Spatial Resolution Land Cover Map Research Articles

Large-Scale Land Cover Mapping Framework Based on Prior Product Label Generation: A Case Study of Cambodia

Large-Scale land cover mapping (LLCM) based on deep learning models necessitates a substantial number of high-precision sample datasets. However, the limited availability of such datasets poses challenges in regularly updating land cover products. A commonly referenced method involves utilizing prior products (PPs) as labels to achieve up-to-date land cover mapping. Nonetheless, the accuracy of PPs at the regional level remains uncertain, and the Remote Sensing Image (RSI) corresponding to the product is not publicly accessible. Consequently, the sample dataset constructed through geographic location matching may lack precision. Errors in such datasets are not only due to inherent product discrepancies, and can also arise from temporal and scale disparities between the RSI and PPs. In order to solve the above problems, this paper proposes an LLCM framework for generating labels for use with PPs. The framework consists of three main parts. First, initial generation of labels, in which the collected PPs are integrated based on D-S evidence theory and initial labels are obtained using the generated trust map. Second, for dynamic label correction, a two-stage training method based on initial labels is adopted. The correction model is pretrained in the first stage, then the confidence probability (CP) correction module of the dynamic threshold value and NDVI correction module are introduced in the second stage. The initial labels are iteratively corrected while the model is trained using the joint correction loss, with the corrected labels obtained after training. Finally, the classification model is trained using the corrected labels. Using the proposed land cover mapping framework, this study used PPs to produce a 10 m spatial resolution land cover map of Cambodia in 2020. The overall accuracy of the land cover map was 91.68% and the Kappa value was 0.8808. Based on these results, the proposed mapping framework can effectively use PPs to update medium-resolution large-scale land cover datasets, and provides a powerful solution for label acquisition in LLCM projects.

Hierarchical-modular framework for habitat mapping through systematic and informed integration of remote sensing data with contextual information

Accurate and up-to-date habitat maps at national or regional levels are critical for informing conservation and restoration actions. Land cover maps generated from Earth Observation (EO) data can, to varying degrees, indicate the extent of some habitat categories but these are often insufficient in terms of the level of detail required. This study introduces a hierarchical-modular framework based on the globally applicable Food and Agriculture Organization (FAO) Land Cover Classification System (LCCS) whereby the land cover classes can be systematically translated to habitat categories and further differentiated by referencing contextual information and expert knowledge. Application is showcased for Wales (United Kingdom) where 10 m spatial resolution land cover maps were first constructed nationally from time-series of Sentinel-1C-band radar and − 2 optical data (2018–2020). The contextual datasets derived from non-EO sources were utilised to associate dominant vegetation types to detailed habitat classes.The accuracy assessment done using field survey data highlights that habitat classes directly derived from EO, including cultivated/managed and forest types, demonstrated higher mapping accuracies (>70% User's and Producer's Accuracies) compared to those that were more heavily dependent on non-EO-derived contextual information. These included neutral grasslands and fen/marsh/swamp that demonstrated lower accuracies (<20%) but also other detailed wetland classes (user's and producer's accuracies ranging from 25 to 60%). These reduced accuracies were largely associated with discrepancies in the contextual datasets used for their differentiation and mapping, and those integrated within field validation datasets. Dependency on contextual datasets diminished as detailed habitat maps were generalized into broader categories. The study proposes that, for enhanced habitat mapping, efforts should focus not only on EO data but also on maximising the accuracy and minimizing inconsistencies in contextual datasets as well as taxonomical systems. The flexible structure of the FAO LCCS hierarchical framework is also highlighted, emphasizing its adaptability for future improvements in habitat mapping by incorporating contextual datasets and more advanced algorithms that can improve EO-derived land cover descriptions.

