Land Cover Applications Research Articles

A new global hybrid map of annual herbaceous cropland at a 500 m resolution for the year 2019

Abstract The global spatial extent of croplands is a crucial input to global and regional agricultural monitoring and modeling systems. Although many new remotely-sensed products are now appearing due to recent advances in the spatial and temporal resolution of satellite sensors, there are still issues with these products that are related to the definition of cropland used and the accuracies of these maps, particularly when examined spatially. To address the needs of the agricultural monitoring community, here we have created a hybrid map of global cropland extent at a 500 m resolution by fusing two of the latest high resolution remotely-sensed cropland products: the European Space Agency’s WorldCereal and the cropland layer from the University of Maryland. We aggregated the two products to a common resolution of 500 m to produce percentage cropland and compared them spatially, calculating two kinds of disagreement: density disagreement, where the two maps differ by more than 80%, and absence-presence of cropland disagreement, where one map indicates the presence of cropland while the other does not. Based on these disagreements, we selected continuous areas of disagreement, referred to in the paper as hotspots of disagreement, for manual correction by experts using the Geo-Wiki land cover application. The hybrid map was then validated using a stratified random sample based on the disagreement layer, where the sample was visually interpreted by a different set of experts using Geo-Wiki. The results show that the hybrid product improves upon the overall accuracy statistics in the areas where the underlying cropland layer from the University of Maryland was improved with the WorldCereal product, but more importantly, it represents an improved spatially explicit cropland mask for early warning and food security assessment purposes.

Transformer-based land use and land cover classification with explainability using satellite imagery

Transformer-based models have greatly improved Land Use and Land Cover (LULC) applications. Their revolutionary ability to analyze and extract key information has greatly advanced the field. However, the high computational cost of these models presents a considerable obstacle to their practical implementation. Therefore, this study aims to strike a balance between computational cost and accuracy when employing transformer-based models for LULC analysis. We exploit transfer learning and fine-tuning strategies to optimize the resource utilization of transformer-based models. Furthermore, transparency is the core principle of our methodology to promote fairness and trust in applying LULC models across various domains, including forestry, environmental studies, and urban or rural planning. To ensure transparency, we have employed Captum, which enables us to uncover and mitigate potential biases and interpret AI-driven decisions. Our results indicate that transfer learning can potentially improve transformer-based models in satellite image classification, and strategic fine-tuning can maintain efficiency with minimal accuracy trade-offs. This research highlights the potential of Explainable AI (XAI) in Transformer-based models for achieving more efficient and transparent LULC analysis, thereby encouraging continued innovation in the field.

Estimating conservation value and natural capital value of land cover classes in the Irish National Land Cover Map and application to a case study area

Conservation science and planning, by measuring proxies of biodiversity and ecosystem services provision, aim to identify priority areas for nature conservation and ecosystem services. In Ireland, fine-scale data on ecosystems functioning and biodiversity are limited, making it challenging to map conservation value (CV) and natural capital value (NCV) accurately. We elicited expert knowledge to rank habitat classes mapped in the recently published National Land Cover Map (NLCM) (EPA and Tailte Éireann, 2023). A scoring system from 0 to 10 was used to score habitats based on their estimated provision of biodiversity (CV) and ecosystem services (NCV). As a case study, we applied this scoring system to a catchment in the south-east of Ireland (>2,000 km2) with land cover information available from the draft NLCM. The expert elicitation showed little overall difference between the scores assigned by the team and the experts invited to validate the CV and NCV scores. However, some scores were revised based on experts’ contributions. Results of the mapping exercise indicated a high correlation between monads with high CV and high NCV scores. Future work should focus on differentiating the weighting assigned to each ecosystem service associated with each land cover class. This could result in changes in the overall NCV scores assigned to each habitat (and monads). Nevertheless, the approach developed here has the potential to identify areas in the landscape that should be targeted for conservation. For reproducibility, we provide the R code for analysis at polygon scale.

