High-resolution Land Cover Map Research Articles

Assessment of Implementing Land Use/Land Cover LULC 2020-ESRI Global Maps in 2D Flood Modeling Application

Floods are one of the most dangerous water-related risks. Numerous sources of uncertainty affect flood modeling. High-resolution land-cover maps along with appropriate Manning’s roughness values are the most significant parameters for building an accurate 2D flood model. Two land-cover datasets are available: the National Land Cover Database (NLCD 2019) and the Land Use/Land Cover for Environmental Systems Research Institute (LULC 2020-ESRI). The NLCD 2019 dataset has national coverage but includes references to Manning’s roughness values for each class obtained from earlier studies, in contrast to the LULC 2020-ESRI dataset, which has global coverage but without an identified reference to Manning’s roughness values yet. The main objectives of this study are to assess the accuracy of using the LULC 2020-ESRI dataset compared with the NLCD 2019 dataset and propose a standard reference to Manning’s roughness values for the classes in the LULC 2020-ESRI dataset. To achieve the research objectives, a confusion matrix using 548,117 test points in the conterminous United States was prepared to assess the accuracy by quantifying the cross-correspondence between the two datasets. Then statistical analyses were applied to the global maps to detect the appropriate Manning’s roughness values associated with the LULC 2020-ESRI map. Compared to the NLCD 2019 dataset, the proposed Manning’s roughness values for the LULC 2020-ESRI dataset were calibrated and validated using 2D flood modeling software (HEC-RAS V6.2) on nine randomly chosen catchments in the conterminous United States. This research’s main results show that the LULC 2020-ESRI dataset achieves an overall accuracy of 72% compared to the NLCD 2019 dataset. The findings demonstrate that, when determining the appropriate Manning’s roughness values for the LULC 2020-ESRI dataset, the weighted average technique performs better than the average method. The calibration and validation results of the proposed Manning’s roughness values show that the overall Root Mean Square Error (RMSE) in depth was 2.7 cm, and the Mean Absolute Error (MAE) in depth was 5.32 cm. The accuracy of the computed peak flow value using LULC 2020-ESRI was with an average error of 5.22% (2.0% min. to 8.8% max.) compared to the computed peak flow values using the NLCD 2019 dataset. Finally, a reference to Manning’s roughness values for the LULC 2020-ESRI dataset was developed to help use the globally available land-use/land-cover dataset to build 2D flood models with an acceptable accuracy worldwide.

High resolution modeling of vegetation reveals large summertime biogenic CO2 fluxes in New York City

Abstract Cities are beginning to monitor atmospheric carbon dioxide (CO2) to assess the efficacy of their climate policies. However, changes in anthropogenic CO2 emissions must be separated from biospheric CO2 fluxes which have a large seasonal cycle. Urban vegetation (e.g. lawns, trees along street and in parks, etc) in developed land covers is often omitted in regional biogenic CO2 flux models. We set up a biosphere model to estimate the regional biogenic CO2 fluxes in New York City (NYC) and assess the importance of vegetation within developed land covers. The model incorporates a high-resolution (30 m) land cover map which identifies the mixture of impervious surfaces and vegetation that is ubiquitous across developed land covers. We designed three model scenarios to evaluate the role of developed land covers in regional biogenic CO2 fluxes by assuming (a) there is no vegetation versus scenarios where all remotely sensed vegetation in developed land covers is either (b) grassland or (c) deciduous forest. Despite relatively low tree canopy cover in NYC, the regional biogenic CO2 fluxes are surprisingly large when vegetation within the developed land covers is included. Furthermore, the types of vegetation within the developed land covers are crucially important for estimating regional biogenic CO2 fluxes, demonstrated by a doubling in estimates of total biogenic CO2 flux when this vegetation is assumed to be grassland compared to forest. Using a Lagrangian atmospheric transport model, we find that the regional biogenic CO2 uptake offsets up to 40% of atmospheric CO2 enhancements attributed to anthropogenic emissions in summer afternoons and completely balances on-road traffic in one of the most congested cities in the United States. Accurate characterization of the vegetation and biogenic carbon fluxes in cities are essential to the development of effective atmospheric monitoring tools. Future measurements should focus on constraining CO2 fluxes in urban grasslands (i.e. lawns).

