Developing remote sensing inversion algorithms
Two directions have been pursued: focusing on improved estimation of individual variables, and exploring a new inversion scheme for estimating a set of variables simultaneously based on data assimilation approach.
The inversion accuracy also depends on the quality of input data, which has to be controlled by different pre-processing techniques. Integration of high-level products as a post-processing technique is another way to improve the accuracy and quality of the final satellite products.
The following text contains the papers in four categories: 1). Review papers; 2). Pre-processing and post-processing techniques; 3). Algorithms for estimating individual variables, including radiative transfer modeling; LAI; FAPAR, FVC, incident shortwave radiation; incident PAR; surface albedo; LST; broadband emissivity; longwave net radiation, all-wave net radiation; ET; soil moisture and other hydrological variables; GPP; biomass and crop yield; TOA fluxes; and image classification; and 4). A data assimilation based new inversion scheme.
· Multiangular remote sensing (Liang et al. 2000a)
· Angular correction in radiation budget (Liang et al. 2000b);
· Canopy radiative transfer modeling(Qin and Liang 2000)
· Snow/soil radiative transfer modeling (Nolin and Liang 2000)
· Biophysical parameter retrieval (Liang 2007)
· Earth’s energy budget (Liang 2017; Liang et al. 2010)
· Surface radiation budget (Liang et al. 2009)
· Integration of high-level products (Liang et al. 2017).
· Calibrating Chinese environmental (HJ) (Jiang et al. 2013) and meteorological (FY) (Kim et al. 2014a) satellites, and also on Landsat-5 TM (Kim et al. 2014b) data;
· Quantifying MODIS thermal errors (Liu et al. 2006b).
· Correcting the effects of spatial variations in aerosol loadings for ETM+ (Fallah-Adl et al. 1997; Fallah-Adl et al. 1996; Liang et al. 1997a; Liang et al. 2001; Liang et al. 2002b), MODIS(Guang et al. 2012; Liang et al. 2006b; Zhong et al. 2007), ASTER(Guang et al. 2011);
· Correcting water vapor effects for AVIRIS and Hyperion hyperspectral data(Liang and Fang 2004; Zhao et al. 2008);
· Atmospheric correction for multi-angular data (Liang and Strahler 1994a).
· Re-constructing surface reflectance from the re-processed NDVI data (Xiao et al. 2017b);
· Reconstructing high-level products (Fang et al. 2007; Fang et al. 2008b; Lu et al. 2007).
· Optimal Interpolation (OI) and Empirical Orthogonal Functions (EOF) for integrating the LAI and evapotranspiration products (Feng et al. 2016a; Wang and Liang 2011, 2014);
· Multi-Resolution Tree (MRT) method for integrating multiple products of land surface broadband albedo (He et al. 2014b), FAPAR (Tao et al. 2017) and land surface emissivity (Shi et al. 2016b);
· Bayesian Model Averaging (BMA) method for integrating surface longwave downward radiation and surface latent heat flux products (Chen et al. 2015; Wu et al. 2012b; Yao et al. 2014a).
A series of radiative transfer models of the Earth’s surface have been developed which coupled elements of the soil-vegetation-atmosphere system. These models have been proven particularly valuable for linking remote sensing observations with environmental variables, and have consequently led to the development of new inversion algorithms;
· Canopy RT modeling(Albers et al. 1990; Liang and Strahler 1993a; Liang and Strahler 1993b; Liang and Strahler 1995; Liang et al. 1997b);
· Atmospheric RT modeling (Liang and Lewis 1996; Liang and Strahler 1994b);
· Soil RT modeling (Liang 1997; Liang and Mishchenko 1997; Liang and Townshend 1996a, b);
· Snow RT modeling (Cheng and Liang 2011; Cheng et al. 2010b);
· Simulating LAI and albedo scaling(Liang 2000) and thermal scaling(Su et al. 2003).
