Nadir BRDF-adjusted Reflectance Research Articles

Global 500 m seamless dataset (2000–2022) of land surface reflectance generated from MODIS products

Abstract. The Moderate Resolution Imaging Spectroradiometer (MODIS) is widely utilized for retrieving land surface reflectance to reflect plant conditions, detect ecosystem phenology, monitor forest fires, and constrain terrestrial energy budgets. However, the state-of-the-art MODIS surface reflectance products suffer from temporal and spatial gaps due to atmospheric conditions (e.g. clouds and aerosols), limiting their use in ecological, agricultural, and environmental studies. Therefore, there is a need for reconstructing spatiotemporally seamless (i.e. gap-filled) surface reflectance data from MODIS products, which is difficult due to the intrinsic inconsistency of observations resulting from various sun/view geometry and the prolonged missing values resulting from polar night or heavy cloud coverage, especially in monsoon season. We built a framework for generating the global 500 m daily seamless data cubes (SDC500) based on MODIS surface reflectance dataset, which contains the generation of a land-cover-based a priori database, bidirectional reflectance distribution function (BRDF) correction, outlier detection, gap filling, and smoothing. The first global spatiotemporally seamless land surface reflectance at 500 m resolution was produced, covering the period from 2000 to 2022. Preliminary evaluation of the dataset at 12 sites worldwide with different land cover demonstrated its robust performance. The quantitative assessment shows that the SDC500 gap-filling results have a root-mean-square error (RMSE) of 0.0496 and a mean absolute error (MAE) of 0.0430. The SDC500 BRDF correction results showed an RMSE of 0.056 and a bias of −0.0085 when compared with MODIS nadir BRDF-adjusted reflectance (NBAR) products, indicating the acceptable accuracy of both products. From a temporal perspective, the SDC500 eliminates abnormal fluctuations while retaining the useful localized feature of rapid disturbances. From a spatial perspective, the SDC500 shows satisfactory spatial continuity. In conclusion, the SDC500 is a well-processed global daily surface reflectance product, which can serve as the fundamental input for large-scale ecological, agricultural, and environmental applications and quantitative remote sensing studies. The SDC500 is available at http://data.starcloud.pcl.ac.cn/resource/27 (Liang et al., 2023b) or https://doi.org/10.12436/SDC500.27.20230701 (Liang et al., 2023a).

Analyzing GOES-R ABI BRDF-adjusted EVI2 time series by comparing with VIIRS observations over the CONUS

The Advanced Baseline Imager (ABI) instruments onboard the Geostationary Operational Environmental Satellite-R series (GOES-R) provide new opportunities to expand their applications in global terrestrial ecology and environmental sciences. It offers a 5–15 min temporal resolution and 0.5–1 km spatial resolution that greatly benefit those environmental applications where spatial resolution can be compromised in favor of more frequent imagery. However, ABI captures reflected radiation at varying sun positions throughout the day, causing significant diurnal variations. The varying off-nadir view across ABI full disk presents further challenges for generating comparable and consistent vegetation observations, such as the two-band Enhanced Vegetation Index (EVI2). To address these challenges, three Bidirectional Reflectance Distribution Function (BRDF) models (RossThick-LiSparse-Reciprocal (RTLSR), Walthall, and Tian models) are examined to generate high-quality BRDF-adjusted EVI2 based on GeoNEX GOES-16 ABI surface reflectance. The ABI BRDF retrievals are then systematically compared to the corresponding VIIRS (Visible Infrared Imaging Radiometer Suite) Nadir BRDF-Adjusted Reflectance (NBAR) products (VNP43 V001) retrieved from RTLSR model. The viewing geometry effects are further analyzed over various land cover types using time series analysis, Pearson Correlation Coefficients (PCC), and Relative Difference (RD) between ABI and VIIRS high-quality time series inversions. The results indicate that the RTLSR model fits well the diurnal ABI observed reflectance and EVI2 values (even for the hot-spot phenomena), and outperforms the Walthall and Tian models for generating ABI high-quality BRDF retrievals with consistently low Root Mean Square Error (RMSE<0.02) and Weight of Determination (WoD < 0.5) across the CONUS. The annual proportion of ABI high-quality BRDF retrievals with a 3-day moving window using the RTLSR model is spatially and statistically comparable to that of VIIRS NBAR with a 16-day retrieval period. Compared to the time series of high-quality VIIRS NBAR EVI2, the BRDF adjusted ABI EVI2 time series still show considerable impacts of view zenith angles over evergreen forest, shrubland, and savanna pixels. However, such impacts are not significant over spatially homogeneous pixels such as cropland and deciduous forests. These results emphasize the need for BRDF correction when considering the impact of off-nadir observations from geostationary satellites on vegetation monitoring. They also demonstrate the advantages of geostationary satellites in capturing rapid changes in greenness trajectories.