Hybrid optimal joint spatial-spectral hyperspectral image classification using modified DHO-based GIF with JRKNN

ABSTRACT Hyperspectral images (HSI) have been extensively utilized in environmental protection, agriculture, and land cover applications. But, it is difficult to classify due to improper exploitation of spatial and spectral features. As a result, this paper proposes a unified guided image filtering (GIF) method that uses a joint representation model of the K-nearest neighbour algorithm (JRKNN) to extract spatial context information and preserve edges while denoising classification results. In addition, a new meta-heuristic optimization algorithm called modified deer hunting optimization (MDHO) is implemented to minimize the cost function in GIF by optimizing the regularization parameter and edge-aware weighting, which is named GIF-MDHO. Finally, the simulation results show that the GIF-MDHO with JRKNN approach performed superior as compared to the existing support vector machine (SVM) by KNN and even the KNN-GIF methods with an improved accuracy of 1.02%, precision, recall , and F1-score of 0.78%, 0.91%, and 0.74%, respectively.

Multimodal Fusion of Mobility Demand Data and Remote Sensing Imagery for Urban Land-Use and Land-Cover Mapping

This paper aims at exploring the potentiality of the multimodal fusion of remote sensing imagery with information coming from mobility demand data in the framework of land-use mapping in urban areas. After a discussion on the function of mobility demand data, a probabilistic fusion framework is developed to take advantage of remote sensing and transport data, and their joint use for urban land-use and land-cover applications in urban and surrounding areas. Two different methods are proposed within this framework, the first based on pixelwise probabilistic decision fusion and the second on the combination with a region-based multiscale Markov random field. The experimental validation is conducted on a case study associated with the city of Genoa, Italy.

WH-MAVS: A Novel Dataset and Deep Learning Benchmark for Multiple Land Use and Land Cover Applications

Over the past decade, many excellent data sharing efforts have enriched the remote sensing scene classification (SC) methods. These datasets have achieved great success in complex high-level semantic information interpretation. However, most existing datasets are collected from standard and ungeoreferenced image patches for algorithm training and evaluation. These datasets do not fit for practical applications and cannot be directly applied in further geographical study. Accordingly, we provide a large range high-resolution SC dataset with multiple time phases, called “ <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">W</b> u <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">h</b> an <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">M</b> ulti <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">a</b> pplication <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">V</b> HR <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">S</b> cene classification dataset (WH-MAVS).” It facilitates the study of SC and scene change detection (SCD) algorithms. Moreover, it can also be directly employed to perform a variety of real-life land use application tasks. To the best of our knowledge, this is the first free, publicly available, georeferenced, and annotated dataset to cover almost an entire megacity. The WH-MAVS was collected and annotated from Google Earth imagery with the same spatial resolution and uniform nonoverlapping patch size, covering the central area of Wuhan, China. The total number of scene samples is 47 137, which belong to 14 classes with 23 567 labeled patches for each time phase in 2014 and 2016, respectively. The geographic coordinates of all samples in both time phases exhibit one-to-one correspondence with 23 202 unchanged image patches of scene categories and 365 changed ones. The distribution of the number of samples in each class is highly imbalanced; moreover, there are large intraclass differences and indistinguishable interclass variances. These characteristics are closer to the real land use/land cover application tasks and introduce further challenges to the related algorithm research. In addition, we conducted benchmark experiments on SC and SCD based on the WH-MAVS dataset with widely used deep learning models. DenseNet169 was found to achieve the best performance. The overall accuracies are 91.07% and 92.09%, respectively, in the 2014 and 2016 validation sets of WH-MAVS. Furthermore, SCD obtained by DenseNet169 has a binary change detection accuracy of 89.56% and a multiple (from–to) change detection accuracy of 86.70%. Over and above the research value of the algorithm, it is also proven to have practical applications in fields such as urban planning, landscape pattern analysis, and urban dynamic monitoring and analysis.

