Search

Cloud is a serious problem that affects the quality of remote-sensing (RS) images. Existing cloud removal techniques suffer from notable limitations, such as being specific to certain data types, cloud conditions, and spatial extents, as well as requiring auxiliary data, which hampers their generalizability and flexibility. To address the issue, we propose a maximum-value compositing approach by generating cloud masks. We acquired 432 daily MOD09GA L2 MODIS imageries covering a vast region with persistent cloud cover and various climates and land-cover types. Labeled datasets for cloud, land, and no-data were collected from selected daily imageries. Subsequently, we trained and evaluated RF, SVM, and U-Net models to choose the best models. Accordingly, SVM and U-Net were chosen and employed to classify all the daily imageries. Then, the classified imageries were converted to two sets of mask layers to mask clouds and no-data pixels in the corresponding daily images by setting the masked pixels' values to −0.999999. After masking, we employed the maximum-value technique to generate two sets of 16-day composite products, MaxComp-1 and MaxComp-2, corresponding to SVM and U-Net-derived cloud masks, respectively. Finally, we assessed the quality of our composite products by comparing them with the reference MOD13A1 16-day composite product. Based on the land-cover classification accuracy, our products yielded a significantly higher accuracy (5–28%) than the reference MODIS product across three classifiers (RF, SVM, and U-Net), indicating the quality of our products and the effectiveness of our techniques. In particular, MaxComp-1 yielded the best results, which further implies the superiority of SVM for cloud masking. In addition, our products appear to be more radiometrically and spectrally consistent and less noisy than MOD13A1, implying that our approach is more efficient in removing shadows and noises/artifacts. Our method yields high-quality products that are vital for investigating large regions with persistent clouds and studies requiring time-series data. Moreover, the proposed techniques can be adopted for higher-resolution RS imageries, regardless of the spatial extent, data volume, and type of clouds. [ABSTRACT FROM AUTHOR]

The use of satellite remote sensing has considerably improved scientific understanding of the heterogeneity of Amazonian rainforests. However, the persistent cloud cover and strong Bidirectional Reflectance Distribution Function (BRDF) effects make it difficult to produce up-to-date satellite image composites over the huge extent of Amazonia. Advanced pre-processing and pixel-based compositing over an extended time period are needed to fill the data gaps caused by clouds and to achieve consistency in pixel values across space. Recent studies have found that the multidimensional median, also known as medoid, algorithm is robust to outliers and noise, and thereby provides a useful approach for pixel-based compositing. Here we describe Landsat-7 and Landsat-8 composites covering all Amazonia that were produced using Landsat data from the years 2013–2021 and processed with Google Earth Engine (GEE). These products aggregate reflectance values over a relatively long time, and are, therefore, especially useful for identifying permanent characteristics of the landscape, such as vegetation heterogeneity that is driven by differences in geologically defined edaphic conditions. To make similar compositing possible over other areas and time periods (including shorter time periods for change detection), we make the workflow available in GEE. Visual inspection and comparison with other Landsat products confirmed that the pre-processing workflow was efficient and the composites are seamless and without data gaps, although some artifacts present in the source data remain. Basin-wide Landsat-7 and Landsat-8 composites are expected to facilitate both local and broad-scale ecological and biogeographical studies, species distribution modeling, and conservation planning in Amazonia.

Cloud is a serious problem that affects the quality of remote-sensing (RS) images. Existing cloud removal techniques suffer from notable limitations, such as being specific to certain data types, cloud conditions, and spatial extents, as well as requiring auxiliary data, which hampers their generalizability and flexibility. To address the issue, we propose a maximum-value compositing approach by generating cloud masks. We acquired 432 daily MOD09GA L2 MODIS imageries covering a vast region with persistent cloud cover and various climates and land-cover types. Labeled datasets for cloud, land, and no-data were collected from selected daily imageries. Subsequently, we trained and evaluated RF, SVM, and U-Net models to choose the best models. Accordingly, SVM and U-Net were chosen and employed to classify all the daily imageries. Then, the classified imageries were converted to two sets of mask layers to mask clouds and no-data pixels in the corresponding daily images by setting the masked pixels’ values to −0.999999. After masking, we employed the maximum-value technique to generate two sets of 16-day composite products, MaxComp-1 and MaxComp-2, corresponding to SVM and U-Net-derived cloud masks, respectively. Finally, we assessed the quality of our composite products by comparing them with the reference MOD13A1 16-day composite product. Based on the land-cover classification accuracy, our products yielded a significantly higher accuracy (5–28%) than the reference MODIS product across three classifiers (RF, SVM, and U-Net), indicating the quality of our products and the effectiveness of our techniques. In particular, MaxComp-1 yielded the best results, which further implies the superiority of SVM for cloud masking. In addition, our products appear to be more radiometrically and spectrally consistent and less noisy than MOD13A1, implying that our approach is more efficient in removing shadows and noises/artifacts. Our method yields high-quality products that are vital for investigating large regions with persistent clouds and studies requiring time-series data. Moreover, the proposed techniques can be adopted for higher-resolution RS imageries, regardless of the spatial extent, data volume, and type of clouds.

