We present a novel Multi-Agent Reinforcement Learning approach to understand and improve policy development by land-shaping agents, such as governments and institutional bodies. We derive the underlying policy decisions by analyzing the land and developing an intelligent system that proposes optimal land conversion strategies. The aim is an efficient method for allocating residential spaces while considering the dynamic population influx in different regions, jurisdictional constraints, and the intrinsic characteristics of the land. Our main goal is to be sustainable, preserving desirable land types such as forests and fluvial lands while optimizing land organization. We introduce an attractiveness metric that quantifies the proximity to different land types and other factors to optimize land usage. It distinguishes two types of agents: ``top-down'' agents, which are policymakers and shareholders, and ``bottom-up'' agents representing individuals or groups with specific housing preferences. Our main objective is to create a synergistic environment where the top-down policy meets the bottom-up preferences to devise a comprehensive land use and conversion strategy. This paper, thus, serves as a pivotal reference point for future urban planning and policy-making processes, contributing to a sustainable and efficient landscape design model.

Fractional vegetation cover (FVC) is a critical component of ecosystems, global climate change and the carbon cycle. Several FVC products have been released, the most widely used of which are the GLASS FVC products (including the GLASS-MODIS and GLASS-AVHRR FVC products). Specifically, the GLASS-MODIS FVC product covers the period from 2000 to present with a 500 m spatial resolution, whereas the GLASS-AVHRR FVC product is available from 1982 to present with a coarser spatial resolution of 5 km. For local monitoring of patterns of change in vegetation, however, there is a great need for fine spatial resolution (e.g., 500 m in this paper) and long-term time-series FVC datasets. To this end, we proposed to reconstruct a 500 m, 8-day historical MODIS FVC dataset (1982–2000) by making full use of the advantages of the existing GLASS-MODIS FVC (fine spatial resolution of 500 m) and GLASS-AVHRR FVC (long-term coverage from 1982 to the present) products covering China in this paper. The known GLASS-AVHRR FVC product was first used to fit the relationship between the FVC data after 2000 and before 2000, based on a random forest (RF) model. The trained relationship was migrated to the GLASS-MODIS FVC product, that is, predicting the MODIS FVC before 2000 based on the input of MODIS FVC after 2000. The validation using 64 scenes of Landsat FVC reference data revealed that the predicted historical MODIS FVC dataset has a reliable accuracy with a correlation coefficient (CC) value of 0.84, root mean square error (RMSE) of 0.14, Bias of 0.04 and unbiased RMSE (ubRMSE) of 0.12. Moreover, an accuracy evaluation in seven different regions in 1999 suggested that the historical MODIS FVC is closer to the Landsat FVC than the GEOV2 FVC product. Overall, the 500 m, 8-day MODIS FVC dataset (1982–2000) in China can provide important historical data for long-term, local monitoring of vegetation, which has great potential in supporting studies in a range of applications areas including ecology, hydrology and climatology. This dataset is available at https://doi.org/10.6084/m9.figshare.24616446.v1.

Solar-induced chlorophyll fluorescence (SIF) is a crucial variable towards timely and effective monitoring of vegetation productivity, as well as physiological and biochemical parameters, across extensive areas. Among these advances, the TROPOspheric Monitoring Instrument (TROPOMI) SIF has significantly increased the spatiotemporal resolution and data coverage compared to previous sensors. However, TROPOMI SIF data suffer from nonuniform sampling, swath gaps and cloud contamination, resulting in numerous instances of missing data. In this paper, we proposed a physical and spatial information-aided gap filling (PSGF) method, which addresses effectively the missing data problem, generating a spatially seamless, 0.05°, daily SIF (S2-SIF) dataset globally at a spatial resolution of 0.05° from 2018 to 2021. Through missing data simulation experiments conducted in six regions worldwide, we demonstrated consistency between the reference SIF and the filled SIF, with a correlation coefficient (CC) of 0.659. Furthermore, validation usingin situdata from 35 SIF and gross primary productivity (GPP) ground sites yielded a CC of approximately 0.70 for the SIF sites and CC values above 0.60 between the ground GPP and filled SIF. Additionally, consistency was observed between the filled SIF datasets and two other SIF products across 11 vegetation types, confirming the reliability of the filled SIF data and the efficacy of the PSGF method. The produced filled SIF data are made publicly available and should increase greatly the applicability of the daily SIF data for a wide range of applications, including quantifying the photosynthesis of vegetation and accurately estimating GPP globally.

Extracting 3D building information from monocular satellite sensor imagery remains a formidable challenge in the field of remote sensing. Multitask frameworks based on deep learning, which particularly for simultaneously predicting 2D building outlines and their respective heights using ortho-rectified satellite imagery, have shown promise in addressing this challenge. Height estimation is notably complex due to the absence of explicit height indicators, limited interaction between semantic-height features, and inadequate representation of building relationships. Moreover, the availability of data sources is a limiting factor for broader application. To overcome these issues, this study introduces an innovative instance-level multitask learning model (named BDH-Net) that leverages off-nadir perspectives and roof-to-footprint offset vectors to enhance modeling. This model comprises four key components: a pixel-wise feature extraction image encoder-decoder, a query transformer decoder, a multitask decoder, and a height decoder that employs intra-instance and inter-instance attention for precise building height estimation. Additionally, we pioneer the use of Google Earth imagery to construct an off-nadir satellite dataset with roof-to-footprint offset vectors specifically designed for building instance segmentation and height prediction, known as the BDH dataset. Comprehensive experiments demonstrate that the proposed BDH-Net significantly improves the accuracy of monocular 3D building data extraction by integrating roof-to-footprint offset vectors and leveraging context specific to each building instance. With the extensive coverage and regular updates of Google Earth imagery, BDH-Net holds substantial potential for wide-ranging and long-term applications. The source code of the proposed BDH-Net and the BDH dataset are publicly available at https://github.com/wishx98/BDHNet.

Spatio-temporal fusion is a general technique used to blend fine spatial resolution and fine temporal resolution remote sensing data from multiple sensors, to generate time-series data with both fine spatial and temporal resolutions. It has received increasing attention in recent years. Drastic changes in land surface, however, pose great challenges for spatio-temporal fusion. To address this issue, this paper proposed a spatio-temporal fusion method which integrates multi-modal data (M2-STF), specifically SAR data with optical data. Considering the scenario of flooding (which causes drastic land surface changes) as an example, this study focused on spatio-temporal fusion based on Sentinel-2 MSI and Sentinel-3 OLCI data, and developed the M2-STF method by integrating Sentinel-1 SAR data. For the changed area, M2-STF integrates the Sentinel-2 image at the known time and the Sentinel-1 SAR image at the prediction time to obtain a more accurate fine spatial resolution classification map at the prediction time. Based on this map, a spatial unmixing model and spatial interpolation model were developed taking into account both homogeneity and heterogeneity characteristics, which were then combined into a homogeneity index. For the unchanged area, a new similar pixel selection strategy was constructed to exclude the influence of similar pixels from the changed area. In the experiments, three regions were selected for validation, and M2-STF was compared with five typical spatio-temporal fusion methods. By integrating Sentinel-1 SAR data at the prediction time, the accuracy of spatio-temporal fusion was increased remarkably, especially when the land surface changes greatly from the known to the prediction times. Specifically, the M2-STF method outperforms all five benchmark methods, by reducing root mean square error (RMSE) by at least 16%.