Continuity of Top-of-Atmosphere, Surface, and Nadir BRDF-Adjusted Reflectance and NDVI between Landsat-8 and Landsat-9 OLI over China Landscape

The successful launch of Landsat-9 marks a significant achievement in preserving the data legacy and ensuring the continuity of Landsat’s calibrated Earth observations. This study comprehensively assesses the continuity of reflectance and the Normalized Difference Vegetation Index (NDVI) between Landsat-8 and Landsat-9 Operational Land Imagers (OLIs) over diverse Chinese landscapes. It reveals that sensor discrepancies minimally impact reflectance and NDVI consistency. Although Landsat-9’s top-of-atmosphere (TOA) reflectance is slightly lower than that of Landsat-8, small root-mean-square errors (RMSEs) ranging from 0.0102 to 0.0248 for VNIR and SWIR bands (and larger RMSE for NDVI at 0.0422) fall within acceptable ranges for Earth observation applications. Applying atmospheric corrections markedly enhances reflectance uniformity and brings regression slopes closer to unity. Further, Bidirectional Reflectance Distribution Function (BRDF) adjustments improve comparability, ensuring measurement reliability, and the NDVI maintains robust consistency across various reflectance types, time series, and land cover classes. These findings affirm Landsat-9’s success in achieving data continuity within the Landsat program, allowing interchangeable use of Landsat-8 and Landsat-9 OLI data for diverse Earth observation purposes. Future research may explore specific sensor correlations across different vegetation types and seasons while integrating data from complementary platforms, such as Sentinel-2, to enhance the understanding of data continuity factors.

Improving Crop Mapping by Using Bidirectional Reflectance Distribution Function (BRDF) Signatures with Google Earth Engine

Recent studies have demonstrated the potential of using bidirectional reflectance distribution function (BRDF) signatures captured by multi-angle observation data to enhance land cover classification and retrieve vegetation architectures. Considering the diversity of crop architectures, we proposed that crop mapping precision may be enhanced by using BRDF signatures. We compared the accuracy of four supervised machine learning classifiers provided by the Google Earth Engine (GEE), namely random forest (RF), classification and regression trees (CART), support vector machine (SVM), and Naïve Bayes (NB), using the moderate resolution imaging spectroradiometer (MODIS) nadir BRDF-adjusted reflectance data (MCD43A4 V6) and BRDF and albedo model parameter data (MCD43A1 V6) as input. Our results indicated that using BRDF signatures leads to a moderate improvement in classification results in most cases, compared to using reflectance data from a single nadir observation direction. Specifically, the overall validation accuracy increased by up to 4.9%, and the validation kappa coefficients increased by up to 0.092. Furthermore, the classifiers were ranked in order of accuracy, from highest to lowest: RF, CART, SVM, and NB. Our study contributes to the development of crop mapping and the application of multi-angle observation satellites.

Land Surface Phenology Retrieval through Spectral and Angular Harmonization of Landsat-8, Sentinel-2 and Gaofen-1 Data

Land Surface Phenology is an important characteristic of vegetation, which can be informative of its response to climate change. However, satellite-based identification of vegetation transition dates is hindered by inconsistencies in different observation platforms, including band settings, viewing angles, and scale effects. Therefore, time-series data with high consistency are necessary for monitoring vegetation phenology. This study proposes a data harmonization approach that involves band conversion and bidirectional reflectance distribution function (BRDF) correction to create normalized reflectance from Landsat-8, Sentinel-2A, and Gaofen-1 (GF-1) satellite data, characterized by the same spectral and illumination-viewing angles as the Moderate-Resolution Imaging Spectroradiometer (MODIS) and Nadir BRDF Adjusted Reflectance (NBAR). The harmonized data are then subjected to the spatial and temporal adaptive reflectance fusion model (STARFM) to produce time-series data with high spatio–temporal resolution. Finally, the transition date of typical vegetation was estimated using regular 30 m spatial resolution data. The results show that the data harmonization method proposed in this study assists in improving the consistency of different observations under different viewing angles. The fusion result of STARFM was improved after eliminating differences in the input data, and the accuracy of the remote-sensing-based vegetation transition date was improved by the fused time-series curve with the input of harmonized data. The root mean square error (RMSE) estimation of the vegetation transition date decreased by 9.58 days. We concluded that data harmonization eliminates the viewing-angle effect and is essential for time-series vegetation monitoring through improved data fusion.