· MODIS(Chai et al. 2012; Fang and Liang 2005; Fang et al. 2008b; Wang et al. 2008a; Xiao et al. 2014; Zhang et al. 2012; Zhou et al. 2017);
· AVHRR(Xiao et al. 2016a);
· VIIRS(Xiao et al. 2016b);
· MISR and VEGETATION(Liu et al. 2014; Ma et al. 2017b; Wan et al. 2009);
· Landsat TM/ETM+ (Fang and Liang 2003; Fang et al. 2003; Fang et al. 2005; Walthall et al. 2004);
· EO1 ALI(Liang et al. 2003a);
· simulation data(Gong et al. 1999);
· Data assimilation based algorithm(Liu et al. 2014; Qin et al. 2008; Wang et al. 2010b; Xiao et al. 2011; Xiao et al. 2009; Xiao et al. 2012);
· Validation and inter-comparison: evaluation of long-time series LAI products from AVHRR data (Xiao et al. 2017a), time series analysis (Jiang et al. 2010), validation of products(Fang et al. 2012), and product intercomparson(Fang et al. 2013a; Fang et al. 2004).
· Landsat/ETM+(Fang et al. 2005);
· MODIS(Tao et al. 2016; Xiao et al. 2016c);
· MISR(Tao et al. 2016);
· AVHRR(Xiao et al. 2016c);
· VIIRS (Xiao et al. 2016b).
· MODIS (Jia et al. 2015a; Yang et al. 2016);
· ETM+ (Jia et al. 2017b; Wang et al. 2017b; Yang et al. 2017);
· GF(Jia et al. 2016b);
· AVHRR (Jia et al. 2015b);
· Evaluating tree cover products(Liu et al. 2006a).
· Parameterization algorithm (Qin et al. 2015);
· Look-up table (LUT) methods for MODIS (Zhang et al. 2014a). GMS-5 (Lu et al. 2010), and MTSAT (Huang et al. 2011);
· Neural network method (Qin et al. 2011a);
· Calibrating satellite product with ground measurements(Zhang et al. 2016a), and comparing satellite products with reanalysis products(Zhang et al. 2016b);
· Product validation(Gui et al. 2010; Schroeder et al. 2009).
· LUT method (Liang et al. 2006a) for MODIS (Liu et al. 2008; Wang et al. 2010a). GOES (Zheng et al. 2008), and AVHRR (Liang et al. 2007) data;
· Estimating daily PAR from sunshine data(Qin et al. 2012) and from MODIS high-level products(Tang et al. 2017).
· Temporal scaling for daily PAR (Wang et al. 2010a; Zheng and Liang 2011)
· “direct estimation method” estimating albedo directly from satellite observations based on extensive radiative transfer simulations, different from the traditional approaches consisting of atmospheric correction, BRDF modeling, narrowband to broadband conversion: (Liang 2003; Liang et al. 1999; Liang et al. 2005a). It has been used for GLASS (Liu et al. 2013a; Liu et al. 2013b; Qu et al. 2014) and the VIIRS albedo production(Wang et al. 2013; Wang et al. 2017a), and for a variety of remotely sensed data, such as MODIS(Wang et al. 2015c), MISR (He et al. 2017a);AVIRIS (He et al. 2014a), Landsat (He et al. 2017b),ALI(Liang et al. 2003a), and HJ (He et al. 2015a);
· Narrowband to broadband albedo conversion (Liang 2001b; Liang et al. 2003b; Liang et al. 2005b);
· Estimating albedo and BRDF using the optimization method(He et al. 2012)and ensemble Kalman filter(Qin et al. 2006), and high-performance computing(Kalluri et al. 2001; Zhang et al. 1998);
· Estimating albedo over oceans (Feng et al. 2016b; Qu et al. 2016) and evaluating existing ocean albedo products(Cao et al. 2016);
· Evaluating and validating the global albedo products: MODIS(Jin et al. 2003a, b; Liang et al. 2002a; Román et al. 2013; Stroeve et al. 2005; Wang et al. 2010d), MISR(Chen et al. 2008; Taberner et al. 2010; Wu et al. 2012a), VIIRS (Zhou et al. 2016 ), and GEWEX/ISCCP(Qin et al. 2011b).