Extensive cloud contamination severely hinders the interpretation of optical remote sensing images. Existing cloud removal methods focus primarily on the reconstruction of individual cloudy images, with few studies addressing the reconstruction of cloudy time-series images. Furthermore, current methods tend to prioritize using cloud-free auxiliary images while overlooking valuable information present in the cloudy auxiliary images that are temporally closer to the target cloudy image. In this paper, we proposed a deep network called Res-cLSTM to reconstruct cloudy time-series images. Res-cLSTM processes time-series images sequentially using convolutional LSTM, synthesizing long- and short-term memory streams to match the complex temporal relationships amongst them. Then, Res-cLSTM further decodes the feature maps using a refined residual module with skip connections, resulting in the final output. Simulated and real cloud removal experiments on Landsat 8 OLI time-series data across five different regions demonstrated that Res-cLSTM is an effective cloud removal method, which can produce more accurate predictions than three benchmark approaches. For example, for reconstruction of the cloudy time-series of three simulated cloudy areas, the average CC of the Res-cLSTM prediction is about 0.01, 0.04 and 0.04 larger than that of the second most accurate method (i.e., autoencoder (AE)). As a lightweight network, Res-cLSTM does not require global sampling of training data and can fully exploit the valuable information in the non-cloud regions of cloudy time-series images to facilitate cloud removal. Moreover, Res-cLSTM demonstrates robustness to thin cloud omission and exhibits a faster convergence rate, thus, holds great potential for practical applications requiring real-time processing.

Temporally neighboring homologous images are crucial to provide auxiliary information for thick cloud removal. Due to the inherent satellite revisit period and frequent cloud obscuration, there is often a significant time interval between the target cloudy images and neighboring cloud-free homologous images, leading to potential land surface condition changes. Moreover, multitemporal cloudy images that may contain valuable complementary information in the noncloudy regions, are often neglected in practice. This article focused on thick cloud removal from Landsat 8 OLI images. We proposed to fuse the temporally more frequent Sentinel-2 MSI images and also cloudy multitemporal images consisting of Sentinel-2 MSI and Landsat 8 OLI time-series. Acquired by a sensor different from Landsat 8 OLI, Sentinel-2 MSI images exhibit great similarities in data characteristics. To fully exploit the spatio-temporal-spectral information in multisource and multitemporal auxiliary images, we proposed a novel deep network called MST-Net. MST-Net was validated using 12 simulated and two real cloudy Landsat 8 OLI images. The results show that the MST-Net can produce more satisfactory predictions than the five benchmark methods. Both the images acquired by a different sensor and homogeneous multitemporal cloudy images are beneficial. Under different sizes of clouds, the MST-Net produces consistently the most accurate predictions. Furthermore, due to the fusion of all bands simultaneously in the temporally closest Sentinel-2 MSI images, the MST-Net is less affected by thin cloud occlusion errors. Overall, the MST-Net shows great potential for cloud removal from optical images produced by a wide range of sensors and, more generally, filling gaps in various global scale products.

Semantic change detection (SCD) involves the simultaneous extraction of changed regions and their corresponding semantic classifications (pre- and post-change) in remote sensing images (RSIs). Despite recent advancements in vision foundation models (VFMs), the fast-segment anything model has demonstrated insufficient performance in SCD. In this article, we propose a novel VFMs architecture for SCD, designated as VFM-ReSCD. This architecture integrates a side adapter (SA) into the VFM-ReSCD to fine-tune the fast segment anything model (FastSAM) network, enabling zero-shot transfer to novel image distributions and tasks. This enhancement facilitates the extraction of spatial features from very high-resolution (VHR) RSIs. Moreover, we introduce a recurrent neural network (RNN) to model semantic correlation and capture feature changes. We evaluated the proposed methodology on two benchmark datasets. Extensive experiments show that our method achieves state-of-the-art (SOTA) performances over existing approaches and outperforms other CNN-based methods on two RSI datasets.

Change detection(CD) is important for Earth observation, emergency response and time-series understanding. Recently, data availability in various modalities has increased rapidly, and multimodal change detection (MCD) is gaining prominence. Given the scarcity of datasets and labels for MCD, unsupervised approaches are more practical for MCD. However, previous methods typically either merely reduce the gap between multimodal data through transformation or feed the original multimodal data directly into the discriminant network for difference extraction. The former faces challenges in extracting precise difference features. The latter contains the pronounced intrinsic distinction between the original multimodal data; direct extraction and comparison of features usually introduce significant noise, thereby compromising the quality of the resultant difference image. In this article, we proposed the MaCon framework to synergistically distill the common and discrepancy representations. The MaCon framework unifies mask reconstruction (MR) and contrastive learning (CL) self-supervised paradigms, where the MR serves the purpose of transformation while CL focuses on discrimination. Moreover, we presented an optimal sampling strategy in the CL architecture, enabling the CL subnetwork to extract more distinguishable discrepancy representations. Furthermore, we developed an effective silent attention mechanism that not only enhances contrast in output representations but stabilizes the training. Experimental results on both multimodal and monomodal datasets demonstrate that the MaCon framework effectively distills the intrinsic common representations between varied modalities and manifests state-of-the-art performance across both multimodal and monomodal CD. Such findings imply that the MaCon possesses the potential to serve as a unified framework in the CD and relevant fields. Source code will be publicly available once the article is accepted.

The extraction of building heights from monocular satellite sensor imagery poses a significant difficulty in remote sensing. Deep learning-based multi-task frameworks have recently emerged, showing potential in concurrently predicting 2D building shapes and heights from ortho-rectified satellite images. These techniques, however, have several limitations, including a lack of direct height markers, inherent uncertainties in height estimation, a heavy reliance on limited digital surface models, suboptimal integration of semantic-height feature, and inadequate depiction of the spatial relationships between buildings. To address these challenges, we develop a novel instance-level multi-task learning model that uses off-nadir imaging angles, roof-to-footprint as a proxy for height, and integrates certain geometric properties to increase height estimation accuracy. This method was validated using a self-constructed dataset, demonstrating significantly superior performance relative to two benchmarks.

Change detection (CD) has become increasingly important in remote sensing and Earth observation. Currently, the data in various modalities has rapidly increased, and multimodal change detection is gaining prominence and holds substantial potential for applications demanding high temporal frequency or rapid response. In this research, we proposed a novel CDR-Net architecture for unsupervised multimodal change detection. The CDR-Net fuses the merits of the mask reconstruction and contrastive learning self-supervised paradigm. Within this architecture, the mask reconstruction subnetwcork pays more attention to low-level details, distilling common representations between multimodal remote sensing images to make them comparable, while the CL subnetwork emphasizes high-level semantics, extracting discrepant representations to facilitate the change detection task. Experimental results demonstrated that the CDR-Net achieved outstanding performance. This implies that the CDR-Net is of great value for resource investigation, emergency response and time-series understanding.

Due to its capacity to recognize detailed spectral differences, hyperspectral data have been extensively used for precise Land Use Land Cover (LULC) mapping. However, recent multi-modal methods have shown their superior classification performance over the algorithms that use single data sets. On the other hand, Convolutional Neural Networks (CNNs) are models extensively utilized for the hierarchical extraction of features. Vision transformers (ViTs), through a self-attention mechanism, have recently achieved superior modeling of global contextual information compared to CNNs. However, to harness their image classification strength, ViTs require substantial training datasets. In cases where the available training data is limited, current advanced multi-layer perceptrons (MLPs) can provide viable alternatives to both deep CNNs and ViTs. In this paper, we developed the SGU-MLP, a deep learning algorithm that effectively combines MLPs and spatial gating units (SGUs) for precise Land Use Land Cover (LULC) mapping using multi-modal data from multi-spectral, LiDAR, and hyperspectral data. Results illustrated the superiority of the developed SGU-MLP classification algorithm over several CNN and CNN-ViT-based models, including HybridSN, ResNet, iFormer, EfficientFormer, and CoAtNet. The SGU-MLP classification model consistently outperformed the benchmark CNN and CNN-ViT-based algorithms. The code will be made publicly available at https: //github.com/aj1365/SGUMLP.