A Refined Pyramid Scene Parsing Network for Polarimetric SAR Image Semantic Segmentation in Agricultural Areas

Polarimetric synthetic aperture radar (PolSAR) image semantic segmentation is currently of great importance for synthetic aperture radar (SAR) image interpretation, especially in agricultural applications. Several convolutional neural networks (CNNs) have been implemented for SAR image semantic segmentation in urban or land cover applications. However, existing CNNs often break one semantic area into several pieces or confuse the adjacent semantic area in a PolSAR image. To address these issues uniformly, a refined pyramid scene parsing network (PSPNet) is proposed for PolSAR image semantic segmentation in an agricultural areas. Compared to conventional PSPNet architecture, the refined PSPNet adopts a multilevel feature fusion design in its decoder to effectively exploit the features learned from its different encoder branches. Besides, a polarimetric channel attention module is incorporated into the network to capture rich polarimetric features in a PolSAR image. Furthermore, an edge-aware loss function is devised to guide the network to refine pixel-level edge information directly from semantic segmentation prediction, separating easily confused agriculture area with sharp contours. Experimental results on one airborne millimeter-wave PolSAR dataset verify that the proposed network achieves promising semantic segmentation accuracy and preferable spatial consistency.

A Semisupervised CRF Model for CNN-Based Semantic Segmentation With Sparse Ground Truth

Convolutional neural networks (CNNs) represent the new reference approach for semantic segmentation of very-high-resolution (VHR) images, due to their ability to automatically capture semantic information while learning relevant features. However, as for most supervised methods, the map accuracy depends on the quantity and quality of ground truth (GT) used to train them. The use of densely annotated data (i.e., a detailed, exhaustive, pixel-level GT) allows to obtain effective CNN models but normally implies high efforts in annotation. Such ground truth is often available in benchmark datasets on which new methods are tested, but not on real data for land-cover applications, where only sparse annotations might be sufficiently cost effective. A CNN model trained with such incomplete GT maps has the tendency to smooth object boundaries because they are never precisely delineated in the GT. To cope with those shortcomings, we propose to exploit the intermediate activation maps of the CNN and to deploy a semisupervised fully connected conditional random field (CRF). In comparison with competitors using the same sparse annotations, the proposed method is able to better fill part of the performance gap compared to a CNN trained on the densely annotated, but generally unavailable, GTs.

Spectral Temporal Information for Missing Data Reconstruction (STIMDR) of Landsat Reflectance Time Series

The number of Landsat time-series applications has grown substantially because of its approximately 50-year history and relatively high spatial resolution for observing long term changes in the Earth’s surface. However, missing observations (i.e., gaps) caused by clouds and cloud shadows, orbit and sensing geometry, and sensor issues have broadly limited the development of Landsat time-series applications. Due to the large area and temporal and spatial irregularity of time-series gaps, it is difficult to find an efficient and highly precise method to fill them. The Missing Observation Prediction based on Spectral-Temporal Metrics (MOPSTM) method has been proposed and delivered good performance in filling large-area gaps of single-date Landsat images. However, it can be less practical for a time series longer than one year due to the lack of mechanics that exclude dissimilar data in time series (e.g., different phenology or changes in land cover). To solve this problem, this study proposes a new gap-filling method, Spectral Temporal Information for Missing Data Reconstruction (STIMDR), and examines its performance in Landsat reflectance time series. Two groups of experiments, including 2000 × 2000 pixel Landsat single-date images and Landsat time series acquired from four sites (Kenya, Finland, Germany, and China), were performed to test the new method. We simulated artificial gaps to evaluate predicted pixel values with real observations. Quantitative and qualitative evaluations of gap-filled images through comparisons with other state-of-the-art methods confirmed the more robust and accurate performance of the proposed method. In addition, the proposed method was also able to fill gaps contaminated by extreme cloud cover for a period (e.g., winter in high-latitude areas). A down-stream task of random forest supervised classification through both gap-filled simulated datasets and the original valid datasets verified that STIMDR-generated products are relevant to the user community for land cover applications.