Time-first approach for land cover mapping using big Earth observation data time-series in a data cube – a case study from the Lake Geneva region (Switzerland)

ABSTRACT Accurate, consistent, and high-resolution Land Use & Cover (LUC) information is fundamental for effectively monitoring landscape dynamics and better apprehending drivers, pressures, state, and impacts on land systems. Nevertheless, the availability of such national products with high thematic accuracy is still limited and consequently researchers and policymakers are constrained to work with data that do not necessarily reflect on-the-ground realities impending to correctly capture details of landscape features as well as limiting the identification and quantification of drivers and rate of change. Hereafter, we took advantage of the Switzerland’s official LUC statistical sampling survey and dense time-series of Sentinel-2 data, combining them with Machine and Deep Learning methods to produce an accurate and high spatial resolution land cover map over the Lake Geneva region. Findings suggest that time-first approach is a valuable alternative to space-first approaches, accounting for the intra-annual variability of classes, hence improving classification performances. Results demonstrate that Deep Learning methods outperform more traditional Machine Learning ones such as Random Forest, providing more accurate predictions with lower uncertainty. The produced land cover map has a high accuracy, an improved spatial resolution, while at the same time preserving the statistical significance (i.e. class proportion) of the official national dataset. This work paves the way towards the objective to produce a yearly high resolution land cover map of Switzerland and potentially implement a continuous land change monitoring capability. However further work is required to properly address challenges such as the need for increased temporal resolution for LUC information or the quality of training samples.

Spatio-temporal subpixel mapping with cloudy images

Spatio-temporal subpixel mapping (STSPM) has shown great potential for monitoring land surfaces, by generating land cover maps with both fine spatial and temporal resolutions. Selecting cloud-free fine spatial resolution images as ancillary data for STSPM can ensure that the temporal dependence term is measured for all subpixels, as in all current STSPM methods. However, such images are generally limited by cloud contamination, thereby resulting in great land cover changes between the available clear image and the desired fine spatial resolution land cover map. This research proposes a cloud-independent STSPM (C-STSPM) method to reconstruct the fine spatial resolution land cover maps by using cloudy images directly, which are assumed to have fewer land cover changes than temporally distant clear images. Cloud-independent spatio-temporal dependence was proposed in the presence of cloudy pixels. Experiments were performed under various cloud conditions involving 21 × 21 pairs of simulated cloudy images. The results demonstrate that by utilizing land cover information of clear pixels in cloudy images, more accurate prediction can be produced by C-STSPM compared to directly discarding those cloudy images, even if the number of cloud pixels increases to 95%. The advantage of C-STSPM is more evident when the clouds are distributed sparsely, which benefits from the increased number of clear pixels at the edge of the cloudy areas. Furthermore, a negative linear correlation was detected between the prediction accuracy and the ratio of overlapping cloudy pixels in the cloudy images. Moreover, the C-STSPM method helps to deal with abrupt changes occurred in the temporally distant cloud-free images by utilizing the temporally adjacent cloudy images with gradual land cover changes. Overall, the C-STSPM method provides a completely new solution to make fuller use of the widely existing cloudy images in multi-scale time-series images.

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.

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.

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.

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.

Spatial Resolution Enhancement for Large-Scale Land Cover Mapping via Weakly Supervised Deep Learning

Multispectral satellite imagery is the primary data source for monitoring land cover change and characterizing land cover globally. However, the consistency of land cover monitoring is limited by the spatial and temporal resolutions of the acquired satellite images. The public availability of daily high-resolution images is still scarce. This paper aims to fill this gap by proposing a novel spatiotemporal fusion method to enhance daily low spatial resolution land cover mapping using a weakly supervised deep convolutional neural network. We merge Sentinel images and moderate resolution imaging spectroradiometer (MODIS )-derived thematic land cover maps under the application background of massive remote sensing data and the large spatial resolution gaps between MODIS data and Sentinel images. The neural network training was conducted on the public data set SEN12MS, while the validation and testing used ground truth data from the 2020 IEEE Geoscience and Remote Sensing Society data fusion contest. The proposed data fusion method shows that the synthesized land cover map has significantly higher spatial resolution than the corresponding MODIS-derived land cover map. The ensemble approach can be implemented for generating high-resolution time series of satellite images by fusing fine images from Sentinel-1 and -2 and daily coarse images from MODIS.