Convolutional Neural Networks (CNNs) are employed extensively in remote sensing due to their capacity to capture intricate features from a broad range of object patterns, irrespective of object size, shape or color. These networks excel at extracting high-frequency spectral information such as angles, edges and outlines. The classification boundary zone, however, becomes hazy for CNNs because they learn characteristics by means of a fixed shape kernel concentrated on the central pixel, and can perform poorly in image classification at class boundaries. Additionally, CNNs are not designed to capture global relations. Thus, in this letter, we propose an Attention Graph Convolutional Network (Attention-GCN) as a solution to the aforementioned shortcomings. The developed model illustrated a high level of superiority over several CNN and ViT-based models. For example, in the Augsburg data benchmark, the developed algorithm exhibited an average accuracy of 61.11%, substantially outperforming other models such as HybridSN, iFormer, Efficient Former, GCN, CoAtNet, 2D-CNN, 3D-CNN, and ResNet by approximately 9, 13, 14, 15, 18, 24, 25 and 29 percentage points, respectively. The code will be made publicly available at https://github.com/aj1365/AGCN.

Multimodal image fusion has recently garnered increasing interest in the field of remote sensing. By leveraging the complementary information in different modalities, the fused results may be more favorable in characterizing objects of interest, thereby increasing the chance of a more comprehensive and accurate perception of the scene. Unfortunately, most existing fusion methods tend to extract modality-specific features independently without considering intermodal alignment and complementarity, leading to a suboptimal fusion process. To address this issue, we propose a novel interactive generative adversarial network (IG-GAN), for the task of multimodal image fusion. IG-GAN comprises guided dual streams tailored for enhanced learning of details and content, as well as cross-modal consistency. Specifically, a details-guided interactive running-in module (GIR1) and a content-guided interactive running-in module (GIR2) are developed, with the stronger modality serving as guidance for detail richness or content integrity, and the weaker one assisting. To fully integrate multigranularity features from dual-modality, a hierarchical fusion and reconstruction branch is established. Specifically, a shallow interactive fusion (SIF) module followed by a multilevel interactive fusion (MIF) module is designed to aggregate multilevel local and long-range features. Concerning feature decoding and fused image generation, a high-level interactive fusion and reconstruction module (HRM) is further developed. In addition, to empower the fusion network to generate fused images with complete content, sharp edges, and high fidelity without supervision, a loss function facilitating the mutual game between the generator and two discriminators is also formulated. Comparative experiments with 14 state-of-the-art methods are conducted on three datasets. Qualitative and quantitative results indicate that IG-GAN exhibits obvious superiority in terms of both visual effect and quantitative metrics. Moreover, experiments on two RGB-IR object detection datasets are also conducted, which demonstrate that IG-GAN can enhance the accuracy of object detection by integrating complementary information from different modalities. The code will be available athttps://github.com/flower6top.

Landsat surface temperature (LST) is an important physical quantity for global climate change monitoring. Over the past decades, several LST products have been produced by satellite thermal infrared (TIR) bands or land surface models (LSMs). Recent research has increased the spatio-temporal resolution of LST products to 2 km, hourly based on Geostationary Operational Environmental Satellites (GOES)-R Advanced Baseline Imager (ABI) LST data. The spatial resolution of 2 km, however, is insufficient for monitoring at the regional scale. This paper investigates the feasibility of applying spatio-temporal fusion to generate reliable 100 m, hourly LST data based on fusion of the newly released 2 km, hourly GOES-16 ABI LST and 100 m Landsat LST data. The most accurate fusion method was identified through a comparison between several popular methods. Furthermore, a comprehensive comparison was performed between fusion (with Landsat LST) involving satellite-derived LST (i.e., GOES) and model-derived LSMs (i.e., European Centre for Medium-range Weather Forecasts (ECMWF) Reanalysisv.5 (ERA5)-Land). The spatial and temporal adaptive reflectance fusion model (STARFM) method was demonstrated to be an appropriate method to generate 100 m, hourly data, which produced an average root mean square error (RMSE) of 2.640 K, mean absolute error (MAE) of 2.159 K and average coefficient of determination (R2) of 0.982 referring to thein situtime-series. Furthermore, inheriting the advantages of direct observation, and the fusion of Landsat and GOES for the generation of 100 m, hourly LST produced greater accuracy compared to the fusion of Landsat and ERA5-Land LST in the experiments. The generated 100 m, hourly LST can provide important diurnal data with fine spatial resolution for various monitoring applications.

Soil moisture (SM) plays a significant role in many natural and anthropogenic systems. Thus, accurate assessment of changes in SM globally is of great value, including long-term historical assessment. The European Space Agency established the Climate Change Initiative (CCI) program to produce long time-series surface SM datasets starting from 1978 to the present. However, the Soil Moisture Active Passive (SMAP) mission, launched in 2015, has shown more satisfactory performance in both spatial accuracy and in capturing the pattern of temporal changes. In this paper, a random forest (RF) model was proposed to extend the SMAP dataset historically (named Hist_SMAP), using the corresponding CCI SM time-series. We assumed that the temporal changes in the SMAP SM dataset are similar generally to those in the available CCI dataset. Accordingly, the RF model was constructed using the temporal (extracted from the CCI SM data), coupled with terrain and location characteristics, and migrated to predict the Hist_SMAP dataset. The availablein-situand the real SMAP data were used as references for validation. Compared with the CCI dataset, the predicted Hist_SMAP dataset is closer to thein-situSM data and the real SMAP data. Moreover, the historical Hist_SMAP dataset is more accurate than the widely used Global Land Evaporation Amsterdam Model (GLEAM) dataset. Thus, the Hist_SMAP dataset was shown to be a reliable substitute for the historical CCI dataset. The new long time-series Hist_SMAP dataset is provided with free access and will be of great value for research and practical application in a range of fields.

Subpixel mapping (SPM) can address the mixed pixel problem by producing land cover maps at a finer spatial resolution than the input images. The Hopfield neural network (HNN) method has shown great advantages in SPM and various extended versions have been developed recently. However, a long-standing issue in the HNN, especially in the multiclass scenario, is its tendency to fall into local optima with vanished gradients, failing to push subpixels to the hard class label of 0 or 1. This can lead to great uncertainties in determining hard class labels and, moreover, the disappearance of many small-sized land cover features and spatial details. In this article, we proposed a hard-constrained HNN (H-HNN) model that introduces hard label-based constraints at both the subpixel and coarse pixel scales. These constraints aim to increase the accuracy of SPM by guiding the optimization process fully toward obtaining hard classification maps at the subpixel level. Experimental evaluations against benchmark methods demonstrated the effectiveness of the H-HNN. The findings reveal that the H-HNN method is a general and robust alternative to the HNN, which can increase the overall accuracy of the SPM results by about 1%. In addition, the H-HNN can effectively reduce the uncertainties by predicting more accurate hard class labels and coarse proportions (with root-mean-square error (RMSE) of the coarse proportions decreased by about 0.015).

Semantic segmentation of remotely sensed images plays an important role in land resource management, yield estimation, and economic assessment. U-Net, a deep encoder–decoder architecture, has been used frequently for image segmentation with high accuracy. In this letter, we incorporate multiscale features generated by different layers of U-Net and design a multiscale skip connected and asymmetric-convolution-based U-Net (MACU-Net), for segmentation using fine-resolution remotely sensed images. Our design has the following advantages: (1) the multiscale skip connections combine and realign semantic features contained in both low-level and high-level feature maps; (2) the asymmetric convolution block strengthens the feature representation and feature extraction capability of a standard convolution layer. Experiments conducted on two remotely sensed data sets captured by different satellite sensors demonstrate that the proposed MACU-Net transcends the U-Net, U-Netpyramid pooling layers (PPL), U-Net 3+, among other benchmark approaches. Code is available athttps://github.com/lironui/MACU-Net.

