Kernel-driven Bidirectional Reflectance Distribution Function Model Research Articles

The influence of BRDF effects and representativeness of training data on tree species classification using multi-flightline airborne hyperspectral imagery

Airborne hyperspectral imagery provides detailed spatial and spectral information that can be used for tree species classification, but its application in practical forest inventories over large areas is limited. This study investigates the impact of (i) bidirectional reflectance distribution function (BRDF) effects on multi-flightline airborne imagery and (ii) the selection strategy for training data on tree species classification accuracy over large forested areas. Hyperspectral imagery was acquired using the Chinese Academy of Forestry’s LiDAR, CCD, and Hyperspectral system (CAF-LiCHy) over the Mengjiagang Forest Farm (MFF, ∼155 km2) in northeastern China, which includes common keystone coniferous species in northern China’s temperate-boreal ecozones. A kernel-driven BRDF model based on mNDVI704.50 stratification was applied to airborne hyperspectral data acquired from multi-flightlines. We developed three training data selection strategies, namely #1 sampling over the entire area, #2 localized sampling over the area for each flight date, and #3 gradually decreasing sampling in size. Additionally, two validation assessment strategies were respectively employed: independent data over the entire area and independent data in areas with inconsistent classification labeling between using BRDF corrected and uncorrected hyperspectral imagery. These strategies were implemented to perform a comprehensive cross-comparison of tree species classification accuracy using BRDF corrected and uncorrected airborne hyperspectral imagery. The study found that the representativeness of the training data (in terms of solar-viewing geometry distribution and size) has a stronger impact on the classification accuracy of tree species than the BRDF effects in airborne hyperspectral imagery. By utilizing BRDF corrected hyperspectral imagery with a majority of representative training data, the mean overall accuracy of tree species classification improves by ≤ 2.9 % compared to uncorrected imagery. Notably, BRDF correction has improved classification accuracy by 11.5 % in inconsistently labeled areas, mainly concentrated in the spatial overlap regions of adjacent flightlines. These findings provide valuable insights for improving tree species classification accuracy in large forested areas using multi-flightline airborne hyperspectral imagery.

Research on the Directional Characteristics of the Reflectance of Oil-Contaminated Sea Ice

Remote sensing has been widely used for oil spill monitoring in open waters. However, research on remote sensing monitoring of oil spills in ice-infested sea waters (IISWs) is still scarce. The spectral characteristics of oil-contaminated sea ice (OCSI) and clean sea ice (CSI) and their differences are an important basis for oil spill detection using visible/near-infrared (VNIR) remote sensing. Such features and differences can change with the observation geometry, affecting the identification accuracy. In this study, we carried out multi-angle reflection observation experiments of oil-contaminated sea ice (OCSI) and proposed a kernel-driven bidirectional reflectance distribution function (BRDF) model, Walthall–Ross thick-Litransit-Lisparse-r-RPV (WaRoLstRPV), which takes into account the strong forward-scattering characteristics of sea ice. We also analyzed the preferred observation geometry for oil spill monitoring in IISWs. In the validation using actual measured data, the proposed WaRoLstRPV performed well, with RMSEs of 0.0031 and 0.0026 for CSI and OCSI, respectively, outperforming the commonly used kernel-driven BRDF models, Ross thick-Li sparse (R-LiSpr), QU-Roujean (Qu-R), QU-Lisparse R-r-RPV (Qu-LiSpr-RrRPV), and Walthall (Wa). The observation geometry with a zenith angle around 50° and relative azimuth ranging from 250° to 290° is preferred for oil spill detection in IISWs.

A new snow bidirectional reflectance distribution function model in spectral regions from UV to SWIR: Model development and application to ground-based, aircraft and satellite observations