SITS-Former: A pre-trained spatio-spectral-temporal representation model for Sentinel-2 time series classification

Sentinel-2 images provide a rich source of information for a variety of land cover, vegetation, and environmental monitoring applications due to their high spectral, spatial, and temporal resolutions. Recently, deep learning-based classification of Sentinel-2 time series becomes a popular solution to vegetation classification and land cover mapping, but it often demands a large number of manually annotated labels. Improving classification performance with limited labeled data is still a challenge in many real-world remote sensing applications. To address label scarcity, we present SITS-Former (SITS stands for Satellite Image Time Series and Former stands for Transformer), a pre-trained representation model for Sentinel-2 time series classification. SITS-Former adopts a Transformer encoder as the backbone and takes time series of image patches as input to learn spatio-spectral-temporal features. According to the principles of self-supervised learning, we pre-train SITS-Former on massive unlabeled Sentinel-2 time series via a missing-data imputation proxy task. Given an incomplete time series with some patches being masked randomly, the network is asked to regress the central pixels of these masked patches based on the residual ones. By doing so, the network can capture high-level spatial and temporal dependencies from the data to learn discriminative features. After pre-training, the network can adapt the learned features to a target classification task through fine-tuning. As far as we know, this is the first study that exploits self-supervised learning for patch-based representation learning and classification of SITS. We quantitatively evaluate the quality of the learned features by transferring them on two crop classification tasks, showing that SITS-Former outperforms state-of-the-art approaches and yields a significant improvement (2.64%∼3.30% in overall accuracy) over the purely supervised model. The proposed model provides an effective tool for SITS-related applications as it greatly reduces the burden of manual labeling. The source code will be released at https://github.com/linlei1214/SITS-Former upon publication.

Recent Advances in Multi- and Hyperspectral Image Analysis

Current advancements in sensor technology bring new possibilities in multi- and hyperspectral imaging. Real-life use cases which can benefit from such imagery span across various domains, including precision agriculture, chemistry, biology, medicine, land cover applications, management of natural resources, detecting natural disasters, and more. To extract value from such highly dimensional data capturing up to hundreds of spectral bands in the electromagnetic spectrum, researchers have been developing a range of image processing and machine learning analysis pipelines to process these kind of data as efficiently as possible. To this end, multi- or hyperspectral analysis has bloomed and has become an exciting research area which can enable the faster adoption of this technology in practice, also when such algorithms are deployed in hardware-constrained and extreme execution environments; e.g., on-board imaging satellites.

Encoding Spectral-Spatial Features for Hyperspectral Image Classification in the Satellite Internet of Things System

Hyperspectral image classification is essential for satellite Internet of Things (IoT) to build a large scale land-cover surveillance system. After acquiring real-time land-cover information, the edge of the network transmits all the hyperspectral images by satellites with low-latency and high-efficiency to the cloud computing center, which are provided by satellite IoT. A gigantic amount of remote sensing data bring challenges to the storage and processing capacity of traditional satellite systems. When hyperspectral images are used in annotation of land-cover application, data dimension reduction for classifier efficiency often leads to the decrease of classifier accuracy, especially the region to be annotated consists of natural landform and artificial structure. This paper proposes encoding spectral-spatial features for hyperspectral image classification in the satellite Internet of Things system to extract features effectively, namely attribute profile stacked autoencoder (AP-SAE). Firstly, extended morphology attribute profiles EMAP is used to obtain spatial features of different attribute scales. Secondly, AP-SAE is used to extract spectral features with similar spatial attributes. In this stage the program can learn feature mappings, on which the pixels from the same land-cover class are mapped as closely as possible and the pixels from different land-cover categories are separated by a large margin. Finally, the program trains an effective classifier by using the network of the AP-SAE. Experimental results on three widely-used hyperspectral image (HSI) datasets and comprehensive comparisons with existing methods demonstrate that our proposed method can be used effectively in hyperspectral image classification.