Land surface phenology retrievals for arid and semi-arid ecosystems

Land surface phenology (LSP) plays a critical role in the regulation of photosynthesis, evapotranspiration, and energy fluxes. Significant progress has been made in extracting LSP information over large areas using satellite data, yet LSP retrievals remain a challenge over vast arid and semi-arid ecosystems because of sparse greenness, high variability and the lack of distinct annual patterns; for example, the MODerate Imaging Spectrometer (MODIS) Land Cover Dynamics Product MCD12Q2 that provides LSP metrics globally often failed to provide LSP information in these ecosystems. In this study, we used a modified threshold algorithm to extract LSP timing metrics, including the start, peak, and end of growing seasons, using the 16-day composite Enhanced Vegetation Index (EVI) time series from MODIS data. We applied this regionally customized algorithm across all arid and semi-arid climate regions of Australia (75% of the continental land area) encompassing shrublands, grasslands, savannas, woodlands, and croplands, extracting LSP metrics annually from 2003 to 2018, with up to two (phenology) seasons accounted for in each year. Our algorithm yielded an average of 64.9% successful rate of retrieval (proportion of pixels with retrieved LSP metrics) across 16 years in Arid and Semi-arid AUStralia (AS-AUS), which was a significant increase compared to the 14.5% rate of retrieval yielded in our study area by the global product and the major cause of the different performances between these two approaches was the different EVI amplitude restrictions utilized to avoid spurious peaks (i.e. EVI amplitude ≥ 0.1 used by the global product and peak EVI ≥ time series average EVI used by our algorithm). Gross primary productivity (GPP) measurements at OzFlux eddy covariance (EC) tower sites were used to cross-compare with the presence/absence of growing seasons detected by our algorithm, and 97% of our retrieved seasons matched with those extracted using EC data. Preliminary tests at five OzFlux sites showed that our algorithm was robust to view angle-induced sensitivity of the input data and showed similar performance when using EVI data calculated using MODIS Nadir BRDF-Adjusted Reflectance product. Our retrieved LSP metrics revealed that vegetation growth in arid ecosystems is highly irregular and can occur at any time of the year, more than once in a year, or can skip a year. The proportion of pixels with two growing seasons was found to be correlated with the average annual precipitation of the study area (p < 0.01), providing an estimation approach of LSP via rainfall. Our study improves the detection and measurement of vegetation phenology in arid and semi-arid regions by improving the spatial extend of LSP retrievals, which contributes to studies on LSP variations and dryland ecosystem resilience to climate change. More evaluation is planned for future work to assess and further improve the accuracy of the retrieved LSP metrics.

Influence of Varying Solar Zenith Angles on Land Surface Phenology Derived from Vegetation Indices: A Case Study in the Harvard Forest

Vegetation indices are widely used to derive land surface phenology (LSP). However, due to inconsistent illumination geometries, reflectance varies with solar zenith angles (SZA), which in turn affects the vegetation indices, and thus the derived LSP. To examine the SZA effect on LSP, the MODIS bidirectional reflectance distribution function (BRDF) product and a BRDF model were employed to derive LSPs under several constant SZAs (i.e., 0°, 15°, 30°, 45°, and 60°) in the Harvard Forest, Massachusetts, USA. The LSPs derived under varying SZAs from the MODIS nadir BRDF-adjusted reflectance (NBAR) and MODIS vegetation index products were used as baselines. The results show that with increasing SZA, NDVI increases but EVI decreases. The magnitude of SZA-induced NDVI/EVI changes suggests that EVI is more sensitive to varying SZAs than NDVI. NDVI and EVI are comparable in deriving the start of season (SOS), but EVI is more accurate when deriving the end of season (EOS). Specifically, NDVI/EVI-derived SOSs are relatively close to those derived from ground measurements, with an absolute mean difference of 8.01 days for NDVI-derived SOSs and 9.07 days for EVI-derived SOSs over ten years. However, a considerable lag exists for EOSs derived from vegetation indices, especially from the NDVI time series, with an absolute mean difference of 14.67 days relative to that derived from ground measurements. The SOSs derived from NDVI time series are generally earlier, while those from EVI time series are delayed. In contrast, the EOSs derived from NDVI time series are delayed; those derived from the simulated EVI time series under a fixed illumination geometry are also delayed, but those derived from the products with varying illumination geometries (i.e., MODIS NBAR product and MODIS vegetation index product) are advanced. LSPs derived from varying illumination geometries could lead to a difference spanning from a few days to a month in this case study, which highlights the importance of normalizing the illumination geometry when deriving LSP from NDVI/EVI time series.

Making Landsat 5, 7 and 8 reflectance consistent using MODIS nadir-BRDF adjusted reflectance as reference