A generalized and flexible snow Bidirectional Reflectance Distribution Function (BRDF) model is proposed to describe the snow anisotropic properties in the spectral ranges from UltraViolet (UV) to Short-Wave InfraRed (SWIR). The new development belongs to the Fast and Accurate Semi-analytical Model of Atmosphere-surface Reflectance (FASMAR) family. FASMAR is developed, applied to and validated by ground-, aircraft- and satellite observations. FASMAR is a kernel-driven BRDF model. It contains four major kernels: isotropy kernel, asymptotic radiative transfer kernel, single-scattering kernel, and auxiliary kernels. These kernels enable FASMAR to characterize the angular-dependent scattering features and wavelength-dependent absorbing patterns of the snow BRDF properties. The comparison between FASMAR and RossThick-LiSparseReciprocal-Snow (RTLSRS) model were performed using radiative transfer simulations, ground-, aircraft- and satellite observations at selected wavelengths 380, 480, 670, 870, 1220, and 2200 nm. For Solar Zenith Angle (SZA) smaller than 70°, FASMAR shows somehow better accuracy compared to the RTLSRS model. FASMAR’s accuracy keeps quite stable for large SZAs (SZA larger than 70°). The relative difference between FASMAR calculations and radiative transfer simulations is within 5% for wavelengths between 380 and 2200 nm. A similar accuracy of FASMAR is observed compared to aircraft BRDF measurements, with a relative difference of less than 5%. FASMAR also shows good agreement with ground-based and satellite observations, with correlation coefficients larger than 0.9 and relative errors smaller than 2% for UV–vis and 4% for NIR-SWIR spectral ranges. The comparisons additionally reveal that the accuracy of FASMAR is not strongly affected by the a priori knowledge of the snow grain size. FASMAR can provide snow BRDF estimation with good accuracy in spectral regions from UV to SWIR under all observation/illumination geometries. FASMAR is used support the operational retrieval of the snow properties from satellite observations.

Evaluation of BRDF Information Retrieved from Time-Series Multiangle Data of the Himawari-8 AHI

Recently, much attention has been given to using geostationary Earth orbit (GEO) meteorological satellite data for retrieving land surface parameters due to their high observation frequencies. However, their bidirectional reflectance distribution function (BRDF) information content with a single viewing angle has not been sufficiently investigated, which lays a foundation for subsequent quantitative estimation. In this study, we aim to comprehensively evaluate BRDF information from time-series observations from the Advanced Himawari Imager (AHI) onboard the GEO satellite Himawari-8. First, ~6.2 km monthly multiangle surface reflectances from POLDER onboard a low-Earth-orbiting (LEO) satellite with good angle distributions over various land types during 2008 were used as reference data, and corresponding 0.05° high-quality MODIS (i.e., onboard LEO satellites) and AHI datasets during four months in 2020 were obtained using cloud and aerosol property products. Then, indicators of angle distribution, BRDF change, and albedos were retrieved by the kernel-driven Ross-Li BRDF model from the three datasets, which were used for comparisons over different time spans. Generally, the quality of sun-viewing geometries varies dramatically for accumulated AHI observations according to the weight-of-determination, and wide-ranging anisotropic flat indices are obtained. The root-mean-square-errors of white sky albedos between AHI and MODIS half-month data are 0.018 and 0.033 in the red and near-infrared bands, respectively, achieving smaller values of 0.004 and 0.007 between the half-month and daily AHI data, respectively, due to small variances in sun-viewing geometries. The generally wide AHI BRDF variances and good consistency in albedo with MODIS show their potential for retrieving anisotropy information and albedo, while angle accumulation quality of AHI time-series observations must be considered.

ESTIMATION OF LAND SURFACE ALBEDO FROM GCOM-C/SGLI SURFACE REFLECTANCE

Abstract. This paper examines algorithms for estimating terrestrial albedo from the products of the Global Change Observation Mission – Climate (GCOM-C) / Second-generation Global Imager (SGLI), which was launched in December 2017 by the Japan Aerospace Exploration Agency. We selected two algorithms: one based on a bidirectional reflectance distribution function (BRDF) model and one based on multi-regression models. The former determines kernel-driven BRDF model parameters from multiple sets of reflectance and estimates the land surface albedo from those parameters. The latter estimates the land surface albedo from a single set of reflectance with multi-regression models. The multi-regression models are derived for an arbitrary geometry from datasets of simulated albedo and multi-angular reflectance. In experiments using in situ multi-temporal data for barren land, deciduous broadleaf forests, and paddy fields, the albedos estimated by the BRDF-based and multi-regression-based algorithms achieve reasonable root-mean-square errors. However, the latter algorithm requires information about the land cover of the pixel of interest, and the variance of its estimated albedo is sensitive to the observation geometry. We therefore conclude that the BRDF-based algorithm is more robust and can be applied to SGLI operational albedo products for various applications, including climate-change research.