Large-scale (large-area), fine spatial resolution satellite sensor images are valuable data sources for Earth observation while not yet fully exploited by research communities for practical applications. Often, such images exhibit highly complex geometrical structures and spatial patterns, and distinctive characteristics of multiple land-use categories may appear at the same region. Autonomous information extraction from these images is essential in the field of pattern recognition within remote sensing, but this task is extremely challenging due to the spectral and spatial complexity captured in satellite sensor imagery. In this research, a semi-supervised deep rule-based approach for satellite sensor image analysis (SeRBIA) is proposed, where large-scale satellite sensor images are analysed autonomously and classified into detailed land-use categories. Using an ensemble feature descriptor derived from pre-trained AlexNet and VGG-VD-16 models, SeRBIA is capable of learning continuously from both labelled and unlabelled images through self-adaptation without human involvement or intervention. Extensive numerical experiments were conducted on both benchmark datasets and real-world satellite sensor images to comprehensively test the validity and effectiveness of the proposed method. The novel information mining technique developed here can be applied to analyse large-scale satellite sensor images with high accuracy and interpretability, across a wide range of real-world applications.

Point cloud data acquired using terrestrial laser scanning (TLS) often need to be semantically segmented to support many applications. To this end, various point-, voxel-, and image-based methods have been developed. For large-scale point cloud data, the former two types of methods often require extensive computational effort. In contrast, image-based methods are favorable from the perspective of computational efficiency. However, existing image-based methods are highly dependent on RGB information and do not provide an effective means of representing and utilizing the local geometric characteristics of point cloud data in images. This not only limits the overall segmentation accuracy but also prohibits their application to situations where the RGB information is absent. To overcome such issues, this research proposes a novel image enhancement method to reveal the local geometric characteristics in images derived by the projection of the point cloud coordinates. Based on this method, various feature channel combinations were investigated experimentally. It was found that the new combination$IZ_{e}D_{e}$(i.e., intensity, enhanced$Z$-coordinate, and enhanced range images) outperformed the conventional$I$RGB and$I$RGB$D$channel combinations. As such, the approach can be used to replace the RGB channels for semantic segmentation. Using this new combination and the pretrained HR-EHNet considered, a mean intersection over union (mIoU) of 74.2% and an overall accuracy (OA) of 92.1% were achieved on the Semantic3D benchmark, which sets a new state of the art (SOTA) for the semantic segmentation accuracy of image-based methods.

Semantic segmentation of remote sensing images plays an important role in a wide range of applications, including land resource management, biosphere monitoring, and urban planning. Although the accuracy of semantic segmentation in remote sensing images has been increased significantly by deep convolutional neural networks, several limitations exist in standard models. First, for encoder–decoder architectures such as U-Net, the utilization of multiscale features causes the underuse of information, where low-level features and high-level features are concatenated directly without any refinement. Second, long-range dependencies of feature maps are insufficiently explored, resulting in suboptimal feature representations associated with each semantic class. Third, even though the dot-product attention mechanism has been introduced and utilized in semantic segmentation to model long-range dependencies, the large time and space demands of attention impede the actual usage of attention in application scenarios with large-scale input. This article proposed a multiattention network (MANet) to address these issues by extracting contextual dependencies through multiple efficient attention modules. A novel attention mechanism of kernel attention with linear complexity is proposed to alleviate the large computational demand in attention. Based on kernel attention and channel attention, we integrate local feature maps extracted by ResNet-50 with their corresponding global dependencies and reweight interdependent channel maps adaptively. Numerical experiments on two large-scale fine-resolution remote sensing datasets demonstrate the superior performance of the proposed MANet. Code is available athttps://github.com/lironui/Multi-Attention-Network.

Spatiotemporal fusion is a technique applied to create images with both fine spatial and temporal resolutions by blending images with different spatial and temporal resolutions. Spatial unmixing (SU) is a widely used approach for spatiotemporal fusion, which requires only the minimum number of input images. However, ignorance of spatial variation in land cover between pixels is a common issue in existing SU methods. For example, all coarse neighbors in a local window are treated equally in the unmixing model, which is inappropriate. Moreover, the determination of the appropriate number of clusters in the known fine spatial resolution image remains a challenge. In this article, a geographically weighted SU (SU-GW) method was proposed to address the spatial variation in land cover and increase the accuracy of spatiotemporal fusion. SU-GW is a general model suitable for any SU method. Specifically, the existing regularized version and soft classification-based version were extended with the proposed geographically weighted scheme, producing 24 versions (i.e., 12 existing versions were extended to 12 corresponding geographically weighted versions) for SU. Furthermore, the cluster validity index of Xie and Beni (XB) was introduced to determine automatically the number of clusters. A systematic comparison between the experimental results of the 24 versions indicated that SU-GW was effective in increasing the prediction accuracy. Importantly, all 12 existing methods were enhanced by integrating the SU-GW scheme. Moreover, the identified most accurate SU-GW enhanced version was demonstrated to outperform two prevailing spatiotemporal fusion approaches in a benchmark comparison. Therefore, it can be concluded that SU-GW provides a general solution for enhancing spatiotemporal fusion, which can be used to update existing methods and future potential versions.

Mixed pixels are a ubiquitous problem in remote sensing images. Spectral unmixing has been used widely for mixed pixel analysis. However, up to now, most spectral unmixing methods require endmembers and cannot consider fully intraclass spectral variation. The recently proposed spatiotemporal spectral unmixing (STSU) method copes with the aforementioned problems through exploitation of the available temporal information. However, this method requires coarse-to-fine spatial image pairs both before and after the prediction time and is, thus, not suitable for important real-time applications (i.e., where the fine spatial resolution data after the prediction time are unknown). In this article, we proposed a real-time STSU (RSTSU) method for real-time monitoring. RSTSU requires only a single coarse-to-fine spatial resolution image pair before, and temporally closest to, the prediction time, coupled with the coarse image at the prediction time, to extract samples automatically to train a learning model. By fully incorporating the multiscale spatiotemporal information, the RSTSU method inherits the key advantages of STSU; it does not need endmembers and can account for intraclass spectral variation. More importantly, RSTSU is suitable for real-time analysis and, thus, facilitates the timely monitoring of land cover changes. The effectiveness of the method was validated by experiments on four Moderate Resolution Imaging Spectroradiometer (MODIS) datasets. RSTSU utilizes and enriches the theory underpinning the advanced STSU method and enhances greatly the applicability of spectral unmixing for time-series data.

Deep learning methods, especially convolutional neural network (CNN)-based methods, have shown promising performance for hyperspectral image (HSI) change detection (CD). It is acknowledged widely that different spectral channels and spatial locations in input image patches may contribute differently to CD. However, they are treated equally in existing CNN-based approaches. To increase the accuracy of HSI CD, we propose an end-to-end Siamese CNN (SiamNet) with a spectral–spatial-wise attention (SSA-SiamNet) mechanism. The proposed SSA-SiamNet method can emphasize informative channels and locations and suppress less informative ones to refine the spectral–spatial features adaptively. Moreover, in the network training phase, the weighted contrastive loss function is used for more reliable separation of changed and unchanged pixels and to accelerate the convergence of the network. SSA-SiamNet was validated using four groups of bitemporal HSIs. The accuracy of CD using the SSA-SiamNet was found to be consistently greater than for ten benchmark methods.

Subpixel mapping (SPM) is a technique to tackle the mixed-pixel problem and produces land cover and land use (LCLU) maps at a finer spatial resolution than the original coarse data. However, uncertainty exists unavoidably in SPM, which is an ill-posed downscaling problem. Spatio-temporal SPM methods have been proposed to deal with this uncertainty, but current methods fail to explore fully the information in the time-series images, especially more rapid changes over a short-time interval. In this article, a fast and slow changes constrained spatio-temporal subpixel mapping (FSSTSPM) method is proposed to account for fast LCLU changes over a short time interval and slow changes over a long time interval. Both fast and slow changes-based temporal constraints are proposed and incorporated simultaneously into the FSSTSPM to increase the accuracy of SPM. The proposed FSSTSPM method was validated using two synthetic datasets with various proportion errors. It was also applied to oil-spill mapping using a real PlanetScope-Sentinel-2 dataset and Amazon deforestation mapping using a real Landsat-Moderate Resolution Imaging Spectroradiometer (MODIS) dataset. The results demonstrate the superiority of FSSTSPM. Moreover, the advantage of FSSTSPM is more obvious with an increase in proportion errors. The concepts of the fast and slow changes, together with the derived temporal constraints, provide a new insight to enhance SPM by taking fuller advantage of the temporal information in the available time-series images.