The use of satellite remote sensing has considerably improved scientific understanding of the heterogeneity of Amazonian rainforests. However, the persistent cloud cover and strong Bidirectional Reflectance Distribution Function (BRDF) effects make it difficult to produce up-to-date satellite image composites over the huge extent of Amazonia. Advanced pre-processing and pixel-based compositing over an extended time period are needed to fill the data gaps caused by clouds and to achieve consistency in pixel values across space. Recent studies have found that the multidimensional median, also known as medoid, algorithm is robust to outliers and noise, and thereby provides a useful approach for pixel-based compositing. Here we describe Landsat-7 and Landsat-8 composites covering all Amazonia that were produced using Landsat data from the years 2013-2021 and processed with Google Earth Engine (GEE). These products aggregate reflectance values over a relatively long time, and are, therefore, especially useful for identifying permanent characteristics of the landscape, such as vegetation heterogeneity that is driven by differences in geologically defined edaphic conditions. To make similar compositing possible over other areas and time periods (including shorter time periods for change detection), we make the workflow available in GEE. Visual inspection and comparison with other Landsat products confirmed that the pre-processing workflow was efficient and the composites are seamless and without data gaps, although some artifacts present in the source data remain. Basin-wide Landsat-7 and Landsat-8 composites are expected to facilitate both local and broad-scale ecological and biogeographical studies, species distribution modeling, and conservation planning in Amazonia., (© 2024 The Authors.)

Urbanization and the associated change in land cover has been intensifying across the globe in recent decades. Regional studies on the rate and amount of urban expansion are critical for understanding how patterns of change differ within and among cities with varying structure and development characteristics. Yet spatially consistent and timely information on urban development is difficult to access particularly across international jurisdictions. Remote sensing based technologies offer a unique perspective on urban land cover with the data offering significant potential to urban studies due to its consistent and ubiquitous nature. In this research we applied a pixel-based image composite technique to generate annual gap-free surface reflectance Landsat composites from 1984 to 2012 for 25 urban environments across 12 countries in the Pacific Rim. Using time series composites, spectral indices were calculated and compared using a hexagonal grid ring model to assess changes in vegetative and urban patterns. Trajectories were then clustered to further investigate the spatio-temporal dynamics and relationships among the 25 cities. Performance of the clustering analyses varied depended on the temporal and spatial metrics however overall clustering results indicated relatively strong spatio-temporal similarities among a number of key cities. Three pairs of cities—Melbourne and Sydney; Tianjin and Manila; and Singapore City and Kuala Lumpur were found to be highly similar in their urban and vegetation dynamics temporally and spatially. In contrast Vancouver and Las Vegas had no similar analogous. This work demonstrates the value of utilising annual Landsat time series composites for assessing urban vegetation and urban dynamics at regional scales and potential use in sustainable urban planning, resources allocation, and policy making. [ABSTRACT FROM AUTHOR]