Extending a Linear Kernel-Driven BRDF Model to Realistically Simulate Reflectance Anisotropy Over Rugged Terrain

Bidirectional reflectance distribution function (BRDF) models are used to correct surface bidirectional effects and estimate land surface albedo. Many operational BRDF/albedo algorithms adopt a Roujean linear kernel-driven BRDF (RLKB) model because of its simple form and good performance in fitting multidirectional surface reflectance values. However, this model does not explicitly consider topographic effects, resulting in errors when applied over rugged terrain. To address this issue, we proposed a hybrid algorithm suitable for both flat and rugged terrain, called topographical kernel-driven (Topo-KD). First, we constructed a linear kernel-driven BRDF model considering terrain (LKB_T) which describes the topographic effects with a mountain radiative transfer (MRT) model. Then, the Topo-KD algorithm adaptively selects the most suitable model (RLKB or LKB_T) according to the terrain conditions and fitting residuals. The performances of Topo-KD and RLKB using the RossThick–LiSparseReciprocal (RTLSR) kernel are compared using simulated data sets and moderate-resolution imaging spectroradiometer (MODIS) observations. The results show that the BRDF of the pixel is affected by topography. But the RTLSR model does not specifically account for it, resulting in larger biases over rugged terrain than the Topo-KD algorithm in both the red and near-infrared (NIR) bands. The experiment using MODIS data sets demonstrates that the Topo-KD algorithm reduces fitting residuals in the red and NIR bands by 21.5% and 27.4% compared with the RTLSR model. These results indicate that the Topo-KD algorithm can be a better choice for retrieving land surface parameters and describing the radiative transfer process in mountainous areas.

A kernel-driven BRDF model to inform satellite-derived visible anvil cloud detection

Abstract. Satellites routinely observe deep convective clouds across the world. The cirrus outflow from deep convection, commonly referred to as anvil cloud, has a ubiquitous appearance in visible and infrared (IR) wavelength imagery. Anvil clouds appear as broad areas of highly reflective and cold pixels relative to the darker and warmer clear sky background, often with embedded textured and colder pixels that indicate updrafts and gravity waves. These characteristics would suggest that creating automated anvil cloud detection products useful for weather forecasting and research should be straightforward, yet in practice such product development can be challenging. Some anvil detection methods have used reflectance or temperature thresholding, but anvil reflectance varies significantly throughout a day as a function of combined solar illumination and satellite viewing geometry, and anvil cloud top temperature varies as a function of convective equilibrium level and tropopause height. This paper highlights a technique for facilitating anvil cloud detection based on visible observations that relies on comparative analysis with expected cloud reflectance for a given set of angles, thereby addressing limitations of previous methods. A 1-year database of anvil-identified pixels, as determined from IR observations, from several geostationary satellites was used to construct a bidirectional reflectance distribution function (BRDF) model to quantify typical anvil reflectance across almost all expected viewing, solar, and azimuth angle configurations, in addition to the reflectance uncertainty for each angular bin. Application of the BRDF model for cloud optical depth retrieval in deep convection is described as well.

Estimating Forest Canopy Height Using MODIS BRDF Data Emphasizing Typical-Angle Reflectances