Improved Spatiotemporal Information Fusion Approach Based on Bayesian Decision Theory for Land Cover Classification

High-spatial-resolution (HSR) images and high-temporal-resolution (HTR) images have their unique advantages and can be replenished by each other effectively. For land cover classification, a series of spatiotemporal fusion algorithms were developed to acquire a high-resolution land cover map. The fusion processes focused on the single level, especially the pixel level, could ignore the different phenology changes and land cover changes. Based on Bayesian decision theory, this paper proposes a novel decision-level fusion for multisensor data to classify the land cover. The proposed Bayesian fusion (PBF) combines the classification accuracy of results and the class allocation uncertainty of classifiers in the estimation of conditional probability, which consider the detailed spectral information as well as the various phenology information. To deal with the scale inconsistency problem at the decision level, an object layer and an area factor are employed for unifying the spatial resolution of distinct images, which would be applied for evaluating the classification uncertainty related to the conditional probability inference. The approach was verified on two cases to obtain the HSR land cover maps, in comparison with the implementation of two single-source classification methods and the benchmark fusion methods. Analyses and comparisons of the different classification results showed that PBF outperformed the best performance. The overall accuracy of PBF for two cases rose by an average of 27.8% compared with two single-source classifications, and an average of 13.6% compared with two fusion classifications. This analysis indicated the validity of the proposed method for a large area of complex surfaces, demonstrating the high potential for land cover classification.

SEG-ESRGAN: A Multi-Task Network for Super-Resolution and Semantic Segmentation of Remote Sensing Images

The production of highly accurate land cover maps is one of the primary challenges in remote sensing, which depends on the spatial resolution of the input images. Sometimes, high-resolution imagery is not available or is too expensive to cover large areas or to perform multitemporal analysis. In this context, we propose a multi-task network to take advantage of the freely available Sentinel-2 imagery to produce a super-resolution image, with a scaling factor of 5, and the corresponding high-resolution land cover map. Our proposal, named SEG-ESRGAN, consists of two branches: the super-resolution branch, that produces Sentinel-2 multispectral images at 2 m resolution, and an encoder–decoder architecture for the semantic segmentation branch, that generates the enhanced land cover map. From the super-resolution branch, several skip connections are retrieved and concatenated with features from the different stages of the encoder part of the segmentation branch, promoting the flow of meaningful information to boost the accuracy in the segmentation task. Our model is trained with a multi-loss approach using a novel dataset to train and test the super-resolution stage, which is developed from Sentinel-2 and WorldView-2 image pairs. In addition, we generated a dataset with ground-truth labels for the segmentation task. To assess the super-resolution improvement, the PSNR, SSIM, ERGAS, and SAM metrics were considered, while to measure the classification performance, we used the IoU, confusion matrix and the F1-score. Experimental results demonstrate that the SEG-ESRGAN model outperforms different full segmentation and dual network models (U-Net, DeepLabV3+, HRNet and Dual_DeepLab), allowing the generation of high-resolution land cover maps in challenging scenarios using Sentinel-2 10 m bands.

A multi-temporal phenology based classification approach for Crop Monitoring in Kenya

The SBAM (Satellite Based Agricultural Monitoring) project, funded by the Italian Space Agency aims at: developing a validated satellite imagery based method for estimating and updating the agricultural areas in the region of Central-Africa; implementing an automated process chain capable of providing periodical agricultural land cover maps of the area of interest and, possibly, an estimate of the crop yield. The project aims at filling the gap existing in the availability of high spatial resolution maps of the agricultural areas of Kenya. A high spatial resolution land cover map of Central-Eastern Africa including Kenya was compiled in the year 2000 in the framework of the Africover project using Landsat images acquired, mostly, in 1995. We investigated the use of phenological information in supporting the use of remotely sensed images for crop classification and monitoring based on Landsat 8 and, in the near future, Sentinel 2 imagery. Phenological information on crop condition was collected using time series of NDVI (Normalized Difference Vegetation Index) based on Landsat 8 images. Kenyan countryside is mainly characterized by a high number of fragmented small and medium size farmlands that dramatically increase the difficulty in classification; 30 m spatial resolution images are not enough for a proper classification of such areas. So, a pan-sharpening FIHS (Fast Intensity Hue Saturation) technique was implemented to increase image resolution from 30 m to 15 m. Ground test sites were selected, searching for agricultural vegetated areas from which phenological information was extracted. Therefore, the classification of agricultural areas is based on crop phenology, vegetation index behaviour retrieved from a time series of satellite images and on AEZ (Agro Ecological Zones) information made available by FAO (FAO, 1996) for the area of interest. This paper presents the results of the proposed classification procedure in comparison with land cover maps produced in the past years by other projects. The results refer to the Nakuru County and they were validated using field campaigns data. It showed a satisfactory overall accuracy of 92.66 % which is a significant improvement with respect to previous land cover maps.

