1. APPLICATIONS OF THE INTEGRAL EQUATION MODEL IN MICROWAVE REMOTE SENSING OF LAND SURFACE PARAMETERS In Honor of Prof. Adrian K. Fung Kun-Shan Chen National Central University, Taiwan Jiancheng Shi Institute of Remote Sensing Applications, CSA, Beijing, China & University of California, Santa Barbara

2. Current Microwave Surface Scattering Models Importance of surface scattering modeling Direct component of soil moisture and ocean properties Boundary conditions for many other investigations of Earth geophysical properties (vegetation, snow, atmospheric properties) Physical based surface scattering and emission models –Tradition models Small Perturbation Model Physical Optical Model Geometrical Optical Model –Integral Equation Model(s) (IEM, AIEM: analytical solution of above 3 models) –Monte Carlo Model

3. Outline 1.Validation of IEM with 3D Monte Carlo simulated data and field measurements 2.Two examples for Multi-frequency AMSR-E and L- band SMOS and SMAP Soil surface parameterized model development; Inversion model development; Validation with ground radiometer measurements

4. Why do we need a simple surface Emission model? 1. Complex and computational intensive of AIEM - Image based analyses for global scale require a simple model 2. The simple model directly serves as the inversion model for soil moisture estimation 3.The simple model also serves as the boundary condition for other geophysical and atmospheric study Microwave signals 4. Current available semi-empirical models Often derived from the limited experimental data. There are many uncertainties Most of available models fails to describe the characteristics of effects of surface roughness on emission signals at large incidence and high frequencies (AMSR-E, SSM/I, SSM/R, WINSAT, CIMS)

5. Numerical Simulations Using IEM&AIEM Development of the parameterized simple models and inversion algorithms from AIEM model simulated database for a wide range of soil dielectric and roughness conditions

6. Effects of Surface Roughness on Effective Reflectivities Common understanding : surface roughness results in a decrease of the surface effective reflectivity or an increase of emissivity It was found: surface roughness can result in a decreasing surface emissivity in V polarization <= both Monte Carlo and IEM models at high angle

7. Monte Carlo Simulation At 50° - 257 cases rms height: 0.035, 0.05, 0.1, 0.12, 0.15, 0.3, and 0.41 wavelength  correlation length: 0.17 – 1.3 wavelength  Dielectric constant: 3.6 – 24.6 EvEv EhEh 40°50° At 40° - 216 cases rms height: 0.05, 0.1, and 0.15 wavelength  correlation length: 0.33 – 1 wavelength  Dielectric constant: 4.06 – 24.6 Both with Gauss function

8. Validation of AIEM for Emission with Monte Carlo Model RMSE=0.01 RMSE=0.008RMSE=0.017 RMSE=0.013

9. Validation of AIEM Model with Field Experimental Data INRA’93 ground multi-frequency (5.05, 10.65, 23.8, and 36.5 GHz) and polarization (V & H) radiometer experimental data at 50°

10. Sensor Specifications Launched on May 4, 2002 Sun-synchronous orbit Equatorial crossing at 13:30 LST (ascending) AQUA Satellite First Example for Soil Moisture Algorithm Development for AMSR-E 12 channel, 6 frequency conically scanning passive microwave radiometer Earth incidence angle of 55° Built by the Japan Aerospace Exploration Agency (JAXA) AMSR-E: Advanced Microwave Scanning Radiometer

11. Comparing Qp and AIEM Models Frequency in GHz 6.925 10.65 18.7 23.8 36.5 0.0016 0.0012 0.0011 0.0011 0.0012 0.0023 0.0022 0.0017 0.0019 0.0016 V Polarization H Polarization New Qp model Qp is the polarization dependent roughness parameters

12. Surface Roughness Parameterization for Qp Model The surface roughness parameters Qp are highly correlated with the ratio of rms height –s and correlation length – l (proportion to random rough surface slope). s/l

13. Relationship in Roughness Parameters Qp High correlation in roughness parameters can be found between Qh and Qv at different frequencies Q h (f) = a (f)+ b(f)*Q v QvQv QhQh 6.925 GHz10.65GHz18.7 GHz36.5 GHz Est. Q v Q v

14. Inverse algorithm for Bare Surface After re-range, the algorithm: Left side of Eq is from the measurements Right side of Eq is only dependent on surface dielectric constant Therefore

15. Inverse algorithm Accuracies from AIEM Simulated Data Input Mv in % Estimated Mv in % 6.925 GHz 36.5 GHz18.7 GHz 10.65 GHz RMSE=0.44% RMSE=0.30% RMSE=0.28%

16. Inverse algorithm Validation with INRA’93 Experimental Data at 50° RMSE=3.7% RMSE=3.5%RMSE=3.6% RMSE=3.5%

17. Inverse algorithm Validation with USDA BARC (1979-1981) Experimental Data RMSE:2.9% RMSE:3.7% RMSE:3.6% RMSE:3.8%

18. Current and Future satellite L-band radiometers: SMOS – Multi-incidence, 50 km resolution, V and H polarization SMAP – Passive: 40 km, V and H polarizations, active: 1 – 3 km, VV, HH, and VH polarizations. SMOS SMAP Second Example: Applications for L-band Sensors

19. The Parameterized L-band Surface Emissivity Model The parameterized surface emissivity Model V H Absolute and ratio accuracies between IEM and the parameterized model RMSE Viewing Angle and are the effective and fresnel reflectivity. A and B are parameters depending on the roughness

20. High correlation in roughness parameters can be found After re-range, the algorithm can be developed AvAv Av/BvAv/Bv Ah/BhAh/Bh AhAh BhBh Bv/BhBv/Bh Then 40° L-band Inversion Model

21. Validation of Bare Surface Algorithm Using L- band Radiometer Measurements (79-82) at USDA- BARC 20°30°40° 50°60° RMSE bias RMSE=2.9 % RMSE=3.1 % RMSE=2.8 %RMSE=2.6 % RMSE=3.6 %

22. Summary on IEM/AIEM Contributions Providing an important tool for algorithm(s) development in Earth surface geophysical properties retrieval Other application examples: 1.Soil Moisture retrieval for L-band radar (SMAP and POLSAR, Sun et al., IGARSS 2010 ) 2.Retrieval vegetation properties for AMRS-E ( Shi et al., RSE, 112(12) 4285-4300, 2008 ) and for SMOS ( Chen et al., IEEE/GRSL 7(1):127-130, 2010 ) 3.Snow parameterized model(s) for AMSR-E ( Jiang et al., RSE, 111 (2-3) 357-366, Nov. 2007 and CoreH2O ( Du et al., RSE, 114 （ 5 ） ： 1089-1098, 2010 )