Deep Subpixel Mapping Based on Semantic Information Modulated Network for Urban Land Use Mapping

Mixed pixel problem is omnipresent in remote sensing images for urban land use interpretation due to the hardware limitations. Subpixel mapping (SPM) is a usual way to solve this problem by improving the observation scale and realizing a finer spatial resolution land cover mapping. Recently, deep learning-based subpixel mapping network (DLSMNet) was proposed, benefited from its strong representation and learning ability, to restore a visually pleasing finer mapping. However, the spatial context features of artifacts are usually aggregated and progressively lost during the forward pass of the network without sufficient representation, which make it difficult to be learned and restored. In this article, a semantic information modulated (SIM) deep subpixel mapping network (SIMNet) is proposed, which uses low-resolution semantic images as prior, to reinforce the representation of spatial context features. In SIMNet, SIM module is proposed to parametrically incorporate the semantic prior into the state-of-the-art (SOTA) feed forward network architecture in an end-to-end training fashion. Furthermore, stacked SIM module with residual blocks (SIM_ResBlock) is adopted to pass the representation of spatial context feature to the deep layers, to get it fully learned during backpropagation. Experiments have been implemented on three public urban scenario data sets, and the SIMNet generates a clearer outline of artificial facilities with sufficient spatial context, and is distinctive for even individual building, which is challenging for other SOTA DLSMNet. The results demonstrate that the proposed SIMNet is a promising way for high-resolution urban land use mapping from easily available lower resolution remote sensing images.Mixed pixel problem is omnipresent in remote sensing images for urban land use interpretation due to the hardware limitations. Subpixel mapping (SPM) is a usual way to solve this problem by improving the observation scale and realizing a finer spatial resolution land cover mapping. Recently, deep learning-based subpixel mapping network (DLSMNet) was proposed, benefited from its strong representation and learning ability, to restore a visually pleasing finer mapping. However, the spatial context features of artifacts are usually aggregated and progressively lost during the forward pass of the network without sufficient representation, which make it difficult to be learned and restored. In this article, a semantic information modulated (SIM) deep subpixel mapping network (SIMNet) is proposed, which uses low-resolution semantic images as prior, to reinforce the representation of spatial context features. In SIMNet, SIM module is proposed to parametrically incorporate the semantic prior into the state-of-the-art (SOTA) feed forward network architecture in an end-to-end training fashion. Furthermore, stacked SIM module with residual blocks (SIM_ResBlock) is adopted to pass the representation of spatial context feature to the deep layers, to get it fully learned during backpropagation. Experiments have been implemented on three public urban scenario data sets, and the SIMNet generates a clearer outline of artificial facilities with sufficient spatial context, and is distinctive for even individual building, which is challenging for other SOTA DLSMNet. The results demonstrate that the proposed SIMNet is a promising way for high-resolution urban land use mapping from easily available lower resolution remote sensing images.

Large-Area, High Spatial Resolution Land Cover Mapping Using Random Forests, GEOBIA, and NAIP Orthophotography: Findings and Recommendations

Despite the need for quality land cover information, large-area, high spatial resolution land cover mapping has proven to be a difficult task for a variety of reasons including large data volumes, complexity of developing training and validation datasets, data availability, and heterogeneity in data and landscape conditions. We investigate the use of geographic object-based image analysis (GEOBIA), random forest (RF) machine learning, and National Agriculture Imagery Program (NAIP) orthophotography for mapping general land cover across the entire state of West Virginia, USA, an area of roughly 62,000 km2. We obtained an overall accuracy of 96.7% and a Kappa statistic of 0.886 using a combination of NAIP orthophotography and ancillary data. Despite the high overall classification accuracy, some classes were difficult to differentiate, as highlight by the low user’s and producer’s accuracies for the barren, impervious, and mixed developed classes. In contrast, forest, low vegetation, and water were generally mapped with accuracy. The inclusion of ancillary data and first- and second-order textural measures generally improved classification accuracy whereas band indices and object geometric measures were less valuable. Including super-object attributes improved the classification slightly; however, this increased the computational time and complexity. From the findings of this research and previous studies, recommendations are provided for mapping large spatial extents.

Super-resolution land cover mapping by deep learning

ABSTRACTSuper-resolution mapping (SRM) is a technique to estimate a fine spatial resolution land cover map from coarse spatial resolution fractional proportion images. SRM is often based explicitly on the use of a spatial pattern model that represents the land cover mosaic at the fine spatial resolution. Recently developed deep learning methods have considerable potential as an alternative approach for SRM, based on learning the spatial pattern of land cover from existing fine resolution data such as land cover maps. This letter proposes a deep learning-based SRM algorithm (DeepSRM). A deep convolutional neural network was first trained to estimate a fine resolution indicator image for each class from the coarse resolution fractional image, and all indicator maps were then combined to create the final fine resolution land cover map based on the maximal value strategy. The results of an experiment undertaken with simulated images show that DeepSRM was superior to conventional hard classification and a suite of popular SRM algorithms, yielding the most accurate land cover representation. Consequently, methods such as DeepSRM may help exploit the potential of remote sensing as a source of accurate land cover information.