Forest-canopy height is an important parameter for the estimation of forest biomass and terrestrial carbon flux and climate-change research at regional and global scales. Currently, various methods combining Light Detection and Ranging (LiDAR) data with various auxiliary data, particularly satellite remotely sensed reflectances, have been widely used to produce spatially continuous canopy-height products. However, current methods in use for remote sensing reflectances mainly focus on the nadir view direction, while anisotropic reflectances, which are theoretically more sensitive to the forest canopy height in the multiangle remote sensing field, have rarely been explored. Here, we attempted to examine the potential of using modeled multiangle reflectances at three typical viewing angles (i.e., from the hotspot, darkspot, and nadir directions) to estimate forest-canopy height as auxiliary data sources. First, the sensitivities of the typical angular reflectances as a function of forest canopy height were fully examined using the Extended Fourier Amplitude Sensitivity Test (EFAST) method based on the 4-scale Bidirectional Reflectance Distribution Function (BRDF) model simulations. This indicated that reflectances in the off-nadir viewing directions are generally sensitive to canopy-height variations. Then, the canopy heights were extracted from airborne Laser Vegetation Imaging Sensor (LVIS) data, which were further divided into training and validation data. Moderate Resolution Imaging Spectroradiometer (MODIS) multiangle reflectances at typical viewing angles were calculated from the MODIS BRDF parameter product (MCD43A1, version 6) as partial training-input data, based on a hotspot-adjusted, kernel-driven linear BRDF model. Subsequently, the Random Forest (RF) machine learning model was trained to acquire the relationship between the extracted canopy heights and the corresponding MODIS typical viewing reflectances. The trained model was further applied to estimate the canopy height metrics in the study areas of Howland Forest, Harvard Forest, and Bartlett Forest. Finally, the estimated canopy heights were independently validated by canopy heights extracted from the LVIS data. The results indicate that the canopy heights modeled through this method exhibit generally high accordance with the LVIS-derived canopy heights (R = 0.65−0.67; RMSE = 3.63−5.78). The results suggest that the MODIS multiangle reflectance data at typical observation angles contain important information regarding forest canopy height and can, therefore, be used to estimate forest canopy height for various ecological applications.

Evaluation of the Snow Albedo Retrieved from the Snow Kernel Improved the Ross-Roujean BRDF Model

The original kernel-driven bidirectional reflectance distribution function (BRDF) models were developed based on soil-vegetation systems. To further improve the ability of the models to characterize the snow surface scattering properties, a snow kernel was derived from the asymptotic radiative transfer (ART) model and used in the kernel-driven BRDF model framework. However, there is a need to further evaluate the influence of using this snow kernel to improve the original kernel-driven models in snow albedo retrieval applications. The aim of this study is to perform such an evaluation using a variety of snow BRDF data. The RossThick-Roujean (RTR) model is used as a framework for taking in the new snow kernel (hereafter named the RTS model) since the Roujean geometric-optical (GO) kernel captures a neglectable hotspot effect and represents a more prominent dome-shaped BRDF, especially at a small solar zenith angle (SZA). We obtained the following results: (1) The RTR model has difficulties in reconstructing the snow BRDF shape, especially at large SZAs, which tends to underestimate the reflectance in the forward direction and overestimate reflectance in the backward direction for various data sources. In comparison, the RTS model performs very well in fitting snow BRDF data and shows high accuracy for all data. (2) The RTR model retrieved snow albedos at SZAs = 30°–70° are underestimated by 0.71% and 0.69% in the red and near-infrared (NIR) bands, respectively, compared with the simulation results of the bicontinuous photon tracking (bic-PT) model, which serve as “real” values. However, the albedo retrieved by the RTS model is significantly improved and generally agrees well with the simulation results of the bic-PT model, although the improved model still somewhat underestimates the albedo by 0.01% in the red band and overestimates the albedo by 0.05% in the NIR band, respectively, at SZAs = 30°–70°, which may be negligible. (3) The albedo derived by these two models shows a high correlation (R2 > 0.9) between the field-measured and Polarization and Directionality of the Earth's Reflectances (POLDER) data, especially for the black-sky albedo. However, the albedo derived using the RTR model is significantly underestimated compared with the RTS model. The RTR model underestimates the black-sky albedo (white-sky albedo) retrievals by 0.62% (1.51%) and 0.93% (2.08%) in the red and NIR bands, respectively, for the field-measured data. The shortwave black-sky and white-sky albedos derived using the RTR model for the POLDER data are underestimated by 1.43% and 1.54%, respectively, compared with the RTS model. These results indicate that the snow kernel in the kernel-driven BRDF model frame is more accurate in snow albedo retrievals and has the potential for application in the field of the regional and global energy budget.

An assessment of the performance of two snow kernels in characterizing snow scattering properties