The Landsat satellite data applications have been broadened by the improvement in data consistency, e.g., advances in calibration and atmospheric correction algorithms. However, there are still inter-sensor inconsistencies of surface reflectance caused by differences in sensor spectral response and residuals in calibration and atmospheric correction. Such inconsistencies tend to entangle with the land surface change, especially for the subtle changes such as vegetation responses to climate change. We propose a Bayesian approach to reduce inter-sensor inconsistencies by making Landsat 5, 7 and 8 reflectance consistent with MODIS nadir BRDF-adjusted reflectance (NBAR). The Bayesian MODIS-consistent Landsat reflectance should be similar to the MODIS NBAR when it is degraded to the MODIS resolution considering the point spread function (PSF). The MODIS NBAR PSF is modelled as a Gaussian function with empirically derived and band specific parameters. The relationships between the MODIS-consistent Landsat reflectance and the two available reflectance (i.e., the MODIS NBAR and observed Landsat reflectance) are built in the Bayesian framework. The Bayesian MODIS-consistent Landsat reflectance is then derived using the computationally efficient conjugate gradient descent method. The algorithm was applied to Landsat 5, 7 and 8 surface reflectance over five study areas in Brazil, China, France, Russia and USA, and compared with a linear regression-based reflectance adjustment method. Visual difference was evident in the MODIS NBAR and the observed Landsat reflectance and the relative root-mean-squared differences (rRMSD) ranged from 3.16% to 67.72% and R2 from 0.037 to 0.934. Both the Bayesian and linearly MODIS-consistent Landsat reflectance are visually more consistent with the MODIS NBAR. The average rRMSD and R2 improvements of the Bayesian MODIS-consistent Landsat reflectance over the observed Landsat reflectance were 21.36% and 0.19, 3.38% and 0.11, 5.25% and 0.07, 3.14% and 0.11, 3.49% and 0.08, and 5.72% and 0.09 for the blue, green, red, near infrared, and two shortwave infrared bands, respectively. The linearly MODIS-consistent Landsat reflectance had less rRMSD improvements: 18.26%, 1.46%, 3.02%, 1.86%, 1.97% and 3.36% for the six bands respectively, and had no R2 improvements. The linear regression derived using the whole image pixels has low performance in adjusting some extreme reflectance values that are not well represented in the image. In contrast, the Bayesian method derives the reflectance image that is consistent to MODIS at pixel level.

Spatiotemporal image fusion in Google Earth Engine for annual estimates of land surface phenology in a heterogenous landscape

Currently, quantifying phenology at landscape to regional scales is not feasible with field data or near-surface sensors. Consequently, the spatial and temporal complexity of phenology has been assessed using satellite-based estimates (land surface phenology, LSP). While estimates from Moderate Resolution Imaging Spectroradiometer (MODIS) capture intraannual patterns of phenology, they have relatively low spatial resolution. Estimates from sensors like Landsat capture finer spatial detail but are often limited by Landsat’s temporal resolution. We implemented a spatio-temporal image fusion method on the Google Earth Engine (GEE) platform and used the resulting dense time series of images to estimate intraannual LSP at 30-meter resolution. We utilized Landsat 8 surface reflectance and MODIS NBAR (Nadir BRDF-Adjusted Reflectance; MCD43A4) images from 2016 and 2017 in the interior Pacific Northwest of the United States. Images predicted from the GEE image fusion algorithm were evaluated with true Landsat observations and compared with the accuracy achieved by executing the original ESTARFM algorithm. Excluding snow and cloud obscured observations, the algorithm produced approximately 215 observations per 30-meter pixel in 2017. Root mean squared prediction error (RMSPE) of Normalized Difference Vegetation Index (NDVI) for the GEE predicted images ranged from 0.032 to 0.066, and the RMSPE for the original ESTARFM predicted images from the ranged from 0.027 to 0.064. Phenometric estimates were evaluated with near-surface sensors (PhenoCams) in shrubland, conifer, and agricultural sites and field observations of phenology in grassland, open-pine, and mixed-conifer sites. Although phenometric estimates were dissimilar at all PhenoCam sites, the general temporal pattern of the GEE image fusion and PhenoCam time series was often similar. The start of season derived from the GEE image fusion time series had closer correspondence to the PhenoCam-derived start of season at the shrubland site (13 days) than the agriculture and conifer sites. The end of season was closest at one of the conifer sites and the agriculture site (22 and 31 days, respectively). Trends of some of the field-based phenology observations aligned with phenometrics estimated from the image fusion time series. At the grassland and open-pine field sites, the phenometrics from GEE image fusion were associated with phenophase trends of dominant plant functional types. Though characterizing LSP within the interior Pacific Northwest remains a challenge, this study demonstrates that image fusion implemented in GEE can produce a densified time series capable of capturing seasonal trends in NDVI related to vegetation phenology, which can be used to estimate intraannual phenometrics.