Relationships between urbanization, tree morphology, and carbon density: An integration of remote sensing, allometric models, and field survey

Urban trees store and sequester large amounts of carbon and are a vital component of natural climate solutions. Despite the well-recognized carbon benefits of urban trees, there is limited effort to examine how spatial distribution of carbon density varies across distinctive social, demographic, and built dimensions of urban landscapes. Moreover, it is unclear whether specific aspects of landscape structure and design could help increase carbon densities in urban trees. Here, we produced a fine-resolution carbon density map of urban trees in New York City (NYC) by integrating high-resolution land cover map, LiDAR-derived tree metrics, i-Tree Eco, and field survey data. We then explored spatial variations of carbon density across the gradients of urban development intensity, social deprivation index, and neighborhood age, and we examined the relationships between carbon density, and fragmentation, aggregation, size, and shape of tree canopy cover. We find that carbon stored in urban trees in NYC is estimated as 1078 Gg, with an average density of 13.8 Mg/ha. This large amount of carbon is unevenly distributed, with carbon densities being highest in Bronx and in open parks and street trees. Furthermore, carbon densities are negatively associated with urban development intensity and the social gradient of deprivation. Regarding the impacts of tree morphology on carbon density, our results show that while the amount of tree cover is the most influential factor in determining carbon density, small-sized forest patches and moderate levels of forest edges are also conductive to increasing carbon densities of urban trees. To incorporate urban forestry into developing innovative, effective, and equitable climate mitigation strategies, planners and decision makers need to identify the optimal spatial configuration of urban forests and invest in tree planting programs in marginalized communities.

Breaking the resolution barrier: A low-to-high network for large-scale high-resolution land-cover mapping using low-resolution labels

Large-scale high-resolution land-cover mapping is a way to comprehend the Earth’s surface and resolve the ecological and resource challenges facing humanity. High-resolution (≤1 m) remotely sensed images can now be captured more easily, with wider coverage, as sensors and satellites develop. Nevertheless, the synchronous renewal of land-cover maps is still challenging when using the common land-cover mapping methods, due to the requirement for high-resolution land-cover labels. Abundant low-resolution (∼30 m) land-cover products are available for use as alternative label sources, but the resolution gap between these products and the growing volume of high-resolution imagery is a barrier yet to be overcome. In this paper, to break through this obstacle, we propose a low-to-high network (L2HNet) to automatically generate high-resolution land-cover maps from high-resolution images by taking only low-resolution land-cover products as the training labels, thus getting rid of the requirement for finely labeled samples during the large-scale map updating process. Firstly, to obtain the mapping results with rich details, we propose a resolution-preserving (RP) backbone that contains parallel multi-scale convolutional layers for extracting the high-resolution features from the images. Furthermore, to settle the label noise issue caused by the mismatched resolution, a confident area selection (CAS) module and a low-to-high (L2H) loss function, with weak and unsupervised strategies, are designed for obtaining reliable supervision information from the coarse labels. The experimental results obtained for six administrative states located in the Chesapeake Bay watershed of the United States show that L2HNet outperforms several of the state-of-the-art methods and the mainstream land-cover mapping methods in creating 1-m resolution land-cover maps by taking 30-m resolution land-cover products as training labels. As a further application, L2HNet was also adopted to produce the first 1-m resolution land-cover map with level II classification hierarchy for the entire state of Maryland in the United States, which covers an area of about 33,872 km2. The land-cover map of Maryland is publicly available at http://hipag.whu.edu.cn/L2HNet.html.