Effects of Impervious Surface on the Spatial Distribution of Urban Waterlogging Risk Spots at Multiple Scales in Guangzhou, South China

An impervious surface is considered one of main factors affecting urban waterlogging. Previous studies found that spatial pattern (composition and configuration) of impervious surfaces affected urban waterlogging. However, their relative importance remains unknown, and the scale-effect of the spatial pattern on urban waterlogging has been ignored. To move forward, our research studied the relationship between spatial patterns on the impervious surface and its subcategories (building and pavement) on urban waterlogging risk spots using Pearson correlation, partial redundancy analysis and performed at three grid scales (1 km × 1 km, 3 km × 3 km, 5 km × 5 km) and the catchment scale based on different spatial resolution land cover maps (2 m, 10 m and 30 m). We identified positively-correlated metrics with urban waterlogging risk spots, such as the composition of impervious surface (i.e., total impervious surface, building, pavement) and the aggregation metric of the total impervious surface at most scales, as well as two negatively correlated metrics (i.e., proximity metric of building, fragmentation metric of total impervious surface). Furthermore, the total variance of urban waterlogging risk spots explained by the spatial pattern of the total impervious surface and its subcategories increased with studied grid and catchment scales while decreasing from a fine to a coarse resolution. The relative contribution of the impervious surface composition and configuration to the variation of urban waterlogging risk spots varied across scales and among impervious surface types. The composition contributed more than the configuration did for the total impervious surface at both grid and catchment scales. Similar to total impervious surface, the composition of buildings was more important than its configuration was at all the grid scales, while the configuration of buildings was more important at the catchment scale. Contrary to the total impervious surface, the configuration of pavement at both the grid and catchment scales mattered more than their compositions did. Furthermore, the composition of the building was more important than that of pavement, but its configuration mattered less. Our study could provide a multi-scale landscape perspective with detailed suggestions for controlling the area of impervious surface and optimizing its spatial configuration in urban waterlogging risk mitigation and urban planning.

Generating a series of fine spatial and temporal resolution land cover maps by fusing coarse spatial resolution remotely sensed images and fine spatial resolution land cover maps

Studies of land cover dynamics would benefit greatly from the generation of land cover maps at both fine spatial and temporal resolutions. Fine spatial resolution images are usually acquired relatively infrequently, whereas coarse spatial resolution images may be acquired with a high repetition rate but may not capture the spatial detail of the land cover mosaic of the region of interest. Traditional image spatial–temporal fusion methods focus on the blending of pixel spectra reflectance values and do not directly provide land cover maps or information on land cover dynamics. In this research, a novel Spatial–Temporal remotely sensed Images and land cover Maps Fusion Model (STIMFM) is proposed to produce land cover maps at both fine spatial and temporal resolutions using a series of coarse spatial resolution images together with a few fine spatial resolution land cover maps that pre- and post-date the series of coarse spatial resolution images. STIMFM integrates both the spatial and temporal dependences of fine spatial resolution pixels and outputs a series of fine spatial–temporal resolution land cover maps instead of reflectance images, which can be used directly for studies of land cover dynamics. Here, three experiments based on simulated and real remotely sensed images were undertaken to evaluate the STIMFM for studies of land cover change. These experiments included comparative assessment of methods based on single-date image such as the super-resolution approaches (e.g., pixel swapping-based super-resolution mapping) and the state-of-the-art spatial–temporal fusion approach that used the Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM) and the Flexible Spatiotemporal DAta Fusion model (FSDAF) to predict the fine-resolution images, in which the maximum likelihood classifier and the automated land cover updating approach based on integrated change detection and classification method were then applied to generate the fine-resolution land cover maps. Results show that the methods based on single-date image failed to predict the pixels of changed and unchanged land cover with high accuracy. The land cover maps that were obtained by classification of the reflectance images outputted from ESTARFM and FSDAF contained substantial misclassification, and the classification accuracy was lower for pixels of changed land cover than for pixels of unchanged land cover. In addition, STIMFM predicted fine spatial–temporal resolution land cover maps from a series of Landsat images and a few Google Earth images, to which ESTARFM and FSDAF that require correlation in reflectance bands in coarse and fine images cannot be applied. Notably, STIMFM generated higher accuracy for pixels of both changed and unchanged land cover in comparison with other methods.