ABSTRACTThe kernel-driven RossThick-LiSparseReciprocal (RTLSR) bidirectional reflectance distribution function (BRDF) model has been widely used in the quantitative remote sensing community. However, the performance of this model is challenged when modelling the optical scattering properties of pure snow surfaces. Recently, two snow kernels have been developed to improve the snow anisotropic reflectance in the kernel-driven RTLSR model framework. However, the performances of these two snow kernels must be assessed to identify their potential applications. Therefore, we assess the performances of these two kernels using various BRDF data sources. Our findings demonstrate their differences in several aspects. (1) These two kernels differ in characterizing the variability in BRDF shape as a function of the solar zenith angle (SZA). As the SZA increases, the shape of snow kernel derived by the asymptotic radiative transfer (ART) model (hereinafter named the ART method) changes from a dome shape to bowl shape, which agrees well with the simulation data of the bicontinuous photon tracking (bic-PT) model. The shape of the snow kernel proposed by Qu et al. (hereinafter named the Qu method) based on the Rahman-Pinty-Verstraete (RPV) model maintains a bowl shape for all SZAs. These differences in the kernel performances affect their abilities to fit snow BRDF data with different SZAs. (2) The corrected RTLSR models, with their respective snow kernels, are generally able to model the forward-scattering properties of snow surfaces compared with the original RTLSR model. However, the ART method performs better in capturing the BRDF variations in snow surfaces than the Qu method. This assessment provides an improved understanding of the performance of these two snow kernels and, thus, suggests further applications for the ART snow kernel in the kernel-driven BRDF model framework in the near future.

Derivation of Kernel Functions for Kernel-Driven Reflectance Model Over Sloping Terrain

The importance of the bidirectional reflectance distribution function (BRDF) has been well documented in quantitative remote sensing. The semiempirical, kernel-driven BRDF model is widely used to generate operational BRDF/albedo products due to its simplicity and accuracy. However, the effect of topography is rarely coupled with a kernel-based BRDF model. In this paper, a new kernel-driven reflectance model for sloping terrain (KDST) was developed based on the framework of the RossThick-LiSparse-Reciprocal (RTLSR) model. The slope, aspect, geotropic nature of the tree crown governed by gravity, component spectra contrasts, and diffuse irradiance were considered in the KDST model. The performance of KDST was evaluated by 3-D discrete anisotropic radiative transfer (DART) simulations, in situ measurements, and HJ-1A/B constellation charge-coupled device satellite observations. Using DART simulations, KDST reduces the maximum biases of the RTLSR from 15.6% and 29.7% to 7.1% and 4.8% for the surface bidirectional reflectance factor and hemispherical-direction reflectance factor, respectively. Compared with in situ measurements, KDST improves the reflectance simulation accuracy for the red and near-infrared bands from 0.0172 (18.65%) and 0.022 (4.77%) to 0.0060 (6.51%) and 0.0043 (1.05%), respectively. The preliminary comparison results indicate that KDST is promising for reflectance simulation over rugged terrain.

Development of a snow kernel to better model the anisotropic reflectance of pure snow in a kernel-driven BRDF model framework

The linear kernel-driven RossThick-LiSparseReciprocal (RTLSR) bidirectional reflectance distribution function (BRDF) model was originally developed from the simplified scenarios of continuous and discrete vegetation canopies, and has been widely used to fit multiangle observations of vegetation-soil systems of the land surface in many fields. Although this model was not developed explicitly for snow surfaces, it can capture the geometric-optical effect caused by the shadowing of rugged or undulating snow surfaces. However, in this study, this model has been further developed to better characterize the scattering properties of snow surface, which can also exhibit strongly forward-scattering behavior. This study proposes a new snow kernel to characterize the reflectance anisotropy of pure snow based on the asymptotic radiative transfer (ART) model that assumes snow can be modeled as a semi-infinite, plane-parallel, weakly absorbing light scattering layer. This new snow kernel adopts a correction term with a free parameter α to correct the analytic form of the ART model that has been reported to underestimate observed snow reflectance in the forward-scattering direction in the principal plane (PP), particularly in cases of a large viewing zenith angle (>60°). This snow kernel has now been implemented in the kernel-driven RTLSR BRDF model framework in conjunction with two additional kernels (i.e., the volumetric scattering kernel and geometric-optical scattering kernel) and is validated using observed and simulated multiangle data from three data sources. Pure snow targets were selected from the extensive archive of the Polarization and Directionality of the Earth's Reflectance (POLDER) BRDF data. Antarctic snow field measurements, which were taken from the top of a 32-m-tall tower at Dome C Station and include 6336 spectral bidirectional reflectance factors (BRFs), were also utilized. Finally, a set of simulated BRFs, generated by a hybrid scattering snow model that combines the geometric optics with vector radiative transfer theory, were used to further assess the proposed method. We first retrieve the value of the free parameter α for a comprehensive analysis using single multiangle snow data with a sufficient BRDF sampling. Then, we determine the optimally fixed value of the α parameter as prior information for potential users. The new snow kernel method is shown to be quite accurate, presenting a high correlation coefficient (R2 = ~0.9) and a negligible bias between the modeled BRFs and the various snow BRDF validation data. The finding demonstrates that this snow kernel provides an improved potential compared to that of the original kernel-driven model framework for a pure snow surface in many applications, particularly those involving the global water cycle and radiation budget, where snow cover plays an important role.