The scan-line corrector (SLC) of the Landsat 7 ETM+ failed permanently in 2003, resulting in about 22% unscanned gap pixels in the SLC-off images, affecting greatly the utility of the ETM+ data. To address this issue, we propose a spatial-spectral radial basis function (SSRBF)-based interpolation method to fill gaps in SLC-off images. Different from the conventional spatial-only radial basis function (RBF) that has been widely used in other domains, SSRBF also integrates a spectral RBF to increase the accuracy of gap filling. Concurrently, global linear histogram matching is applied to alleviate the impact of potentially large differences between the known and SLC-off images in feature space, which is demonstrated mathematically in this article. SSRBF fully exploits information in the data themselves and is user-friendly. The experimental results on five groups of data sets covering different heterogeneous regions show that the proposed SSRBF method is an effective solution to gap filling, and it can produce more accurate results than six popular benchmark methods.

Optical earth observation satellite sensors often provide a coarse spatial resolution (CR) multispectral (MS) image together with a fine spatial resolution (FR) panchromatic (PAN) image. Pansharpening is a technique applied to such satellite sensor images to generate an FR MS image by injecting spatial detail taken from the FR PAN image while simultaneously preserving the spectral information of MS image. Pansharpening methods are mostly applied on a per-pixel basis and use the PAN image to extract spatial detail. However, many land cover objects in FR satellite sensor images are not illustrated as independent pixels, but as many spatially aggregated pixels that contain important semantic information. In this article, an object-based pansharpening approach, termed object-based area-to-point regression kriging (OATPRK), is proposed. OATPRK aims to fuse the MS and PAN images at the object-based scale and, thus, takes advantage of both the unified spectral information within the CR MS images and the spatial detail of the FR PAN image. OATPRK is composed of three stages: image segmentation, object-based regression, and residual downscaling. Three data sets acquired from IKONOS and Worldview-2 and 11 benchmark pansharpening algorithms were used to provide a comprehensive assessment of the proposed OATPRK approach. In both the synthetic and real experiments, OATPRK produced the most superior pan-sharpened results in terms of visual and quantitative assessment. OATPRK is a new conceptual method that advances the pixel-level geostatistical pansharpening approach to the object level and provides more accurate pan-sharpened MS images.

This paper proposes a new classification method for spatial data by adjusting prior class probabilities according to local spatial patterns. First, the proposed method uses a classical statistical classifier to model training data. Second, the prior class probabilities are estimated according to the local spatial pattern and the classifier for each unseen object is adapted using the estimated prior probability. Finally, each unseen object is classified using its adapted classifier. Because the new method can be coupled with both generative and discriminant statistical classifiers, it performs generally more accurately than other methods for a variety of different spatial datasets. Experimental results show that this method has a lower prediction error than statistical classifiers that take no spatial information into account. Moreover, in the experiments, the new method also outperforms spatial auto-logistic regression and Markov random field-based methods when an appropriate estimate of local prior class distribution is used.

Spatial downscaling is an ill-posed, inverse problem, and information loss (IL) inevitably exists in the predictions produced by any downscaling technique. The recently popularized area-to-point kriging (ATPK)-based downscaling approach can account for the size of support and the point spread function (PSF) of the sensor, and moreover, it has the appealing advantage of the perfect coherence property. In this article, based on the advantages of ATPK and the conceptualization of IL, an IL-guided image fusion (ILGIF) approach is proposed. ILGIF uses the fine spatial resolution images acquired in other wavelengths to predict the IL in ATPK predictions based on the geographically weighted regression (GWR) model, which accounts for the spatial variation in land cover. ILGIF inherits all the advantages of ATPK, and its prediction has perfect coherence with the original coarse spatial resolution data which can be demonstrated mathematically. ILGIF was validated using two data sets and was shown in each case to predict downscaled images more accurately than the compared benchmark methods.

Based on the measured data of 12 meteorological stations in western Jilin province from 1961 to 2017, the prediction factors of NCEP reanalysis data were selected through correlation analysis. Combined with the global climate model HadCM3 data in the two scenarios of A2 and B2, the Statistical DownScaling Model (SDSM) for western Jilin province was established. By means of SDSM, this study simulated the changes of temperature, precipitation and eight climate extreme indices (SU25, FD0, CDD, CWD, TXn, TNn, TXx and TNx in western Jilin province in the four periods (2030s, 2050s, 2070s and 2090s). In the two scenarios of A2 and B2, the interannual temperature increases in western Jilin province would be 1-5 °C and 1.5-4 °C respectively. In the A2 scenario, SU25 would increase by 3-9 days and FD0 would decrease by 10-18 days. In the B2 scenario, SU25 would increase by 3-7 days and FD0 would decrease by 9-14 days. Under the both A2 and B2 scenarios, six indices (SU25, CWD, TNn, TXn, TNx and TXx) would increase obviously, while the other two indices (CDD and FD0) would decrease. Under the scenario B2, the increase of eight climate extreme indices in western Jilin province would be less than those under the scenario A2.

In this letter, we propose a new approach for remote sensing scene classification by creating an ensemble of the recently introduced massively parallel deep (fuzzy) rule-based (DRB) classifiers trained with different levels of spatial information separately. Each DRB classifier consists of a massively parallel set of human-interpretable, transparent zero-order fuzzy IF...THEN... rules with a prototype-based nature. The DRB classifier can self-organize “from scratch” and self-evolve its structure. By employing the pretrained deep convolution neural network as the feature descriptor, the proposed DRB ensemble is able to exhibit human-level performance through a transparent and parallelizable training process. Numerical examples using benchmark data set demonstrate the superior accuracy of the proposed approach together with human-interpretable fuzzy rules autonomously generated by the DRB classifier.

Soil moisture (SM) plays an important role in the land surface energy balance and water cycle. Microwave remote sensing has been applied widely to estimate SM. However, the application of such data is generally restricted because of their coarse spatial resolution. Downscaling methods have been applied to predict fine-resolution SM from original data with coarse spatial resolution. Commonly, SM is highly spatially variable and, consequently, such local spatial heterogeneity should be considered in a downscaling process. Here, a hybrid geostatistical approach, which integrates geographically weighted regression and area-to-area kriging, is proposed for downscaling microwave SM products. The proposed geographically weighted area-to-area regression kriging (GWATARK) method combines fine-spatial-resolution optical remote sensing data and coarse-spatial-resolution passive microwave remote sensing data, because the combination of both information sources has great potential for mapping fine-spatial-resolution near-surface SM. The GWATARK method was evaluated by producing downscaled SM at 1-km resolution from the 25-km-resolution daily AMSR-2 SM product. Comparison of the downscaled predictions from the GWATARK method and two benchmark methods on three sets of covariates with in situ observations showed that the GWATARK method is more accurate than the two benchmarks. On average, the root-mean-square error value decreased by 20%. The use of additional covariates further increased the accuracy of the downscaled predictions, particularly when using topography-corrected land surface temperature and vegetation-temperature condition index covariates.

Recent advances in computer vision and pattern recognition have demonstrated the superiority of deep neural networks using spatial feature representation, such as convolutional neural networks (CNNs), for image classification. However, any classifier, regardless of its model structure (deep or shallow), involves prediction uncertainty when classifying spatially and spectrally complicated very fine spatial resolution (VFSR) imagery. We propose here to characterize the uncertainty distribution of CNN classification and integrate it into a regional decision fusion to increase classification accuracy. Specifically, a variable precision rough set (VPRS) model is proposed to quantify the uncertainty within CNN classifications of VFSR imagery and partition this uncertainty into positive regions (correct classifications) and nonpositive regions (uncertain or incorrect classifications). Those “more correct” areas were trusted by the CNN, whereas the uncertain areas were rectified by a multilayer perceptron (MLP)-based Markov random field (MLP-MRF) classifier to provide crisp and accurate boundary delineation. The proposed MRF-CNN fusion decision strategy exploited the complementary characteristics of the two classifiers based on VPRS uncertainty description and classification integration. The effectiveness of the MRF-CNN method was tested in both urban and rural areas of southern England as well as semantic labeling data sets. The MRF-CNN consistently outperformed the benchmark MLP, support vector machine, MLP-MRF, CNN, and the baseline methods. This paper provides a regional decision fusion framework within which to gain the advantages of model-based CNN, while overcoming the problem of losing effective resolution and uncertain prediction at object boundaries, which is especially pertinent for complex VFSR image classification.