Radiometric Normalization of Multitemporal Landsat and Sentinel-2 Images Using a Reference MODIS Product Through Spatiotemporal Filtering

Radiometric normalization is an essential preprocessing step for almost all remote sensing applications such as change detection, image mosaic, and 3-D reconstruction. This article proposes a novel radiometric normalizing method based on spatiotemporal filtering using a reference moderate resolution imaging spectroradiometer (MODIS) product. This differs from traditional relative radiometric normalization (RRN) methods in two folds: first, the number of reference images is more than one, which introduces more complexities than RRN with a single reference image; second, the resolution of MODIS product is significantly lower, thus requiring the algorithms to accommodate scale differences. To address, our approach extends the traditional spatiotemporal filtering method with per image bias that represents both internal (e.g., sensor characteristics) and external (e.g., atmosphere and topography) against the reference data. In addition, we use the Kullback-Leibler divergence metric to statistically determine the resemblance degree between the temporal images for weighting. We applied our proposed method to normalize Landsat Operational Land Imager, Enhanced Thematic Mapper Plus +, and Sentinel MSI using MODIS Nadir BRDF-adjusted reflectance product, covering two study areas of 30 × 15 km2 and 32 × 52 km2, respectively, and we show a notable radiometric consistency over both temporal and spatial dimension after the processing through three comparative experiments with state-of-the-art methods. 1) 3–7% improvement in the contexts of transfer learning, which favors only images with consistent radiometric properties and 2) Mosaic results using our processed images show no apparent seamlines as compared with images processed by other methods.

Assimilating optical satellite remote sensing images and field data to predict surface indicators in the Western U.S.: Assessing error in satellite predictions based on large geographical datasets with the use of machine learning

Indicators of vegetation composition, vegetation structure, bare ground cover, and gap size in drylands potentially gives information about the condition of ecosystems, in part because they are strongly related to factors such as erosion, wildlife habitat characteristics, and the suitability for some land uses. Field data collection based on points does not produce spatially continuous information about surface indicators and cannot cover vast geographic areas. Remote sensing is possibly a labor- and time-saving method to estimate important biophysical indicators of vegetation and surface condition at both temporal and spatial scales impossible with field methods. Regression models based on machine learning algorithms, such as random forest (RF), can build relationships between field and remotely sensed data, while also providing error estimates. In this study, field data including over 15,000 points from the Assessment, Inventory, and Monitoring (AIM) and Landscape Monitoring Framework (LMF) programs on Bureau of Land Management (BLM) lands throughout the Western U.S., Moderate Resolution Imaging Spectroradiometer (MODIS) bidirectional reflectance distribution function (BRDF) parameters, MODIS nadir BRDF-adjusted reflectance (NBAR), and Landsat 8 Operational Land Imager (OLI) surface reflectance products with ancillary data were used as predictor variables in a k-fold cross-validation approach to RF modeling. RF regression models were built to predict fourteen indicators of vegetation cover and height, as well as bare gap parameters. The RF model estimates exhibited good correlations with independent samples, with a low bias and a low RMSE. External cross-validation showed good agreement with out-of-bag (OOB) errors produced by RF and also allowed mapping prediction uncertainty. Predicted distribution maps of the surface indicators were produced by using these relationships across the arid and semiarid Western U.S. The bias and RMSE distribution maps show that the sample insufficiency and unevenly pattern of sample strongly impact the accuracy of the RF regression and prediction. The results from this study clearly show the utility of RF as a means to estimate multiple dryland surface indicators from remotely sensed data, and the reliability of the OOB errors in assessing the accuracy of the predictions.

Landsat-8 and Sentinel-2 burned area mapping - A combined sensor multi-temporal change detection approach

The free-availability of global coverage Landsat-8 and Sentinel-2 data provides the opportunity for systematic generation of medium spatial resolution land products. This paper presents a combined Landsat-8 Sentinel-2 burned area mapping algorithm. The data handling integrates recent research on Landsat-8 and Sentinel-2 pre-processing to generate registered, surface nadir BRDF-adjusted reflectance (NBAR) sensor time series that are used as an input. The different sensor data are combined through a random forest change regression, trained with synthetic data built from laboratory and field spectra and using a spectral model of fire effects on reflectance. The random forest regression is applied independently at each gridded 30 m pixel location on a temporally rolling basis considering three months of sensor data to map the central month. Temporal consistency checks are used to reduce commission errors due to non-fire related spectral changes, and a region growing algorithm is used to reduce omission errors due to temporally sparse observations. In the resulting product, each 30 m pixel is labelled as burned, unburned or unmapped. At burned pixels the estimated day of burning, and a single value that provides an estimate of the product of the subpixel fraction burned (f) and the combustion completeness (cc), henceforth termed “f.cc”, and an associated quality measure, are defined. The algorithm is demonstrated using six months of every available Landsat-8 and Sentinel-2A acquisition over 10° × 10° of Southern Africa. Experiments comparing the mapped burned areas considering Sentinel-2A only and both Landsat-8 and Sentinel-2A data indicate a greater area burned and a smaller number of unmapped pixels when both sensors are used. The results are compared with contemporaneous NASA MODIS fire products to gain insights into their temporal and spatial reporting differences. Temporally, the Sentinel-2A and Landsat-8 30 m product reports the day of burning on average three days later than the MODIS 500 m burned area product, because of the lower revisit frequency of the Sentinel-2A and Landsat-8 observations. Spatially, the Sentinel-2A and Landsat-8 30 m product captures more detail than the MODIS 500 m burned area product, with systematically higher burned area estimates. Despite the areal differences, the spatial pattern of the two products is similar, as reflected by the correlation (r2 ~ 0.7) and slope (>0.8) of regressions of the proportions of area burned defined in coarse resolution cells between the two products, taking into account the 3 day temporal bias. A comparison with multi-date PLANET 3 m data, shows high visual agreement in the location and approximate day of burning and higher f.cc values in the interiors of distinct burns than at the edges, and lower values in areas containing mixes of burned and unburned 3 m pixels. The Landsat-8/Sentinel-2A burned area mapping results are validated by comparison with burned areas interpreted visually from two date Sentinel-2A data, over 30 km × 30 km boxes selected by systematic sampling. The results mapped less area burned than the interpreted maps, as reflected by a 0.24 omission error and a negative relative bias (−0.19), with a small 0.06 commission error. Regression between the proportions of area burned defined in 1.5 km cells provide a high correlation (r2 = 0.89) and slope close to unity (0.82) and indicate that the Landsat-8/Sentinel-2A results identified similar burned area spatial patterns as the interpreted maps.