Weakly supervised high spatial resolution land cover mapping based on self-training with weighted pseudo-labels

Despite its success, deep learning in land cover mapping requires a massive amount of pixel-wise labeled images. It typically assumes that the training and test scenes are similar in data distribution. The performance of models trained on any particular dataset could degrade significantly on a new dataset due to the domain shift or domain gap across datasets, resulting in new training data requiring labor-intensive manual pixel-wise labeling. This paper proposes a land cover mapping framework combining Feature Pyramid Network (FPN) and self-training. In the FPN, we integrate ConvNeXt with a Pyramid Pooling Module (PPM). Combining the FPN and the PPM improves the segmentation performance, which benefits from the multiscale aggregation of pyramid features. To fully exploit pseudo-labels, we design an Unsupervised Domain Adaptation (UDA) land cover mapping scheme with self-training using weighted pseudo-labels of the target samples. The proposed land cover mapping framework could benefit from multiscale aggregation of pyramid features and the full use of the pseudo-labels. Comparison results on the LoveDA dataset, the latest large-scale unsupervised domain adaptation dataset for land cover mapping, empirically demonstrated that our land cover mapping approach significantly outperforms the baselines in both UDA scenarios, i.e., Urban → Rural and Rural → Urban . The models of this paper are now publicly available on GitHub. 1 1 https://github.com/csliujw/uda-self-training . • A land cover mapping network based on an FPN using ConvNeXt with a PPM to improve the land cover mapping performance. • A UDA scheme with self-training using weighted pseudo-labels. • Our scheme is significantly superior to the baselines on both UDA scenarios: Urban → Rural and Rural → Urban.

High Resolution Land Cover Mapping and Crop Classification in the Loukkos Watershed (Northern Morocco): An Approach Using SAR Sentinel-1 Time Series

Remote sensing has become more and more a reliable tool for mapping land cover and monitoring cropland. Much of the work done in this field uses optical remote sensing data. In Morocco, active remote sensing data remain under-exploited despite their importance in monitoring spatial and temporal dynamics of land cover and crops even during cloudy weather. This study aims to explore the potential of C-band Sentinel-1 data in the production of a high-resolution land cover mapping and crop classification within the irrigated Loukkos watershed agricultural landscape in northern Morocco. The work was achieved by using 33 dual-polarized images in vertical-vertical (VV) and vertical-horizontal (VH) polarizations. The images were acquired in ascending orbits between April 16 and October 25, 2020, with the purpose to track the backscattering behavior of the main crops and other land cover classes in the study area. The results showed that the backscatter increased with the phenological development of the monitored crops (rice, watermelon, peanuts, and winter crops), strongly for the VH and VV bands, and slightly for the VH/VV ratio. The other classes (water, built-up, forest, fruit trees, permanent vegetation, greenhouses, and bare lands) did not show significant variation during this period. Based on the backscattering analysis and the field data, a supervised classification was carried out, using the Random Forest Classifier (RF) algorithm. Results showed that radiometric characteristics and 6 days’ time resolution covered by Sentinel-1 constellation gave a high classification accuracy by dual-polarization with Radar Ratio (VH/VV) or Radar Vegetation Index and textural features (between 74.07% and 75.19%). Accordingly, this study proves that the Sentinel-1 data provide useful information and a high potential for multi-temporal analyses of crop monitoring, and reliable land cover mapping which could be a practical source of information for various purposes in order to undertake food security issues.

The influence of early-life residential exposure to different vegetation types and paved surfaces on early childhood development: A population-based birth cohort study

BackgroundGrowing evidence suggests that exposure to green space is associated with improved childhood health and development, but the influence of different green space types remains relatively unexplored. In the present study, we investigated the association between early-life residential exposure to vegetation and early childhood development and evaluated whether associations differed according to land cover types, including paved land. MethodsEarly childhood development was assessed via kindergarten teacher-ratings on the Early Development Instrument (EDI) in a large population-based birth cohort (n = 27,539) in Metro Vancouver, Canada. The residential surrounding environment was characterized using a high spatial resolution land cover map that was linked to children by six-digit residential postal codes. Early-life residential exposure (from birth to time of EDI assessment, mean age = 5.6 years) was calculated as the mean of annual percentage values of different land cover classes (i.e., total vegetation, tree cover, grass cover, paved surfaces) within a 250 m buffer zone of postal code centroids. Multilevel models were used to analyze associations between respective land cover classes and early childhood development. ResultsIn adjusted models, one interquartile range increase in total vegetation percentage was associated with a 0.33 increase in total EDI score (95% CI: 0.21, 0.45). Similar positive associations were observed for tree cover (β-coefficient: 0.26, 95% CI: 0.15, 0.37) and grass cover (β-coefficient: 0.12, 95% CI: 0.02, 0.22), while negative associations were observed for paved surfaces (β-coefficient: −0.35, 95% CI: −0.47, −0.23). ConclusionsOur findings indicate that increased early-life residential exposure to vegetation is positively associated with early childhood developmental outcomes, and that associations may be stronger for residential exposure to tree cover relative to grass cover. Our results further indicate that childhood development may be negatively associated with residential exposure to paved surfaces. These findings can inform urban planning to support early childhood developmental health.