This research describes the seasonal phenological pattern of Africa's vegetation and its recent trends using MODIS EVI time-series data with a relatively fine spatial resolution of 500 m and a long temporal range of 15 years (2001-2015). The objectives were to measure the vegetation phenology of the major land cover types and determine the temporal trends across the geographical sub-regions of Africa. An improved representation of the land surface phenology (LSP) of Africa is provided, revealing which land cover types and regions have undergone significant changes in phenology over the period 2001-2015. Recommendations are given for future studies needed to determine and distinguish all the drivers of vegetation phenology.

Sentinel-2 is a wide-swath and fine spatial resolution satellite imaging mission designed for data continuity and enhancement of the Landsat and other missions. The Sentinel-2 data are freely available at the global scale, and have similar wavelengths and the same geographic coordinate system as the Landsat data, which provides an excellent opportunity to fuse these two types of satellite sensor data together. In this paper, a new approach is presented for the fusion of Landsat 8 Operational Land Imager and Sentinel-2 Multispectral Imager data to coordinate their spatial resolutions for continuous global monitoring. The 30 m spatial resolution Landsat 8 bands are downscaled to 10 m using available 10 m Sentinel-2 bands. To account for the land-cover/land-use (LCLU) changes that may have occurred between the Landsat 8 and Sentinel-2 images, the Landsat 8 panchromatic (PAN) band was also incorporated in the fusion process. The experimental results showed that the proposed approach is effective for fusing Landsat 8 with Sentinel-2 data, and the use of the PAN band can decrease the errors introduced by LCLU changes. By fusion of Landsat 8 and Sentinel-2 data, more frequent observations can be produced for continuous monitoring (this is particularly valuable for areas that can be covered easily by clouds, thereby, contaminating some Landsat or Sentinel-2 observations), and the observations are at a consistent fine spatial resolution of 10 m. The products have great potential for timely monitoring of rapid changes.

Forest mapping from satellite sensor imagery provides important information for the timely monitoring of forest growth and deforestation, bioenergy potential assessment, and modeling of carbon flux, among others. Due to the daily global revisit rate and wide swath width, MODerate-resolution Imaging Spectroradiometer (MODIS) images are used commonly for satellite-derived forest mapping at both regional and global scales. However, the spatial resolution of MODIS images is too coarse to observe fine spatial variation in forest cover. The last few decades have seen the production of several fine-spatial-resolution satellite-derived global forest cover maps, such as Hansen’s global tree canopy cover map of 2000, which includes abundant spectral, temporal, and spatial prior information about forest cover at a fine spatial resolution. In this paper, a novel learning-based spatial–temporal superresolution mapping approach is proposed to integrate both current MODIS images and prior maps of Hansen’s tree canopy cover, to map present forest cover with a fine spatial resolution. The novel approach is composed of three main stages: 1) automatic generation of 240-m forest proportion images from both 240- and 480-m MODIS images using a nonlinear learning-based spectral unmixing method; 2) downscaling the 240-m forest proportion images to 30 m to predict the class possibilities at the subpixel scale using a temporal-example learning-based downscaling method; and 3) final production of the fine-spatial-resolution forest map by solving a regularization-based optimization problem. The novel approach produced more accurate fine-spatial-resolution forest cover maps in terms of both visual and quantitative evaluation than traditional pixel-based classification and the latest subpixel based superresolution mapping methods. The results show the great efficiency and potential of the novel approach for producing fine-spatial-resolution forest maps from MODIS images.

The formation of an anisotropic landscape is influenced by natural and/or human processes, which can then be inferred on the basis of geometric indices. In this study, two minimal bounding rectangles in consideration of the principles of mechanics (i.e. minimal width bounding (MWB) box and moment bounding (MB) box) were introduced. Based on these boxes, four novel shape indices, namely MBLW (the length-to-width ratio of MB box), PAMBA (area ratio between patch and MB box), PPMBP (perimeter ratio between patch and MB box) and ODI (orientation difference index between MB and MWB boxes), were introduced to capture multiple aspects of landscape features including patch elongation, patch compactness, patch roughness and patch symmetry. Landscape pattern was, thus, quantified by considering both patch directionality and patch shape simultaneously, which is especially suitable for anisotropic landscape analysis. The effectiveness of the new indices were tested with real landscape data consisting of three kinds of saline soil patches (i.e. the elongated shaped slightly saline soil class, the circular or half-moon shaped moderately saline soil, and the large and complex severely saline soil patches). The resulting classification was found to be more accurate and robust than that based on traditional shape complexity indices.

The aim of this study was to examine the relationship between the anisotropy direction of exposed gravel bed and flow direction. Previous studies have shown that the anisotropy direction of a gravel bed surface can be visually determined in the elliptical contours of 2-D variogram surface (2DVS). In this letter, airborne laser scanning (ALS) point clouds were acquired at a gravel bed, and the whole data set was divided into a series of 6 m × 6 m subsets. To estimate the direction of anisotropy, we proposed an ellipse-fitting-based automatic procedure with consideration given to the grain size characteristic d50to estimate the primary axis of anisotropy [hereafter referred to as the primary continuity direction (PCD)] in the 2DVS. The ALS-derived PCDs were compared to the flow directions (for both high and low flow) derived from hydrodynamic model simulation. Comparison of ALS-derived PCDs and simulated flow directions suggested that ALS-derived PCDs could be used to infer flow direction at different flow rates. Furthermore, we found that the ALS-derived PCDs estimated from any elliptical contour of the 2DVS exhibited a similar orientation when the contours of the 2DVS reveal the clear anisotropic structure, demonstrating the robustness of the technique.

Land cover/land use (LCLU) information extraction from multitemporal sequences of remote sensing imagery is becoming increasingly important. Mixed pixels are a common problem in Landsat and MODIS images that are used widely for LCLU monitoring. Recently developed subpixel mapping (SPM) techniques can extract LCLU information at the subpixel level by dividing mixed pixels into subpixels to which hard classes are then allocated. However, SPM has rarely been studied for time-series images (TSIs). In this paper, a spatiotemporal SPM approach was proposed for SPM of TSIs. In contrast to conventional spatial dependence-based SPM methods, the proposed approach considers simultaneously spatial and temporal dependences, with the former considering the correlation of subpixel classes within each image and the latter considering the correlation of subpixel classes between images in a temporal sequence. The proposed approach was developed assuming the availability of one fine spatial resolution map which exists among the TSIs. The SPM of TSIs is formulated as a constrained optimization problem. Under the coherence constraint imposed by the coarse LCLU proportions, the objective is to maximize the spatiotemporal dependence, which is defined by blending both spatial and temporal dependences. Experiments on three data sets showed that the proposed approach can provide more accurate subpixel resolution TSIs than conventional SPM methods. The SPM results obtained from the TSIs provide an excellent opportunity for LCLU dynamic monitoring and change detection at a finer spatial resolution than the available coarse spatial resolution TSIs.