Fine Land-Cover Mapping in China Using Landsat Datacube and an Operational SPECLib-Based Approach

Fine resolution land cover information is a vital foundation of Earth science. In this paper, a novel SPECLib-based operational method is presented for the classification of multi-temporal Landsat imagery using reflectance spectra from the spatial-temporal spectral library (SPECLib) for 30 m land-cover mapping for the whole of China. Firstly, using the European Space Agency (ESA) Climate Change Initiative Global Land Cover (CCI_LC) product and the MODIS Version 6 Nadir bidirectional reflectance distribution function adjusted reflectance (NBAR) product (MCD43A4), a global SPECLib with a spatial resolution of 158.85 km (equivalent to 1.43° at the equator) and a temporal resolution of eight days was developed in the sinusoidal projection. Then, the Landsat datacube covering the whole of China was developed using all available observations of Landsat OLI imagery in 2015. Thirdly, the multi-temporal random forest method based on SPECLib was presented to produce an annual land-cover map with 22 land-cover types using the Landsat datacube. Finally, the annual China land-cover map was validated by two different validation systems using approximately 11,000 interpretation points. The mapping results achieved the overall accuracy of 71.3% and 80.7% and the kappa coefficient of 0.664 and 0.757 for the level-2 validation system (19 land-cover types) and the level-1 validation system (nine land-cover types), respectively. Therefore, the case study in China indicates that the proposed SPECLib method is an operational and accurate method for regional/global fine land-cover mapping at a spatial resolution of 30 m.

Calibration and validation of the Australian fractional cover product for MODIS collection 6

ABSTRACTFractional cover of photosynthetic vegetation (FPV), non-photosynthetic vegetation (FNPV) and bare soil (FBS) is an important input for assessment of the productivity of global pastures and rangelands. Here we describe the updating of this product using the new Moderate Resolution Imaging Spectroradiometer (MODIS) Nadir BRDF-Adjusted Reflectance (NBAR) Collection 6 reflectance product (C6) and a major expansion of the field calibration database. Fractional cover based on the C6 input exhibited reduced bias and root mean square error compared with the Collection 5 (C5) product. The expanded calibration database with more sites in arid areas provided greater separation between reflectance values of end-members for FNPV and FBS. Specific site variations in FNPV and FBS in arid areas could be traced to small but consistent changes in blue and green band, and occasional changes in short wave infrared reflectance in C6 when compared to C5. The recalibration described here provides an Australian fractional cover product with reduced uncertainty and improves the basis for a prototype global product for use in modelling of rangeland and pasture productivity.

Rainfall Variability, Wetland Persistence, and Water–Carbon Cycle Coupling in the Upper Zambezi River Basin in Southern Africa

The Upper Zambezi River Basin (UZRB) delineates a complex region of topographic, soil and rainfall gradients between the Congo rainforest and the Kalahari Desert. Satellite imagery shows permanent wetlands in low-lying convergence zones where surface&ndash;groundwater interactions are vigorous. A dynamic wetland classification based on MODIS Nadir BRDF-Adjusted Reflectance is developed to capture the inter-annual and seasonal changes in areal extent due to groundwater redistribution and rainfall variability. Simulations of the coupled water&ndash;carbon cycles of seasonal wetlands show nearly double rates of carbon uptake as compared to dry areas, at increasingly lower water-use efficiencies as the dry season progresses. Thus, wetland extent and persistence into the dry season is key to the UZRB&rsquo;s carbon sink and water budget. Whereas groundwater recharge governs the expansion of wetlands in the rainy season under large-scale forcing, wetland persistence in April&ndash;June (wet&ndash;dry transition months) is tied to daily morning fog and clouds, and by afternoon land&ndash;atmosphere interactions (isolated convection). Rainfall suppression in July&ndash;September results from colder temperatures, weaker regional circulations, and reduced instability in the lower troposphere, shutting off moisture recycling in the dry season despite high evapotranspiration rates. The co-organization of precipitation and wetlands reflects land&ndash;atmosphere interactions that determine wetland seasonal persistence, and the coupled water and carbon cycles.