Landscape indicators as a tool for explaining heavy metal concentrations in urban streams

In urban watersheds, the toxicity and persistence of metals found in urban runoff has serious consequences for aquatic organisms. Landscape indicators - measures of the amount and arrangement of land cover - have been widely used as proxies for monitoring stream water quality. Yet, landscape indicators of heavy metal concentrations remain largely unexamined. Here, we investigated the utility of landscape indicators for explaining spatio-temporal trends of heavy metal concentrations in 58 streams throughout the Greater Vancouver Region in British Columbia, Canada. We asked: 1) How effective are landscape indicators in explaining instream heavy metal concentrations over different spatio-temporal scales? 2) Does explanatory power differ for landscape composition versus configuration indicators? We developed landscape indicators using a high resolution (5-m) land cover map and then used these landscape indicators to develop statistical models explaining Copper (Cu), Lead (Pb), Zinc (Zn), Cadmium (Cd) and Iron (Fe) concentrations in streams. Overall, riparian indicators explained 5–11% more variation of wet season metals than dry season, whereas watershed indicators only explained more variability in Pb during the dry season. Combining indicators mapped across multiple scales improved explanatory power of models, explaining >60% of the variability in heavy metal concentrations, regardless of season. When considering spatial arrangement of land cover, edge contrast between deciduous patches and impervious surroundings explained as much as 39% of the variability in Cu concentrations. Our approach is cost-effective, transportable, and especially useful in cities where high resolution land cover is readily available yet resources for in-stream monitoring are scarce.

Modelling Multi-Species Connectivity at the Kafue-Zambezi Interface: Implications for Transboundary Carnivore Conservation

Linking wildlife areas with corridors facilitating species dispersal between core habitats is a key intervention to reduce the deleterious effects of population isolation. Large heterogeneous networks of areas managed for wildlife protection present site- and species-scale complexity underpinning the scope and performance of proposed corridors. In Southern Africa, the Kavango-Zambezi Transfrontier Conservation Area seeks to link Kafue National Park to a cluster of wildlife areas centered in Namibia and Botswana. To assess and identify potential linkages on the Zambian side, we generated a high-resolution land cover map and combined empirical occurrence data for Lions (Panthera leo), Leopards (Panthera pardus) and Spotted Hyena (Crocuta crocuta) to build habitat suitability maps. We then developed four connectivity models to map potential single and multi-species corridors between Kafue and the Zambezi River border with Namibia. Single and multi-species connectivity models selected corridors follow broadly similar pathways narrowing significantly in central-southern areas of the Kafue-Zambezi interface, indicating a potential connectivity bottleneck. Capturing the full extent of human disturbance and barriers to connectivity remains challenging, suggesting increased risk to corridor integrity than modelled here. Notwithstanding model limitations, these data provide important results for land use planners at the Kafue-Zambezi Interface, removing much speculations from existing connectivity narratives. Failure to control human disturbance and secure corridors will leave Kafue National Park, Zambia’s majority component in the Kavango-Zambezi Transfrontier Conservation Area, isolated.

A Dual Network for Super-Resolution and Semantic Segmentation of Sentinel-2 Imagery

There is a growing interest in the development of automated data processing workflows that provide reliable, high spatial resolution land cover maps. However, high-resolution remote sensing images are not always affordable. Taking into account the free availability of Sentinel-2 satellite data, in this work we propose a deep learning model to generate high-resolution segmentation maps from low-resolution inputs in a multi-task approach. Our proposal is a dual-network model with two branches: the Single Image Super-Resolution branch, that reconstructs a high-resolution version of the input image, and the Semantic Segmentation Super-Resolution branch, that predicts a high-resolution segmentation map with a scaling factor of 2. We performed several experiments to find the best architecture, training and testing on a subset of the S2GLC 2017 dataset. We based our model on the DeepLabV3+ architecture, enhancing the model and achieving an improvement of 5% on IoU and almost 10% on the recall score. Furthermore, our qualitative results demonstrate the effectiveness and usefulness of the proposed approach.