The availability of the panchromatic (PAN) band in remote sensing images gives birth to so-called image fusion techniques for increasing the spatial resolution of images to that of the PAN band. The spatial resolution of such spatially sharpened images, such as for the MODIS and Landsat sensors, however, may not be sufficient to provide the required detailed land-cover/land-use information. This paper proposes an area-to-point regression kriging (ATPRK)-based geostatistical solution to increase the spatial resolution of remote sensing images beyond that of any input images, including the PAN band. The new approach is a two-stage approach, including covariate downscaling and ATPRK-based image fusion. The new approach treats the PAN band as the covariate and takes advantages of its textural information. It explicitly accounts for the size of support, spatial correlation, and the point spread function of the sensor and has the characteristic of perfect coherence with the original coarse data. Moreover, the new downscaling approach can be extended readily by incorporating other ancillary information. The proposed approach was examined using both Landsat and MODIS images. The results show that it can produce more accurate sharpened images than four benchmark approaches.

Terrestrial laser scanning (TLS) has become a popular tool for acquiring source data points which can be used to construct digital elevation models (DEMs) for a wide number of applications. A TLS point cloud often has a very fine spatial resolution, which can represent well the spatial variation of a terrain surface. However, the uncertainty in DEMs created from this relatively new type of source data is not well understood, which forms the focus of this letter. TLS survey data representing four terrain surfaces of different characteristics were used to explore the effects of surface complexity and typical TLS data density (in terms of data point spacing) on DEM accuracy. The spatial variation in TLS data can be decomposed into parts corresponding to the signal of spatial variation (of terrain surfaces) and noise due to measurement error. We found a linear relation between the DEM error and the typical TLS data spacings considered (30-100 mm) which arises as a function of the interpolation error, and a constant contribution from the propagated data noise. This letter quantifies these components for each of the four surfaces considered and shows that, for the interpolation method considered here, higher density sampling would not be beneficial.

The kernel function plays an important role in machine learning methods such as the support vector machine. In this letter, a new kernel framework is developed for hyperspectral image classification. In contrast to existing composite kernels constructed via a linearly weighted combination, the multiple-mapping kernel proposed in this letter is obtained through repeated nonlinear mappings. Experiments indicate that the proposed multiple-mapping kernel framework (MMKF) is effective for hyperspectral image classification. Compared to the single kernel methods, the MMKF tends to be more advantageous in terms of classification accuracy, particularly for the situation with a small-size training set.

Indicator cokriging (ICK) has been shown to be an effective subpixel mapping (SPM) algorithm. It is noniterative and involves few parameters. The original ICK-based SPM method, however, requires the semivariogram of land cover classes from prior information, usually in the form of fine spatial resolution training images. In reality, training images are not always available, or laborious work is needed to acquire them. This paper aims to seek spatial structure information for ICK when such prior land cover information is not obtainable. Specifically, the fine spatial resolution semivariogram of each class is estimated by the deconvolution process, taking the coarse spatial resolution semivariogram extracted from the class proportion image as input. The obtained fine spatial resolution semivariogram is then used to estimate class occurrence probability at each subpixel with the ICK method. Experiments demonstrated the feasibility of the proposed ICK with the deconvolution approach. It obtains comparable SPM accuracy to ICK that requires semivariogram estimated from fine spatial resolution training images. The proposed method extends ICK to cases where the prior spatial structure information is unavailable.

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.

Digital elevation models (DEMs) have been widely used for a range of applications and form the basis of many GIS-related tasks. An essential aspect of a DEM is its accuracy, which depends on a variety of factors, such as source data quality, interpolation methods, data sampling density and the surface topographical characteristics. In recent years, point measurements acquired directly from land surveying such as differential global positioning system and light detection and ranging have become increasingly popular. These topographical data points can be used as the source data for the creation of DEMs at a local or regional scale. The errors in point measurements can be estimated in some cases. The focus of this article is on how the errors in the source data propagate into DEMs. The interpolation method considered is a triangulated irregular network (TIN) with linear interpolation. Both horizontal and vertical errors in source data points are considered in this study. An analytical method is derived for the error propagation into any particular point of interest within a TIN model. The solution is validated using Monte Carlo simulations and survey data obtained from a terrestrial laser scanner.

Cork oak is one of the most valuable natural forest genera in the Mediterranean basin. Modelling cork oak growth has been a challenge for foresters in recent years because of strong site and genetic influences, below-ground competition, management regimes and age effects. Because cork productivity is related to forest height, which is, in turn, related directly with site characteristics, an increase in the accuracy of height prediction implies improved productivity estimation. A Bayesian maximum entropy (BME) geostatistical model was applied to characterize the space–time pattern of height of young cork oak in a forest stand from central Sardinia in the years 2000, 2002, 2003, 2006 and 2008. Cork oak height maps were produced for each of the 5 years. The main goals were to analyse and interpret through time (i) the changes in spatial correlation and (ii) the changes in spatial distribution of cork oak height. The plantation was characterized by an increasing spatial dependence through time, whereas the temporal range was 2 years. Cork oak height was significantly correlated with wind speed (reduced by a neighbouring forest) in all the years implying a single trend. The correlations were larger for 2006 and 2008 than for previous years. Three other environmental variables (shade, elevation and slope) were less significant and their influence restricted to 2 years only. This research has several implications for the management of cork oak in the young phase.

Phenological information can be provided globally using remote sensing based time-series vegetation indices. Basic differences in the data and methods used can yield different results. This study analysed such differences in the phenological information, mainly onset of greenness (OG), estimated using the Enhanced Vegetation Index (EVI) from Moderate Resolution Imaging Spectroradiometer (MODIS) data and the Terrestrial Chlorophyll Index (MTCI) from Medium Resolution Imaging Spectrometer (MERIS) data. The two datasets were processed independently using different techniques to provide weekly estimates. Differences in the OG results were analysed for two years (2003 & 2006) and at four levels: a) full study area, b) within land cover classes, c) within core zones of each class and d) at the edge zones of each class. It was found that the trend of OG estimated from MODIS and MERIS were spatially similar, although not the same. From 15 Biome classes found in the study area the classes with the greatest differences were evergreen needle leaf, mixed forest and cropland. The differences were mainly due to the characteristic nature of the indices and also, to some extent, due to false internal flags in the algorithms.

The main aim of this paper is to show the implementation and application of downscaling cokriging for super-resolution image mapping. By super-resolution, we mean increasing the spatial resolution of satellite sensor images where the pixel size to be predicted is smaller than the pixel size of the empirical image with the finest spatial resolution. It is assumed that coregistered images with different spatial and spectral resolutions of the same scene are available. The main advantages of cokriging are that it takes into account the correlation and cross correlation of images, it accounts for the different supports (i.e., pixel sizes), it can explicitly take into account the point spread function of the sensor, and it has the property of prediction coherence. In addition, ancillary images (topographic maps, thematic maps, etc.) as well as sparse experimental data could be included in the process. The main problem is that super-resolution cokriging requires several covariances and cross covariances, some of which are not empirically accessible (i.e., from the pixel values of the images). In the adopted solution, the fundamental concept is that of covariances and cross-covariance models with point support. Once the set of point-support models is estimated using linear systems theory, any pixel-support covariance and cross covariance can be easily obtained by regularization. We show the performance of the method using Landsat Enhanced Thematic Mapper Plus images.

This paper focuses on two common problems encountered when using Light Detection And Ranging (LiDAR) data to derive digital elevation models (DEMs). Firstly, LiDAR measurements are obtained in an irregular configuration and on a point, rather than a pixel, basis. There is usually a need to interpolate from these point data to a regular grid so it is necessary to identify the approaches that make best use of the sample data to derive the most accurate DEM possible. Secondly, raw LiDAR data contain information on above‐surface features such as vegetation and buildings. It is often the desire to (digitally) remove these features and predict the surface elevations beneath them, thereby obtaining a DEM that does not contain any above‐surface features. This paper explores the use of geostatistical approaches for prediction in this situation. The approaches used are inverse distance weighting (IDW), ordinary kriging (OK) and kriging with a trend model (KT). It is concluded that, for the case studies presented, OK offers greater accuracy of prediction than IDW while KT demonstrates benefits over OK. The absolute differences are not large, but to make the most of the high quality LiDAR data KT seems the most appropriate technique in this case.