LUCAS Copernicus 2018: Earth-observation-relevant in situ data on land cover and use throughout the European Union

Abstract. The Land Use/Cover Area frame Survey (LUCAS) is an evenly spaced in situ land cover and land use ground survey exercise that extends over the whole of the European Union. LUCAS was carried out in 2006, 2009, 2012, 2015, and 2018. A new LUCAS module specifically tailored to Earth observation (EO) was introduced in 2018: the LUCAS Copernicus module. The module surveys the land cover extent up to 51 m in four cardinal directions around a point of observation, offering in situ data compatible with the spatial resolution of high-resolution sensors. However, the use of the Copernicus module being marginal, the goal of the paper is to facilitate its uptake by the EO community. First, the paper summarizes the LUCAS Copernicus protocol to collect homogeneous land cover on a surface area of up to 0.52 ha. Secondly, it proposes a methodology to create a ready-to-use dataset for Earth observation land cover and land use applications with high-resolution satellite imagery. As a result, a total of 63 364 LUCAS points distributed over 26 level-2 land cover classes were surveyed on the ground. Using homogeneous extent information in the four cardinal directions, a polygon was delineated for each of these points. Through geospatial analysis and by semantically linking the LUCAS core and Copernicus module land cover observations, 58 426 polygons are provided with level-3 land cover (66 specific classes including crop type) and land use (38 classes) information as inherited from the LUCAS core observation. The open-access dataset supplied with this paper (https://doi.org/10.6084/m9.figshare.12382667.v4 d'Andrimont, 2020) provides a unique opportunity to train and validate decametric sensor-based products such as those obtained from the Copernicus Sentinel-1 and Sentinel-2 satellites. A follow-up of the LUCAS Copernicus module is already planned for 2022. In 2022, a simplified version of the LUCAS Copernicus module will be carried out on 150 000 LUCAS points for which in situ surveying is planned. This guarantees a continuity in the effort to find synergies between statistical in situ surveying and the need to collect in situ data relevant for Earth observation in the European Union.

A Review Paper on Monitoring Environmental Consequences of Land Cover Dynamics with The Help of Geo-informatics Technologies

Land cover dynamics is a challenging and vigorous process that associates natural and human systems that have undeviating effects on atmosphere, water and soil which lead to many environmental problems worldwide. Urbanization is one of a major land cover change that is highly correlated with many environmental problems that need emphasis. This paper aimed to review and present level and effect of land use land cover changes, urbanization, factors affecting land cover change and application of geographic information system &amp; remote sensing in monitoring land cover changes. Over the past 300 years, about 1.2 million kilometer square of forests and 5.6 million kilometer square of pasture and rangeland were replaced by other uses worldwide, while cultivated land increased by 12 million km2. In 1950, only 30 percent of the world population lived in urban settings, the fraction raised to 55% by 2018. This led to about roughly 60% of the ecosystem services are being destroyed or used in unsustainable ways worldwide. Population expansion, change of technology, high land value, corruption, lack of awareness, migration of people and political pressure are among major driving force of land cover changes. Geo-informatics technology specially GIS and Remote Sensing is found to be an excellent tool for study of land cover change that enables observation across large area of earth’s surface with low cost, better efficient and high accuracy. Therefore monitoring, analyzing and evaluation of land cover dynamics with the help of geo-informatics is decisive for improved management &amp; characterizing land cover alteration processes, and determining its environmental consequences.&#x0D; Keywords : land use; land cover change; urbanization; GIS &amp; remote sensing; environment&#x0D; Copyright (c) 2020 Geosfera Indonesia Journal and Department of Geography Education, University of Jember&#x0D; This work is licensed under a Creative Commons Attribution-Share A like 4.0 International License

Hybrid Compact Polarimetric SAR for Environmental Monitoring with the RADARSAT Constellation Mission