Aerosol Optical Depth Retrieval From Landsat 8 OLI Images Over Urban Areas Supported by MODIS BRDF/Albedo Data

This letter presents a new algorithm that allows the retrieval of the aerosol optical depth (AOD) at a high (500 m) spatial resolution from Landsat 8 Operational Land Imager (OLI) data over urban areas. Because of the complex structure over urban surfaces, the bidirectional reflectance characteristic is obvious; however, most of the current aerosol retrieval algorithms over land do not account for the anisotropic effect of the surface. This letter improves the quality of AOD retrieval by providing the surface reflectance based on the multiyear MODIS bidirectional reflectance distribution function (BRDF)/Albedo model parameters product (MCD43A1) and the RossThick-LiSparse reciprocal kernel-driven BRDF model. The ground-based Aerosol Robotic Network (AERONET) AOD measurements from five sites located in urban and suburban areas are used to validate the AOD retrievals, and the MODIS Terra Collection 6 (C6) dark target/deep blue AOD products (MOD04) at 10-km spatial resolution are obtained for comparison. The validation results show that the AOD retrievals from the OLI images are well correlated with the AERONET AOD measurements ( $R = 0.987$ ), with a low root-mean-square error of 0.07, a mean absolute error of 0.036, and a relative mean bias of 1.029; approximately 95.3% of the collocations fall within the expected error. The analysis indicates that the BRDF is essential in ensuring the accuracy of AOD retrieval. Compared with the MOD04 AOD retrievals, the OLI AOD retrievals have better spatial continuity and higher accuracy. The new algorithm can provide continuous and detailed spatial distributions of the AOD over complex urban surfaces.

ANALYSIS OF ACCURACY OF MODIS BRDF PRODUCT (MCD43 C6) BASED ON MISR LAND SURFACE BRF PRODUCT – A CASE STUDY OF THE CENTRAL PART OF NORTHEAST ASIA

Abstract. EOS/MODIS land surface Bi-directional Reflectance Distribution Function (BRDF) product (MCD43), with the latest version C6, is one of the most important operational BRDF products with global coverage. The core sub-product MCD43A1 stores 3 parameters of the RossThick-LiSparseR semi-empirical kernel-driven BRDF model. It is important for confident use of the product to evaluate the accuracy of bi-directional reflectance factor (BRF) predicted by MCD43A1 BRDF model (mBRF). A typical region in the central part of Northeast Asia is selected as the study area. The performance of MCD43A1 BRDF model is analyzed in various observation geometries and phenological phases, using Multi-angle Imaging SpectroRadiometer (MISR) land-surface reflectance factor product (MILS_BRF) as the reference data. In addition, MODIS products MCD12Q1 and MOD/MYD10A1 are used to evaluate the impacts of land cover types and snow covers on the model accuracy, respectively. The results show an overall excellent performance of MCD43A1 in representing the anisotropic reflectance of land surface, with root mean square error (RMSE) of 0.0262 and correlation coefficient (R) of 0.9537, for all available comparable samples of MILS_BRF and mBRF pairs. The model accuracy varies in different months, which is related to the phenological phases of the study area. The accuracy for pixels labelled as ‘snow’ by MCD43 is obviously low, with RMSE/R of 0.0903/0.8401. Ephemeral snowfall events further decrease the accuracy, with RMSE/R of 0.1001/0.7715. These results provide meaningful information to MCD43 users, especially those, whose study regions are subject to phenological cycles as well as snow cover and change.