Observation of the winter regional evaporative fraction using a UAV-based eddy covariance system over wetland area

The observation of regional evaporative fraction (EF) over a heterogeneous area is crucial to surface energy balance modeling and satellite-based evapotranspiration (ET) validation over complex areas. However, regional EF observations are often lacking or not directly acquired because the sparse distribution and inadequate spatial representativeness of ground measurements cannot represent the region's heterogeneity in landscape and land-surface processes. The UAV-based eddy-covariance (EC) system can be used to observe the regional EF at large areas. UAV-based EC measurements were performed in December 2020 at around 90 m above the ground level over Yancheng coastal wetland, which included 7 flights for comparison with ground EC measurements and 4 flights for investigating the regional EF. The results of comparison showed that the sensible and latent heat fluxes from UAV were consistent with those measured from the ground, with the R2 of 0.84, RMSE of 15.7 W/m2, bias of 50.7 W/m2 for sensible heat flux, and the R2 of 0.77, RMSE of 13.1 W/m2, bias of -3.67 W/m2 for latent heat flux. Then, 53 flux observations were obtained from the 4 regional flights after applying a quality filter. A footprint model in conjunction with a high-resolution land cover map were used to determine the fluxes contribution source areas and to determine the accumulated footprint weight of the various land cover classes within the footprint areas. Thirdly, a multiple linear regression model was used to dis-aggregate the observed fluxes into component land-cover-class-specific fluxes, and to resolve the EF of each land cover class. The dis-aggregation results revealed that cropland had the highest EF (0.69 ± 0.11), followed by Spartina alterniflora (0.3 ± 0.11) and Phragmites australis (0.27 ± 0.09). Lastly, the EF of Phragmites australis was compared with ground measurement and had a relative error of 2.9%, demonstrating that the UAV-based EC system provided a reliable observation of regional EF.

Mapping salt marsh along coastal South Carolina using U-Net

Coastal wetland mapping is often difficult because of the heterogeneous vegetation compositions and associated tidal effects. In this study, we employed the U-Net and developed an adaptive deep learning approach to map statewide salt marshes in estuarine emergent wetlands of South Carolina (SC), USA, from 20 Sentinel-2A&B images. Considering the spatial heterogeneity of the coastal environment, two NOAA National Estuarine Research Reserves (NERRs) in SC were examined, the North Inlet-Winyah Bay (NIWB) NERR for model training and the ACE Basin NERR for testing. A high-resolution land cover map in the NIWB was downloaded for the training process. Ground reference points recorded by the NERR, as well as Google Earth were utilized during the accuracy assessment. The highest overall accuracy (90%) was achieved when all scenes with 10 Sentinel bands and the Normalized Difference Vegetation Index (NDVI) were included. The time used to train the model was 5 h, while then the statewide classification was performed in 20 min. Low marsh and high marsh distributions were successfully delineated. Compared to the national marsh maps from the NOAA Coastal Change Analysis Program (C-CAP), this study refined the land cover details concerning low marsh and high marsh distributions on the SC coast. Owing to the computational power of the U-Net, the seasonality and tide influence on marsh classification were mitigated by using multi-temporal images. With images available, the deep learning approach developed in this study could be easily adopted in other coastal areas.

Forecasting deforestation in the Brazilian Amazon to prioritize conservation efforts

As Amazon deforestation rates reach the highest levels observed in the past decade, it is extremely important to direct conservation efforts to regions containing preserved forests with a high risk of deforestation. This requires forecasting deforestation, a complex endeavor due to the interplay of multiple socioeconomic, political, and environmental factors across different spatial and temporal scales. Here we couple high-resolution land-cover maps with Bayesian hierarchical spatial models to identify the main drivers of recent deforestation in the Brazilian Amazon and predict which areas are likely to lose a larger proportion of forest in the next 3 years. Recent deforestation was positively associated with forest edge density, the length of roads and waterways, elevation and terrain slope; and negatively associated with distance to urban areas, roads, and indigenous lands, area designated as protected or indigenous territory, and municipality GDP per capita. From these variables, forest edge density and distance to roads showed the largest effect sizes and highest predictive power. Predictive accuracy was highest for shorter time windows and larger grid sizes. Predicted deforestation was largely concentrated in the North-Eastern portions of the Brazilian Amazon, and amounted to roughly 3, 5, and 6 million hectares for 2020, 2021, and 2022, respectively. About 50% of this predicted deforestation is expected to occur inside protected areas or indigenous lands. Our short-term forecasts can help plan preventive measures to limit deforestation while meeting the specific needs of local areas.