Superresolution mapping is a set of techniques to increase the spatial resolution of a land cover map obtained by soft-classification methods. In addition to the information from the land cover proportion images, supplementary information at the subpixel level can be used to produce more detailed and accurate land cover maps. The proposed method in this research aims to use fused imagery as an additional source of information for superresolution mapping using the Hopfield neural network (HNN). Forward and inverse models were incorporated in the HNN to support a new reflectance constraint added to the energy function. The value of the function was calculated based on a linear mixture model. In addition, a new model was used to calculate the local endmember spectra for the reflectance constraint. A set of simulated images was used to test the new technique. The results suggest that fine spatial resolution fused imagery can be used as supplementary data for superresolution mapping from a coarser spatial resolution land cover proportion imagery.

Superresolution mapping is a set of techniques to obtain a subpixel map from land cover proportion images produced by soft classification. Together with the information from the land cover proportion images, supplementary information at the subpixel level can be used to produce more detailed and accurate land cover maps. This research aims to use the elevation data from light detection and ranging (lidar) as an additional source of information for superresolution mapping using the Hopfield neural network (HNN). A new height function was added to the energy function of the HNN for superresolution mapping. The value of the height function was calculated for each subpixel of a certain class based on the Gaussian distribution. A set of simulated data was used to test the new technique. The results suggest that 0.8-m spatial resolution digital surface models can be combined with optical data at 4-m spatial resolution for superresolution mapping.

Methods for mapping the shoreline at a sub-pixel level are evaluated. The most accurate predictions of shoreline location were made from an approach based on simulated annealing applied to the output of a soft classification (RMSE=2.25 m).

The combined effect of positional uncertainty of field data and pixel size on the accuracy of the empirical line method is examined and quantified. Positional uncertainty reduces accuracy, although this effect decreases as sample size\nand pixel size increases. For a pre–defined accuracy requirement, this information is used to specify the sample size required for a given pixel size.

Twenty six texture measures (derived from local statistics, grey-level co-occurrence matrix (GLCM), sum and difference histogram (SADH) and variograms) were calculated for simulated images and their ability to discriminate image texture independently of image contrast was determined. The seven texture measures able to discriminate texture independently of contrast (and therefore able to estimate biomass independently of backscatter) were entropy (derived from local statistics), contrast, entropy, correlation and chi-square (derived from the GLCM), mean of sum vector (derived from the SADH) and range (derived form the variogram). These measures were calculated for Japanese Earth Resources Satellite (JERS-1) Synthetic Aperture Radar (SAR) images and related to the biomass of regenerating forest and mature forest plots from two study areas in Brazilian Amazonia. It was hypothesised that texture (a measure of both biomass and canopy unevenness) could be related to tropical forest biomass up to and beyond the saturation of the backscatter/biomass relationship. The results showed that only GLCM derived contrast increased the correlation between backscatter and biomass. The combination of GLCM contrast with backscatter has the potential to increase the accuracy of biomass estimation over the use of backscatter alone.

In this paper, land cover was predicted from Landsat TM imagery and used to generate spatially-distributed friction coefficients. Flood inundation was then predicted using the raster-based model LISFLOOD-FP, based on friction and three different elevation models. Sensitivity of LISFLOOD-FP to spatially-distributed friction was assessed. It was found that effect of friction on the flood wave is small when compared to the underlying elevation, but is greatest during the recession phase of the hydrograph. Hydraulic models of channel and overland flow allow river discharge to be related to flood inundation extent, by simu- lating flooding based on a scenario discharge. An important boundary condition for such models is surface friction. In particular, floodplain land cover is related to friction that affects the movement of the flood wave. However, the selection of appropriate friction coefficients for hydraulic models is difficult and it is recognised that floodplain friction coefficients have considerable uncertainty and sensitivity associated with them (1). Specific problems are (i) a lack of data sources, (ii) the often coarse spatial resolution of data and (iii) the use of stationary models (e.g., use of average friction over the whole catchment). Remote sensing can provide spatially distributed estimates of land cover, allowing a more informative and accurate representation of floodplain friction in flood inundation mod- els. However, hard classification may not provide sufficiently detailed land cover data at the sub-pixel scale. For example, where Landsat Thematic Mapper (TM) (spatial resolution of 30 m) imagery is used and the flood is only several hundred metres across, hard classification may be inappropriate. There- fore, in this paper, soft classification was used to estimate friction coefficients on a continuous scale for each cell on the floodplain. The objective was then to assess the sensitivity of a flood inundation model to uncertainty in these coefficients.

Land cover class composition of remotely sensed image pixels can be estimated using soft classification techniques increasingly available in many GIS packages. However, their output provides no indication of how such classes are distributed spatially within the instantaneous field of view represented by the pixel. Techniques that attempt to provide an improved spatial representation of land cover have been developed, but not tested on the difficult task of mapping from real satellite imagery. The authors investigated the use of a Hopfield neural network technique to map the spatial distributions of classes reliably using information of pixel composition determined from soft classification previously. The approach involved designing the energy function to produce a ‘best guess’ prediction of the spatial distribution of class components in each pixel. In previous studies, the authors described the application of the technique to target identification, pattern prediction and land cover mapping at the sub-pixel scale, but only for simulated imagery. We now show how the approach can be applied to Landsat Thematic Mapper (TM) agriculture imagery to derive accurate estimates of land cover and reduce the uncertainty inherent in such imagery. The technique was applied to Landsat TM imagery of small-scale agriculture in Greece and largescale agriculture near Leicester, UK. The resultant maps provided an accurate and improved representation of the land covers studied, with RMS errors for the Landsat imagery of the order of 0.1 in the new fine resolution map recorded. The results showed that the neural network represents a simple efficient tool for mapping land cover from operational satellite sensor imagery and can deliver requisite results and improvements over traditional techniques for the GIS analysis of practical remotely sensed imagery at the sub pixel scale.

Fuzzy classification techniques have been developed recently to estimate the class composition of image pixels, but their output provides no indication of how these classes are distributed spatially within the instantaneous field of view represented by the pixel. As such, while the accuracy of land cover target identification has been improved using fuzzy classification, it remains for robust techniques that provide better spatial representation of land cover to be developed. Such techniques could provide more accurate land cover metrics for determining social or environmental policy, for example. The use of a Hopfield neural network to map the spatial distribution of classes more reliably using prior information of pixel composition determined from fuzzy classification was investigated. An approach was adopted that used the output from a fuzzy classification to constrain a Hopfield neural network formulated as an energy minimization tool. The network converges to a minimum of an energy function, defined as a goal and several constraints. Extracting the spatial distribution of target class components within each pixel was, therefore, formulated as a constraint satisfaction problem with an optimal solution determined by the minimum of the energy function. This energy minimum represents a "best guess" map of the spatial distribution of class components in each pixel. The technique was applied to both synthetic and simulated Landsat TM imagery, and the resultant maps provided an accurate and improved representation of the land covers studied, with root mean square errors (RMSEs) for Landsat imagery of the order of 0.09 pixels in the new fine resolution image recorded.

The support is a geostatistical term used to describe the size, geometry and orientation of the space on which an observation is defined. In remote sensing, the size of support is equivalent to the spatial resolution. The relation of size of support with the precision of estimating the mean of several properties is evaluated by kriging. The authors chose three examples; estimating the dry biomass of pasture on May 6, 1988, and estimating the percentage cover of clover in the pasture and its NDVI (measured using a ground-based radiometer) on Aug. 6, 1988. The modelled experimental variograms of these properties were deregularized to estimate the punctual variograms and these functions regularized to new sizes of support. The regularized variograms were then used to estimate the kriging variances attainable by sampling on a square grid. The kriging variances were plotted against grid spacing for each new size of support and the optimal sampling strategy read from the graph. In each case, there were several optimal sampling strategies, and the final choice depended on the cost of measurement. In some cases increasing the size of support was more efficient than increasing the sampling intensity.>