· Optimization method to estimate LST from multispectral thermal data(Liang 2001c), and also validating the MODIS (Wang et al. 2008b) and ASTER LST products(Wang and Liang 2009b);
· Estimating LST and spectral emissivity from hyperspectral data (Cheng et al. 2011a; Cheng et al. 2010a).
· The GLASS longwave broadband emissivity product algorithms based on conversion of shortwave spectral albedos for soils (Cheng and Liang 2013a; Cheng and Liang 2013b; Cheng et al. 2011b) and radiative transfer calculations for vegetation (Cheng et al. 2016);
· Angular effects (Cheng and Liang 2014), spectral range (Cheng et al. 2013b), validation (Cheng et al. 2014),and product evaluation(Cheng et al. 2013a); Empirical algorithm for determining the vegetation emissivity(Ren et al. 2013);
· Emissivity product applied to Earth system model simulation (Jin and Liang 2006);
· Ocean emissivity estimation algorithm(Cheng et al. 2017a).
· Downward and upwelling longwave radiation are estimated separately. Upwelling longwave radiation can be calculated from LST and broadband longwave emissivity, but the uncertainties of these two components may cause large errors (Wang and Liang 2009b).
· The direct estimation methods have been developed for estimating upwelling, downward and net longwave radiation (Cheng and Liang 2016; Cheng et al. 2017b; Wang and Liang 2009d, 2010; Wang et al. 2009b);
· Meteorological observations have been also used for calculating downward radiation(Wang and Liang 2009c);
· Product validation(Gui et al. 2010).
· Instead of adding all components together, we have developed the algorithms for converting shortwave net radiation in conjunction with other information. The algorithms for estimating shortwave net radiation have been developed for various sensors, such as MODIS(Huang et al. 2012; Kim and Liang 2010); Landsat(Wang et al. 2014),MERSI(Wang et al. 2015a), AVIRIS(He et al. 2015b; Wang et al. 2015b) , and incident shortwave radiation using other methods(Zhang et al. 2016a; Zhang et al. 2016b);
· Comparing different linear formulae (Jiang et al. 2015), machine learning techniques (Jiang et al. 2014), the MARS algorithm is used for producing the GLASS day-time all-wave net radiation (Jiang et al. 2016);
· Empirical algorithms (Wang and Liang 2009a);
· Validation and comparison with other products (Jia et al. 2016a; Jia et al. 2017a).
· Empirical ET algorithms (Wang et al. 2009a; Wang et al. 2010c; Wang et al. 2010e, f; Wang and Liang 2008; Yao et al. 2010a; Yao et al. 2010b);
· Priestley–Taylor type algorithms(Yao et al. 2015; Yao et al. 2017c; Yao et al. 2014c; Yao et al. 2013);
· Penman-Monteith type algorithm (Li et al. 2014; Sun et al. 2013; Yuan et al. 2012b);
· Integrated algorithms (Feng et al. 2015; Feng et al. 2016a; Yao et al. 2017a; Yao et al. 2014a; Yao et al. 2017b; Yao et al. 2014b; Yao et al. 2016)
· Energy budget based model from ASTER (Galleguillos et al. 2011)
· Regression tree method(Xia et al. 2014a);
· Product evaluations (Chen et al. 2014);
· Data assimilation methods for estimating ET/heat fluxes (Bateni and Liang 2012; Qin et al. 2005; Qin et al. 2007; Xu et al. 2015; Xu et al. 2014; Xu et al. 2011a; Xu et al. 2011b).
· Data assimilation methods for predicting river variables (Meng et al. 2017; Xie et al. 2014)