Canada’s successful space-based earth-observation (EO) radar program has earned widespread and expanding user acceptance following the launch of RADARSAT-1 in 1995. RADARSAT-2, launched in 2007, while providing data continuity for its predecessor’s imaging capabilities, added new polarimetric modes. Canada’s follow-up program, the RADARSAT Constellation Mission (RCM), launched in 2019, while providing continuity for its two predecessors, includes an innovative suite of polarimetric modes. In an effort to make polarimetry accessible to a wide range of operational users, RCM uses a new method called hybrid compact polarization (HCP). There are two essential elements to this approach: (1) transmit only one polarization, circular; and (2) receive two orthogonal polarizations, for which RCM uses H and V. This configuration overcomes the conventional dual and full polarimetric system limitations, which are lacking enough polarimetric information and having a small swath width, respectively. Thus, HCP data can be considered as dual-pol data, while the resulting polarimetric classifications of features in an observed scene are of comparable accuracy as those derived from the traditional fully polarimetric (FP) approach. At the same time, RCM’s HCP methodology is applicable to all imaging modes, including wide swath and ScanSAR, thus overcoming critical limitations of traditional imaging radar polarimetry for operational use. The primary image data products from an HCP radar are different from those of a traditional polarimetric radar. Because the HCP modes transmit circularly polarized signals, the data processing to extract polarimetric information requires different approaches than those used for conventional linearly polarized polarimetric data. Operational users, as well as researchers and students, are most likely to achieve disappointing results if they work with traditional polarimetric processing tools. New tools are required. Existing tutorials, older seminar notes, and reference papers are not sufficient, and if left unrevised, could succeed in discouraging further use of RCM polarimetric data. This paper is designed to provide an initial response to that need. A systematic review of studies that used HCP SAR data for environmental monitoring is also provided. Based on this review, HCP SAR data have been employed in oil spill monitoring, target detection, sea ice monitoring, agriculture, wetland classification, and other land cover applications.

Detection of Changes in Land Use/Cover in Muhayil Asir (Saudi Arabia) during 1990 – 2019, Using Remote Sensing and Geographic Information System

The research aimed to study and analyze the changes of Land Use/Cover and the vegetation patterns in Muhayil Asir Governorate during the period 1990-2019 using Remote Sensing and Geographic Information System. Imageries from Landsat Thematic Mapper (TM) and Sentinel-2 were used. The visuals underwent the pre-processing and classification analysis using the supervised classification method. The results of the accuracy assessment were 91.02% for the producer's accuracy and 89.9% for the consumer's accuracy. The results showed that the study area has four types of land cover: agricultural land, urban areas, soil (unutilized land) and rocks. The area of agricultural land has increased 18.08 km² (0.94%) during the period 1990 – 2019 and extended towards the areas of urbanization. This change coincided with increased the built areas by 65.19 km² (3.40%). The increase in agricultural and built areas occurred at the expense of unutilized lands which decreased by (-4.34%). There was no change in rock outcrops over the study period. The study concluded that the most important reasons that affected the change in the land cover pattern and applications increase in the population at a rate of 2.03% population growth and the increase in income and expansion of services, which required increasing the horizontal and vertical expansion, which would affect changing patterns of Land Use/Cover for study area. Keywords: Land use, change detection, Remote Sensing, GIS.

Assessing the effectiveness of PlanetScope synthesized panchromatic bands for spatial enhancement of Sentinel-2 data