REVIEW OF HIGH-RESOLUTION GLOBAL LAND COVER

Abstract. The land cover detection on our planet at high spatial resolution has a key role in many scientific and operational applications, such as climate modeling, natural resources management, biodiversity studies, urbanization analyses and spatial demography. Thanks to the progresses in Remote Sensing, accurate and high-resolution land cover maps have been developed over the last years, aiming at detecting the spatial resolution of different types of surfaces. In this paper we propose a review of the high-resolution global land cover products developed through Earth Observation technologies. A series of general information regarding imagery and data used to produce the map, the procedures employed for the map development and for the map accuracy assessment have been provided for every dataset. The land cover maps described in this paper concern the global distribution of settlements (Global Urban Footprint, Global Human Settlement Built-Up, World Settlement Footprint), water (Global Surface Water), forests (Forest/Non-forest, Tree canopy cover), and a two land cover maps describing world in 10 generic classes (GlobeLand30 and Finer Resolution Observation and Monitoring of Global Land Cover). The advantages and shortcomings of these maps and of the methods employed to produce them are summarized and compared in the conclusions.

Using machine learning to produce a very high resolution land-cover map for Ireland

Abstract. Land-cover classifications in the form of maps are required for numerical modelling of weather and climate. Such maps are often of coarse resolution and are infrequently updated. Here we propose a novel approach for land-cover classification using a Convolutional Neural Network machine learning algorithm to segment satellite images into various land-cover classes. Sentinel-2 satellite imagery, the CORINE land-cover database and the BigEarthNet dataset are used. A 10 m resolution map, called the Ulmas-Walsh map, has been created for Ireland that outperforms ECO-SG in terms of accuracy, as well as demonstrating a capacity for identifying features not labelled correctly in CORINE. The map can be updated on demand for any time of the year, subject to cloud cover. This is particularly useful for regions with large seasonal variation in land classifications such as Turloughs – seasonal lakes, flood plains and rotational crops.

Application and Evaluation of a Deep Learning Architecture to Urban Tree Canopy Mapping

Urban forest is a dynamic urban ecosystem that provides critical benefits to urban residents and the environment. Accurate mapping of urban forest plays an important role in greenspace management. In this study, we apply a deep learning model, the U-net, to urban tree canopy mapping using high-resolution aerial photographs. We evaluate the feasibility and effectiveness of the U-net in tree canopy mapping through experiments at four spatial scales—16 cm, 32 cm, 50 cm, and 100 cm. The overall performance of all approaches is validated on the ISPRS Vaihingen 2D Semantic Labeling dataset using four quantitative metrics, Dice, Intersection over Union, Overall Accuracy, and Kappa Coefficient. Two evaluations are performed to assess the model performance. Experimental results show that the U-net with the 32-cm input images perform the best with an overall accuracy of 0.9914 and an Intersection over Union of 0.9638. The U-net achieves the state-of-the-art overall performance in comparison with object-based image analysis approach and other deep learning frameworks. The outstanding performance of the U-net indicates a possibility of applying it to urban tree segmentation at a wide range of spatial scales. The U-net accurately recognizes and delineates tree canopy for different land cover features and has great potential to be adopted as an effective tool for high-resolution land cover mapping.

Unsupervised Domain Adaption for High-Resolution Coastal Land Cover Mapping with Category-Space Constrained Adversarial Network

Coastal land cover mapping (CLCM) across image domains presents a fundamental and challenging segmentation task. Although adversaries-based domain adaptation methods have been proposed to address this issue, they always implement distribution alignment via a global discriminator while ignoring the data structure. Additionally, the low inter-class variances and intricate spatial details of coastal objects may entail poor presentation. Therefore, this paper proposes a category-space constrained adversarial method to execute category-level adaptive CLCM. Focusing on the underlying category information, we introduce a category-level adversarial framework to align semantic features. We summarize two diverse strategies to extract category-wise domain labels for source and target domains, where the latter is driven by self-supervised learning. Meanwhile, we generalize the lightweight adaptation module to multiple levels across a robust baseline, aiming to fine-tune the features at different spatial scales. Furthermore, the self-supervised learning approach is also leveraged as an improvement strategy to optimize the result within segmented training. We examine our method on two converse adaptation tasks and compare them with other state-of-the-art models. The overall visualization results and evaluation metrics demonstrate that the proposed method achieves excellent performance in the domain adaptation CLCM with high-resolution remotely sensed images.