Can C-band synthetic aperture radar be used to estimate soil organic carbon storage in tundra?

Abstract. A new approach for the estimation of soil organic carbon (SOC) pools north of the tree line has been developed based on synthetic aperture radar (SAR; ENVISAT Advanced SAR Global Monitoring mode) data. SOC values are directly determined from backscatter values instead of upscaling using land cover or soil classes. The multi-mode capability of SAR allows application across scales. It can be shown that measurements in C band under frozen conditions represent vegetation and surface structure properties which relate to soil properties, specifically SOC. It is estimated that at least 29 Pg C is stored in the upper 30 cm of soils north of the tree line. This is approximately 25 % less than stocks derived from the soil-map-based Northern Circumpolar Soil Carbon Database (NCSCD). The total stored carbon is underestimated since the established empirical relationship is not valid for peatlands or strongly cryoturbated soils. The approach does, however, provide the first spatially consistent account of soil organic carbon across the Arctic. Furthermore, it could be shown that values obtained from 1 km resolution SAR correspond to accounts based on a high spatial resolution (2 m) land cover map over a study area of about 7 × 7 km in NE Siberia. The approach can be also potentially transferred to medium-resolution C-band SAR data such as ENVISAT ASAR Wide Swath with ∼ 120 m resolution but it is in general limited to regions without woody vegetation. Global Monitoring-mode-derived SOC increases with unfrozen period length. This indicates the importance of this parameter for modelling of the spatial distribution of soil organic carbon storage.

High spatial resolution land use and land cover mapping of the Brazilian Legal Amazon in 2008 using Landsat-5/TM and MODIS data

ABSTRACT Understanding spatial patterns of land use and land cover is essential for studies addressing biodiversity, climate change and environmental modeling as well as for the design and monitoring of land use policies. The aim of this study was to create a detailed map of land use land cover of the deforested areas of the Brazilian Legal Amazon up to 2008. Deforestation data from and uses were mapped with Landsat-5/TM images analysed with techniques, such as linear spectral mixture model, threshold slicing and visual interpretation, aided by temporal information extracted from NDVI MODIS time series. The result is a high spatial resolution of land use and land cover map of the entire Brazilian Legal Amazon for the year 2008 and corresponding calculation of area occupied by different land use classes. The results showed that the four classes of Pasture covered 62% of the deforested areas of the Brazilian Legal Amazon, followed by Secondary Vegetation with 21%. The area occupied by Annual Agriculture covered less than 5% of deforested areas; the remaining areas were distributed among six other land use classes. The maps generated from this project - called TerraClass - are available at INPE's web site (http://www.inpe.br/cra/projetos_pesquisas/terraclass2008.php).

Assessing a Temporal Change Strategy for Sub-Pixel Land Cover Change Mapping from Multi-Scale Remote Sensing Imagery

Remotely sensed imagery is an attractive source of information for mapping and monitoring land cover. Fine spatial resolution imagery is typically acquired infrequently, but fine temporal resolution systems commonly provide coarse spatial resolution imagery. Sub-pixel land cover change mapping is a method that aims to use the advantages of these multiple spatial and temporal resolution sensing systems. This method produces fine spatial and temporal resolution land cover maps, by updating fine spatial resolution land cover maps using coarse spatial resolution remote sensing imagery. A critical issue for sub-pixel land cover change mapping is downscaling coarse spatial resolution fraction maps estimated by soft classification to a fine spatial resolution land cover map. The relationship between a historic fine spatial resolution map and a contemporary fine spatial resolution map to be estimated at a more recent date plays an important role in the downscaling procedure. A change strategy based on the assumption that the change for each land cover class in a coarse spatial resolution pixel is unidirectional was shown to be a promising means to describe this relationship. This paper aims to assess this change strategy by analyzing the factors that affect the accuracy of the change strategy, using six subsets of the National Land Cover Database (NLCD) of USA. The results show that the spatial resolution of coarse pixels, the time interval of the previous fine resolution land cover map and the current coarse spatial resolution images, and the thematic resolution of the used land cover class scheme have considerable influence on the accuracy of the change strategy. The accuracy of the change strategy decreases with the coarsening of spatial resolution, an increase of time interval, and an increase of thematic resolution. The results also indicate that, when the historic land cover map has a 30 m resolution, like the NLCD, the average accuracy of the change strategy is still as high as 92% when the coarse spatial resolution data used had a resolution of ~1000 m, confirming the effectiveness of the change strategy used in sub-pixel land cover change mapping for use with popular remote sensing systems.