Following the objectives of the mission, Sentinel-2 (S2) makes a significant contribution to land monitoring, climate change, emergency management and security, and related problems. To increase the spatial resolution, researchers have been challenged to obtain a synthesized panchromatic band with a fine resolution for all S2 bands. The capacity to collect satellite imagery with a short revisit time at different spatial resolutions increases the ability for more accurate data fusion applications. Our work presents an investigation of different synthesized panchromatic bands for producing high-resolution S2 data. We produced a synthesized environment from PlanetScope (PS) data to evaluate the performance of different panchromatic bands for the enhancement of S2 10- and 20-m bands. For this, imagery over three different study areas with different characteristics was chosen: Pakistan, North Macedonia, and Turkey. After the three different synthesized panchromatic bands were produced, we fused the panchromatic bands with the 10- and 20-m S2 bands with three different state-of-the-art pansharpening techniques. Experimental comparison between the three newly produced panchromatic bands indicates that all synthesized bands can enhance the spatial resolution of the original multispectral bands. Since the difference in spatial resolution may be critical for more detailed image classification, for future studies, we recommend investigation of the fused data for different land cover applications.

Synthesis of Multispectral Optical Images From SAR/Optical Multitemporal Data Using Conditional Generative Adversarial Networks

The synthesis of realistic data using deep learning techniques has greatly improved the performance of classifiers in handling incomplete data. Remote sensing applications that have profited from those techniques include translating images of different sensors, improving the image resolution and completing missing temporal or spatial data such as in cloudy optical images. In this context, this letter proposes a new deep-learning-based framework to synthesize missing or corrupted multispectral optical images using multimodal/multitemporal data. Specifically, we use conditional generative adversarial networks (cGANs) to generate the missing optical image by exploiting the correspondent synthetic aperture radar (SAR) data with a SAR-optical data from the same area at a different acquisition date. The proposed framework was evaluated in two land-cover applications over tropical regions, where cloud coverage is a major problem: crop recognition and wildfire detection. In both applications, our proposal was superior to alternative approaches tested in our experiments. In particular, our approach outperformed recent cGAN-based proposals for cloud removal, on average, by 7.7% and 8.6% in terms of overall accuracy and F1-score, respectively.

Collect Earth: An online tool for systematic reference data collection in land cover and use applications

Land cover monitoring efforts are important for resource planning and ecosystem services in many countries. Collect Earth Online (CEO) is a new, free open source and user-friendly software tool for land monitoring. It is the product of a collaborative effort between NASA, Food and Agriculture Organization of the United Nations (FAO), US Forest Service and Google. This paper provides a full overview of CEO's structure and functionality. Based on the cloud, CEO's structure supports simultaneous data entry by multiple users. No desktop installation is required and only an internet connection is required setting minimal requirements for using the software. Google Earth Engine widgets can be created for assisted plot interpretation such as image collection, time series graphs featuring indices such as Normalized Difference Vegetation Index (NDVI) and related statistics. We also provide a case study and related findings from a CEO workshop held in Myanmar.

Exploring geo-tagged photos for land cover validation with deep learning

Land cover validation plays an important role in the process of generating and distributing land cover thematic maps, which is usually implemented by high cost of sample interpretation with remotely sensed images or field survey. With an increasing availability of geo-tagged landscape photos, the automatic photo recognition methodologies, e.g., deep learning, can be effectively utilised for land cover applications. However, they have hardly been utilised in validation processes, as challenges remain in sample selection and classification for highly heterogeneous photos. This study proposed an approach to employ geo-tagged photos for land cover validation by using the deep learning technology. The approach first identified photos automatically based on the VGG-16 network. Then, samples for validation were selected and further classified by considering photos distribution and classification probabilities. The implementations were conducted for the validation of the GlobeLand30 land cover product in a heterogeneous area, western California. Experimental results represented promises in land cover validation, given that GlobeLand30 showed an overall accuracy of 83.80% with classified samples, which was close to the validation result of 80.45% based on visual interpretation. Additionally, the performances of deep learning based on ResNet-50 and AlexNet were also quantified, revealing no substantial differences in final validation results. The proposed approach ensures geo-tagged photo quality, and supports the sample classification strategy by considering photo distribution, with accuracy improvement from 72.07% to 79.33% compared with solely considering the single nearest photo. Consequently, the presented approach proves the feasibility of deep learning technology on land cover information identification of geo-tagged photos, and has a great potential to support and improve the efficiency of land cover validation.