Characterization of Sentinel-2A and Landsat-8 top of atmosphere, surface, and nadir BRDF adjusted reflectance and NDVI differences

The medium spatial resolution satellite data from the polar-orbiting Sentinel-2A Multi Spectral Instrument (MSI) and Landsat-8 Operational Land Imager (OLI) sensors provide 10 m to 30 m multi-spectral global coverage with a better than 5-day revisit. There are a number of differences between the sensor data that need to be considered before the data can be used together reliably. Sentinel-2A and Landsat-8 data for approximately 10° × 10° of southern Africa acquired in two summer (December and January) and in two winter (June and July) months of 2016 were compared. The data were registered and each orbit projected into 30 m fixed non-overlapping tiles defined in the sinusoidal equal area projection. Only corresponding sensor observations of each 30 m tile pixel that were both not cloudy, shadow, saturated, or cirrus contaminated, and that were acquired within one-day of each other, were compared. Both the Sentinel-2A MSI and Landsat-8 OLI data were atmospherically corrected using the Land Surface Reflectance Code (LaSRC) and were also corrected to nadir BRDF adjusted reflectance (NBAR). Top of atmosphere (TOA), surface reflectance, and NBAR, for the spectrally corresponding visible, near infrared (NIR) and shortwave infrared (SWIR) MSI and OLI bands, and derived normalized difference vegetation index (NDVI) (from the narrow NIR band for MSI), were compared and their sensor differences quantified by regression analyses. Atmospheric contamination and bi-directional reflectance differences were evident in the 65 million pairs of contemporaneous MSI and OLI observations considered. The MSI surface reflectance was greater than the OLI surface reflectance for all the bands except the green, red, and the broad MSI NIR bands, and the MSI surface NDVI was greater than the OLI surface NDVI. This pattern was also found in the NBAR sensor comparisons except for the red bands. Simulated MSI and OLI reflectance derived using the sensor spectral response functions and laboratory spectra showed similar results in the red, NIR and SWIR bands as the real data comparisons. Ordinary least squares (OLS) linear regressions of the 65 million pairs of contemporaneous MSI and OLI data for the three processing levels had good fits (r2 > 0.87 for the TOA data comparisons, r2 > 0.89 for the atmospherically corrected data comparisons, r2 > 0.90 for the NBAR data comparisons; p-values < 0.0001). The OLS regression coefficients are provided so that they can be used to help improve the consistency between Sentinel-2A MSI and Landsat-8 OLI data.

Improvement of Countrywide Vegetation Mapping over Japan and Comparison to Existing Maps

This paper presents an improved classification and mapping of vegetation types for all of Japan by utilizing the Moderate-resolution Imaging Spectroradiometer (MODIS) data. The Nadir BRDF-Adjusted Reflectance (MCD43A4 product) data were compared to the conventional Surface Reflectance (MOD09A1/MOY09A1 products) data for the classification of vegetation types: evergreen coniferous forest, evergreen broadleaf forest, deciduous coniferous forest, deciduous broadleaf forest, shrubs, herbaceous, arable; and non-vegetation. Very rich spectral and temporal features were prepared from MCD43A4 and MOD09A1/MOY09A1 products. Random Forests classifier was employed for the classification of vegetation types with the support of ground truth data prepared in the research. Accuracy metrics—confusion matrix, overall accuracy, and kappa coefficient calculated through 10-fold cross-validation approach—were used for quantitative comparison of MCD43A4 and MOD09A1/MOY09A1 products. The cross-validation results indicated better performance of the MCD43A4 (Overall accuracy = 0.73; Kappa coefficient = 0.69) product than conventional MOD09A1/MOY09A1 products (Overall accuracy = 0.70; Kappa coefficient = 0.66) for the classification. McNemar’s test was also used to confirm a significant difference (p-value = 0.0003) between MCD43A4 and MOD09A1/MOY09A1 products. Based on these results, by utilizing the MCD43A4 features, a new vegetation map was produced for all of Japan. The newly produced map showed better accuracy than the extant, MODIS Land Cover Type product (MCD12Q1) and Global Land Cover by National Mapping Organizations (GLCNMO) product in Japan.