Deriving burn severity from multispectral satellite data is a widely adopted approach to infer the degree of environmental change caused by fire. Burn severity maps obtained by thresholding bi-temporal indices based on pre- and post-fire Normalized Burn Ratio (NBR) can vary substantially depending on temporal constraints such as matched acquisition and optimal seasonal timing. Satisfying temporal requirements is crucial to effectively disentangle fire and non-fire induced spectral changes and can be particularly challenging when only a few cloud-free images are available. Our study focuses on 10 wildfires that occurred in mountainous areas of the Piedmont Region (Italy) during autumn 2017 following a severe and prolonged drought period. Our objectives were to: (i) generate reflectance composites using Sentinel-2 imagery that were optimised for seasonal timing by embedding spatial patterns of long-term land surface phenology (LSP); (ii) produce and validate burn severity maps based on the modelled relationship between bi-temporal indices and field data; (iii) compare burn severity maps obtained using either a pair of cloud-free Sentinel-2 images, i.e. paired images, or reflectance composites. We proposed a pixel-based compositing algorithm coupling the weighted geometric median and thematic spatial information, e.g. long-term LSP metrics derived from the MODIS Collection 6 Land Cover Dynamics Product, to rank all the clear observations available in the growing season. Composite Burn Index data and bi-temporal indices exhibited a strong nonlinear relationship (R 2 > 0.85) using paired images or reflectance composites. Burn severity maps attained overall classification accuracy ranging from 76.9% to 83.7% (Kappa between 0.61 and 0.72) and the Relative differenced NBR (RdNBR) achieved the best results compared to other bi-temporal indices (differenced NBR and Relativized Burn Ratio). Improvements in overall classification accuracy offered by the calibration of bi-temporal indices with the dNBR offset were limited to burn severity maps derived from paired images. Reflectance composites provided the highest overall classification accuracy and differences with paired images were significant using uncalibrated bi-temporal indices (4.4% to 5.2%) while they decreased (2.8% to 3.2%) when we calibrated bi-temporal indices derived from paired images. The extent of the high severity category increased by ~19% in burn severity maps derived from reflectance composites (uncalibrated RdNBR) compared to those from paired images (calibrated RdNBR). The reduced contrast between healthy and burnt conditions associated with suboptimal seasonal timing caused an underestimation of burnt areas. By embedding spatial patterns of long-term LSP metrics, our approach provided consistent reflectance composites targeted at a specific phenological stage and minimising non-fire induced inter-annual changes. Being independent from the multispectral dataset employed, the proposed pixel-based compositing approach offers new opportunities for operational change detection applications in geographic areas characterised by persistent cloud cover. • We mapped burn severity using Sentinel-2 paired images and reflectance composites. • Bi-temporal indices benefit from a reduction in non-fire induced spectral changes. • Reflectance composites exhibit spectral consistency in space and time. • Our approach increased the overall classification accuracy up to 5.2%. • Gaining contrast between healthy and burnt vegetation improves burn severity maps. [ABSTRACT FROM AUTHOR]

Remote sensing assessments of land use and land cover change (LULCC) are critical to improve understanding of socio-economic, institutional and ecological processes that lead to and stem from land use change. This is particularly crucial in the emerging frontiers of Southern Africa, where there is a paucity of LULCC studies relative to the humid tropics. This study focuses on Gurué District (5606 km2) of Zambezia province of Mozambique, one of many countries in the region that has experienced a recent growth in foreign investments in agriculture through large-scale land acquisitions, often resulting in land use conversions and modifications. Previous LULCC assessments covering Mozambique have focused on dynamics between natural and anthropogenic land categories, with limited efforts to distinguish the different land use agents associated with these changes, and relating this with social, economic and technological processes. In this study we built a new LULC assessment methodology that leverages the power of open remote sensing data and tools to integrate categorical and continuous training and validation data obtained from field surveys and Collect Earth software within Google Earth Engine. We then examined the suitability of five pixel-based compositing techniques for generating cloud-free Landsat images that can support analysis of land use dynamics in persistently cloudy, mosaic landscapes with more limited Landsat archives. Drawing upon the spectral and textural features of Landsat data in pixel-based composites, we classified land use over three time periods, 2006, 2012 and 2016, and characterized land use change, focusing on changes between small-scale cropland, large-scale mechanized cropland, and other land uses. This method can be upscaled and applied in many parts of Africa with similar historic image availability challenges, and similar economic contexts with great disparities between small-scale unmechanized cropland and very large-scale mechanized cropland, to explore land consolidation dynamics and agent-specific pathways of land use change. • Pixel-based compositing impacts classification results. • Median compositing performed best for disaggregating cropland by field size. • Collect Earth data can be used to train and test classification algorithms. • Textural features facilitate the separation of small- and large-scale cropland. • Mapping cropland scale dynamics can improve understanding of land and food systems. [ABSTRACT FROM AUTHOR]