An algorithm for the retrieval of the clumping index (CI) from the MODIS BRDF product using an adjusted version of the kernel-driven BRDF model

The clumping index (CI) characterizes the grouping of foliage relative to a random spatial distribution of leaves and is an important structural parameter for plant canopies that can influence canopy radiation regimes. Consequently, the CI is very useful for ecological and meteorological models. One method used to retrieve the CIs of plant canopies is to construct a linear relationship between the CI and the normalized difference between hotspot and dark spot (NDHD) angular index. This method requires a particularly accurate reconstruction of hotspot signatures, which are difficult to measure. In this study, we propose a framework to retrieve CIs from Moderate Resolution Imaging Spectroradiometer (MODIS) bidirectional reflectance distribution function (BRDF) parameters, which are generally based on linear CI-NDHD equations. The main algorithm is designed to retrieve CIs in the closed interval [0.33, 1.00]. This range is derived from the CI-NDHD equations and is thus called as the physical range here, although a modified lower boundary can be implemented in the future if necessary. If CIs are outside of this range, then a backup algorithm is designed to reprocess these so-called outlier CIs. The hotspot-adjusted version of the RossThick-LiSparseReciprocal (RTLSR) model (i.e., the RTCLSR model) is employed to reconstruct the hotspot signatures for the MODIS BRDF parameters. This method simplifies the hotspot reconstruction by using two hotspot parameters that are not distinctly scale-dependent particularly in the context of an inhomogeneous coarse spatial resolution. To evaluate this algorithm framework, we collect dozens of global field-measured CIs and calculate their determination coefficient (R2), root mean square error (RMSE) and bias relative to MODIS CIs derived using both the main algorithm and the backup algorithm. Our results show that this framework can derive MODIS CIs with a high accuracy (i.e., R2 = 0.80 (0.72), RMSE = 0.07 (0.12), bias = −0.03 (−0.10)) using the main (backup) algorithms and that it shows promise for various ecological applications, especially in combination with the leaf area index (LAI).

Potential Investigation of Linking PROSAIL with the Ross-Li BRDF Model for Vegetation Characterization

Methods that link different models for investigating the retrieval of canopy biophysical/structural variables have been substantially adopted in the remote sensing community. To retrieve global biophysical parameters from multiangle data, the kernel-driven bidirectional reflectance distribution function (BRDF) model has been widely applied to satellite multiangle observations to model (interpolate/extrapolate) the bidirectional reflectance factor (BRF) in an arbitrary direction of viewing and solar geometries. Such modeled BRFs, as an essential information source, are then input into an inversion procedure that is devised through a large number of simulation analyses from some widely used physical models that can generalize such an inversion relationship between the BRFs (or their simple algebraic composite) and the biophysical/structural parameter. Therefore, evaluation of such a link between physical models and kernel-driven models contributes to the development of such inversion procedures to accurately retrieve vegetation properties, particularly based on the operational global BRDF parameters derived from satellite multiangle observations (e.g., MODIS). In this study, the main objective is to investigate the potential for linking a popular physical model (PROSAIL) with the widely used kernel-driven Ross-Li models. To do this, the BRFs and albedo are generated by the physical PROSAIL in a forward model, and then the simulated BRFs are input into the kernel-driven BRDF model for retrieval of the BRFs and albedo in the same viewing and solar geometries. To further strengthen such an investigation, a variety of field-measured multiangle reflectances have also been used to investigate the potential for linking these two models. For simulated BRFs generated by the PROSAIL model at 659 and 865 nm, the two models are generally comparable to each other, and the resultant root mean square errors (RMSEs) are 0.0092 and 0.0355, respectively, although some discrepancy in the simulated BRFs can be found at large average leaf angle (ALA) values. Unsurprisingly, albedos generated by the method are quite consistent, and 99.98% and 97.99% of the simulated white sky albedo (WSA) has a divergence less than 0.02. For the field measurements, the kernel-driven model presents somewhat better model-observation congruence than the PROSAIL model. The results show that these models have an overall good consistency for both field-measured and model-simulated BRFs. Therefore, there is potential for linking these two models for looking into the retrieval of canopy biophysical/structural variables through a simulation method, particularly from the current archive of the global routine MODIS BRDF parameters that were produced by the kernel-driven BRDF model; however, erectophile vegetation must be further examined.