Adjustment of Sentinel-2 Multi-Spectral Instrument (MSI) Red-Edge Band Reflectance to Nadir BRDF Adjusted Reflectance (NBAR) and Quantification of Red-Edge Band BRDF Effects

Optical wavelength satellite data have directional reflectance effects over non-Lambertian surfaces, described by the bidirectional reflectance distribution function (BRDF). The Sentinel-2 multi-spectral instrument (MSI) acquires data over a 20.6° field of view that have been shown to have non-negligible BRDF effects in the visible, near-infrared, and short wave infrared bands. MSI red-edge BRDF effects have not been investigated. In this study, they are quantified by an examination of 6.6 million (January 2016) and 10.7 million (April 2016) pairs of forward and back scatter reflectance observations extracted over approximately 20° × 10° of southern Africa. Non-negligible MSI red-edge BRDF effects up to 0.08 (reflectance units) across the 290 km wide MSI swath are documented. A recently published MODIS BRDF parameter c-factor approach to adjust MSI visible, near-infrared, and short wave infrared reflectance to nadir BRDF-adjusted reflectance (NBAR) is adapted for application to the MSI red-edge bands. The red-edge band BRDF parameters needed to implement the algorithm are provided. The parameters are derived by a linear wavelength interpolation of fixed global MODIS red and NIR BRDF model parameters. The efficacy of the interpolation is investigated using POLDER red, red-edge, and NIR BRDF model parameters, and is shown to be appropriate for the c-factor NBAR generation approach. After adjustment to NBAR, red-edge MSI BRDF effects were reduced for the January data (acquired close to the solar principal where BRDF effects are maximal) and the April data (acquired close to the orthogonal plane) for all the MSI red-edge bands.

Evaluation of the VIIRS BRDF, Albedo and NBAR products suite and an assessment of continuity with the long term MODIS record

Bidirectional Reflectance Distribution Function (BRDF) model parameters, Albedo quantities, and Nadir BRDF Adjusted Reflectance (NBAR) products derived from the Visible Infrared Imaging Radiometer Suite (VIIRS), on the Suomi-NPP (National Polar-orbiting Partnership) satellite are evaluated with spatially representative in situ tower albedometer measurements and through comparison with the MODerate Resolution Imaging Spectroradiometer (MODIS) long term record. White Sky Albedo (WSA), Black Sky Albedo (BSA), NBAR, and Quality Assurance (QA) results show that the VIIRS algorithm is performing well at the global scale and is able to provide global data products comparable with the heritage MODIS (with an absolute bias of 0.0069 for shortwave broadband WSA) despite the spectral, angular, and effective spatial resolution differences between the two sensors. Both VIIRS and MODIS albedo are shown to agree well with in situ albedo measurements at a number of spatially representative sites. This evaluation provides confidence that the high quality, daily VIIRS BRDF model parameters, Albedo, and NBAR products will be able to provide the long term continuity required by the modeling and monitoring communities.

Examination of Sentinel-2A multi-spectral instrument (MSI) reflectance anisotropy and the suitability of a general method to normalize MSI reflectance to nadir BRDF adjusted reflectance

The Sentinel-2A multi-spectral instrument (MSI) acquires multi-spectral reflective wavelength observations with directional effects due to surface reflectance anisotropy and changes in the solar and viewing geometry. Directional effects were examined by considering two ten day periods of Sentinel-2A data acquired close to the solar principal and orthogonal planes over approximately 20°×10° of southern Africa. More than 6.6 million (January 2016) and 10.6 million (April 2016) pairs of reflectance observations sensed 3 or 7days apart in the forward and backscatter directions in overlapping Sentinel-2A orbit swaths were considered. The Sentinel-2A data were projected into the MODIS sinusoidal projection but first had to be registered due to a misregistration issue evident in the overlapping orbits. The top of atmosphere reflectance data were corrected to surface reflectance using the SEN2COR atmospheric correction software. Only pairs of forward and backward reflectance values that were cloud and snow-free, unsaturated, and had no significant change in their 3 or 7day separation, were considered. The maximum observed Sentinel-2A view zenith angle was 11.93°. Greater BRDF effects were apparent in the January data (acquired close to the solar principal plane) than the April data (acquired close to the orthogonal plane) and at higher view zenith angle. For the January data the average difference between the surface reflectance in the forward and backward scatter directions at the Sentinel-2A scan edges increased with wavelength from 0.035 (blue), 0.047 (green), 0.057 (red), 0.078 (NIR), to about 0.1 (SWIR). These differences may constitute a significant source of noise for certain applications.The suitability of a recently published methodology developed to generate Landsat nadir BRDF-adjusted reflectance (NBAR) was examined for Sentinel-2A application. The methodology uses fixed MODIS BRDF spectral parameters and is attractive because it has little sensitivity to the land cover type, condition, or surface disturbance and can be derived in a computationally efficient manner globally. It was applied to the southern Africa Sentinel-2A data and shown to reduce Sentinel-2A BRDF effects. The average difference between the reflectance in the forward and backward scatter directions at the Sentinel-2A scan edges was smaller in the NBAR data than in the corresponding surface reflectance data. Residual BRDF effects in the Sentinel-2A NBAR data occurred likely because of atmospheric correction and sensor calibration errors and inadequacies in the NBAR derivation approach. These issues are discussed with recommendations for future research including global and red-edge Sentinel-2A NBAR derivation that were not considered in this study.