performance at all conditions in terms of roughness, polarizations and incident angles. (3) A non-zero Q configuration is preferred in parameter retrieval experiments and the observed linear relationship between Delta N and root-mean-square height (sigma) or sigma(2)/L-c can only be reproduced when Qis non-zero in parameter retrieval. (4) The effective roughness parameters (Q, N-p, and h) generally increase when soil get rougher. The calibrated Q, N-h, and N-v, show exponential dependence on the effective parameter h. The calibrated h still shows dependence on surface soil moisture after accounting the impact from soil sampling depth and also shows strong power-law dependence on Tb at incident angle of 40 degrees. The non-zero-Q fitting models have comparable performance in Tb modeling with zero-Q models but may be more physically realistic. (C) 2017 Elsevier Inc. All rights reserved.

通讯作者地址: Peng, B; Zhao, TJ (通讯作者)

Chinese Acad Sci, Inst Remote Sensing & Digital Earth, State Key Lab Remote Sensing Sci, Beijing 100101, Peoples R China.

地址:

[ 1 ] Chinese Acad Sci, Inst Remote Sensing & Digital Earth, State Key Lab Remote Sensing Sci, Beijing 100101, Peoples R China

[ 2 ] Univ Illinois, Natl Ctr Supercomp Applicat, Urbana, IL 61801 USA

[ 3 ] Univ Illinois, Dept Nat Resources & Environm Sci, Urbana, IL 61801 USA

[ 4 ] Tsinghua Univ, Key Lab Earth Syst Modeling, Minist Educ, Beijing 100084, Peoples R China

[ 5 ] Tsinghua Univ, Dept Earth Syst Sci, Beijing 100084, Peoples R China

[ 6 ] Univ Toulouse, IRD, CNRS, UPS,CNES,CESBIO, F-31400 Toulouse, France

[ 7 ] Univ Pittsburgh, Dept Civil & Environm Engn, Pittsburgh, PA 15261 USA