BIDIRECTIONAL REFLECTANCE MODELING OF THE GEOSTATIONARY SENSOR HIMAWARI-8/AHI USING A KERNEL-DRIVEN BRDF MODEL

Himawari-8/AHI is a new geostationary sensor that can observe the land surface with high temporal frequency. Bidirectional reflectance derived by the Advanced Himawari Imager (AHI) includes information regarding land surface properties such as albedo, vegetation condition, and forest structure. This information can be extracted by modeling bidirectional reflectance using a bidirectional reflectance distribution function (BRDF). In this study, a kernel-driven BRDF model was applied to the red and near infrared reflectance observed over 8 hours during daytime to express intraday changes in reflectance. We compared the goodness of fit for six combinations of model kernels. The Ross-Thin and Ross-Thick kernels were selected as the best volume kernels for the red and near infrared bands, respectively. For the geometric kernel, the Li-sparse-Reciprocal and Li-Dense kernels displayed similar goodness of fit. The coefficient of determination and regression residuals showed a strong dependency on the azimuth angle of land surface slopes and the time of day that observations were made. Atmospheric correction and model adjustment of the terrain were the main issues encountered. These results will help to improve the BRDF model and to extract surface properties from bidirectional reflectance.

BIDIRECTIONAL REFLECTANCE MODELING OF THE GEOSTATIONARY SENSOR HIMAWARI-8/AHI USING A KERNEL-DRIVEN BRDF MODEL

Abstract. Himawari-8/AHI is a new geostationary sensor that can observe the land surface with high temporal frequency. Bidirectional reflectance derived by the Advanced Himawari Imager (AHI) includes information regarding land surface properties such as albedo, vegetation condition, and forest structure. This information can be extracted by modeling bidirectional reflectance using a bidirectional reflectance distribution function (BRDF). In this study, a kernel-driven BRDF model was applied to the red and near infrared reflectance observed over 8 hours during daytime to express intraday changes in reflectance. We compared the goodness of fit for six combinations of model kernels. The Ross-Thin and Ross-Thick kernels were selected as the best volume kernels for the red and near infrared bands, respectively. For the geometric kernel, the Li-sparse-Reciprocal and Li-Dense kernels displayed similar goodness of fit. The coefficient of determination and regression residuals showed a strong dependency on the azimuth angle of land surface slopes and the time of day that observations were made. Atmospheric correction and model adjustment of the terrain were the main issues encountered. These results will help to improve the BRDF model and to extract surface properties from bidirectional reflectance.

Performance evaluation of four directional emissivity analytical models with thermal SAIL model and airborne images

Land surface emissivity is a crucial parameter in the surface status monitoring. This study aims at the evaluation of four directional emissivity models, including two bi-directional reflectance distribution function (BRDF) models and two gap-frequency-based models. Results showed that the kernel-driven BRDF model could well represent directional emissivity with an error less than 0.002, and was consequently used to retrieve emissivity with an accuracy of about 0.012 from an airborne multi-angular thermal infrared data set. Furthermore, we updated the cavity effect factor relating to multiple scattering inside canopy, which improved the performance of the gap-frequency-based models.

Angular Normalization of Land Surface Temperature and Emissivity Using Multiangular Middle and Thermal Infrared Data

This paper aimed at the case of nonisothermal pixels and proposed a daytime temperature-independent spectral indices (TISI) method to retrieve directional emissivity and effective temperature from daytime multiangular observed images in both middle and thermal infrared (MIR and TIR) channels by combining the kernel-driven bidirectional reflectance distribution function (BRDF) model and the TISI method. Four groups of angular observations and two groups of MIR and TIR channels with narrow and broad bandwidths were used to investigate the influence of angular observations and bandwidth on the retrieval accuracy. Model sensitivity analysis indicated that the new method can generally obtain directional emissivity and temperature with an error less than 0.015 and 1.5 K if the noise included in the measured directional brightness temperature (DBT) and atmospheric data was no more than 1.0 K and 10%, respectively. The analysis also indicated that 1) large-angle intervals among the angular observations and a larger viewing zenith angle, with respect to nadir direction, can improve the retrieval accuracy because those angle conditions can result in significant difference for components' fractions and DBT under different viewing directions; 2) narrow channels can produce better results than broad channels. The new method was finally applied to a multiangular MIR and TIR data set acquired by an airborne system, and a modified kernel-driven BRDF model was used for angular normalization to the surface temperature for the first time. The difference of the retrieved emissivity and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) emissivity was found to be approximately 0.012 in the study area.