2015 Land cover map of Southeast Asia at 250 m spatial resolution

ABSTRACTIn this letter, we present the methodology and accuracy assessment results of a new 250 m spatial resolution land cover map of Southeast Asia (covering Brunei, Cambodia, Indonesia, Laos, Malaysia, Myanmar, Philippines, Singapore, Thailand, Timor-Leste and Vietnam). The map is based on semi-automated classification of Moderate Resolution Imaging Spectroradiometer data (acquired January to November 2015) supported by various auxiliary data sets. The classification scheme with 18 land cover classes reflects the characteristics of land cover in Southeast Asia and aims to provide useful input for various regional-level analyses requiring information on land cover distribution. The accuracy of the map is assessed separately for the seasonal (north of 9° N) and humid (south of 9° N) parts of the region, reaching 82% and 86% overall accuracies respectively. The map will be made available at: https://ormt-crisp.nus.edu.sg/ormt/Home/Disclaimer

Fast Subpixel Mapping Algorithms for Subpixel Resolution Change Detection

Due to rapid changes on the Earth's surface, it is important to perform land cover change detection (CD) at a fine spatial and fine temporal resolution. However, remote sensing images with both fine spatial and temporal resolutions are commonly not available or, where available, may be expensive to obtain. This paper attempts to achieve fine spatial and temporal resolution land cover CD with a new computer technology based on subpixel mapping (SPM): The fine spatial resolution land cover maps (FRMs) are first predicted through SPM of the coarse spatial but fine temporal resolution images, and then, subpixel resolution CD is performed by comparison of class labels in the SPM results. For the first time, five fast SPM algorithms, including bilinear interpolation, bicubic interpolation, subpixel/pixel spatial attraction model, Kriging, and radial basis function interpolation methods, are proposed for subpixel resolution CD. The auxiliary information from the known FRM on one date is incorporated in SPM of coarse images on other dates to increase the CD accuracy. Based on the five fast SPM algorithms and the availability of the FRM, subpixels for each class are predicted by comparison of the estimated soft class values at the target fine spatial resolution and borrowing information from the FRM. Experiments demonstrate the feasibility of the five SPM algorithms using FRM in subpixel resolution CD. They are fast methods to achieve subpixel resolution CD.

Land Cover Change Detection at Subpixel Resolution With a Hopfield Neural Network

In this paper, a new subpixel resolution land cover change detection (LCCD) method based on the Hopfield neural network (HNN) is proposed. The new method borrows information from a known fine spatial resolution land cover map (FSRM) representing one date for subpixel mapping (SPM) from a coarse spatial resolution image on another, closer date. It is implemented by using the thematic information in the FSRM to modify the initialization of neuron values in the original HNN. The predicted SPM result was compared to the original FSRM to achieve subpixel resolution LCCD. The proposed method was compared with the original unmodified HNN method as well as six state-of-the-art methods for LCCD. To explore the effect of uncertainty in spectral unmixing, which mainly originates from spectral separability in the input, coarse image, and the point spread function (PSF) of the sensor, a set of synthetic multispectral images with different class separabilities and PSFs was used in experiments. It was found that the proposed LCCD method (i.e., HNN with an FSRM) can separate more real changes from noise and produce more accurate LCCD results than the state-of-the-art methods. The advantage of the proposed method is more evident when the class separability is small and the variance in the PSF is large, that is, the uncertainty in spectral unmixing is large. Furthermore, the utilization of an FSRM can expedite the HNN-based processing required for LCCD. The advantage of the proposed method was also validated by applying to a set of real Landsat-Moderate Resolution Imaging Spectroradiometer (MODIS) images.