Produce Land Cover Maps Research Articles

Assessment of fine-resolution land cover mapping products in the Changbai Mountain Range, China

Assessment of fine-resolution land cover mapping products in the Changbai Mountain Range, China

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

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

Map of Land Cover Agreement: Ensambling Existing Datasets for Large-Scale Training Data Provision

Land cover information plays a critical role in supporting sustainable development and informed decision-making. Recent advancements in satellite data accessibility, computing power, and satellite technologies have boosted large-extent high-resolution land cover mapping. However, retrieving a sufficient amount of reliable training data for the production of such land cover maps is typically a demanding task, especially using modern deep learning classification techniques that require larger training sample sizes compared to traditional machine learning methods. In view of the above, this study developed a new benchmark dataset called the Map of Land Cover Agreement (MOLCA). MOLCA was created by integrating multiple existing high-resolution land cover datasets through a consensus-based approach. Covering Sub-Saharan Africa, the Amazon, and Siberia, this dataset encompasses approximately 117 billion 10m pixels across three macro-regions. The MOLCA legend aligns with most of the global high-resolution datasets and consists of nine distinct land cover classes. Noteworthy advantages of MOLCA include a higher number of pixels as well as coverage for typically underrepresented regions in terms of training data availability. With an estimated overall accuracy of 96%, MOLCA holds great potential as a valuable resource for the production of future high-resolution land cover maps.

Scalable approach for high-resolution land cover: a case study in the Mediterranean Basin

The production of land cover maps is an everyday use of image classification applications on remote sensing. However, managing Earth observation satellite data for a large region of interest is challenging in the task of creating land cover maps. Since satellite imagery is getting more precise and extensive, Big Data techniques are becoming essential to handle the rising quantity of data. Furthermore, given the complexity of managing and analysing the data, defining a methodology that reduces the complexity of the process into different smaller steps is vital to data processing. This paper presents a Big Data methodology for creating land cover maps employing artificial intelligence algorithms. Machine Learning algorithms are contemplated for remote sensing and geodata classification, supported by explainable artificial intelligence. Furthermore, the process considers aspects related to downloading data from different satellites, Copernicus and ASTER, executing the pre-processing and processing of the data in a distributed environment, and depicting the visualisation of the result. The methodology is validated in a test case for er map of the Mediterranean Basin.

Automatic forest change detection through a bi-annual time series of satellite imagery: Toward production of an integrated land cover map

Land cover mapping is fundamental for national and international agencies to monitor forest resources. However, monitoring forest disturbances by direct comparison of these maps poses several difficulties and challenges. As a result, different methodologies have been explored to detect forest disturbances. However, most of them cannot be fully integrated with land cover map production since they require additional input data, while others are not suitable for monitoring small land parcels. This study presents a methodology that fulfils the need to integrate land cover mapping with land cover change detection. Specifically, this methodology was designed to complement the Sentinel-2-based land cover mapping used in Galicia, northwest Spain, a region characterized by small land parceling. First, two previously obtained land cover maps from 2019 and 2020 were compared to identify all the pixels with potential land cover changes using QGIS. The behavior of spectral indexes in a time series were then analyzed to identify which of the previously identified pixels correspond to forest disturbances. This step was implemented in the software R. Using the Normalized Difference Vegetation Index (NDVI) to detect different land cover changes it was obtained an overall accuracy of 82%, considering the existence of varying phenologies, diverse topographic conditions, and areas with a high level of stand fragmentation. This study could help agencies that have already developed their own land cover maps to easily advance the integration of their maps with land cover change detection, since this technique can be applied with any land cover mapping methodology based on multitemporal analysis of satellite images, without the need for additional input data.

Simultaneous Update of High-Resolution Land-Cover Mapping Attempt: Wuhan and the Surrounding Satellite Cities Cartography Using L2HNet

Land-cover mapping is important for urban planning and management, and current land-cover mapping products are unable to meet the needs of cities due to frequent land surface changes. In this study, based on the L2HNet network, we generate a high-resolution land-cover mapping product for Wuhan and its surrounding areas. In this paper, we adopt a simplified L2HNet by removing the CAS and the L2H loss module to shorten the cycle time of entire mapping process. The mapping process used ESA LandCover (2021) as LR labels and Google Maps as high-resolution remote sensing images. In the course of the experiment, we also calculate the four indicators MIoU, Oa, FWIoU, and Kappa, evaluate the accuracy of our product in predicting fine feature structure using a point-based test method, and compare it with six mainstream land-cover mapping products. The product achieves a 1m-resolution land-cover product in study areas while maintaining an over 75.21% MIoU. Oa, FWIoU and Kappa all maintain values above 85.00%, showing excellent prediction results. In quantitative analysis, compared to ESA LandCover(2021), the L2HNet product has a significant improvement in mapping accuracy for build-up and permanent water, including an exciting 21.08% improvement in permanent water accuracy and an amazing improvement in build-up. The comparison with mainstream products also shows the credibility and practicality of the product. The end result of this research fills a gap in Wuhan and its surrounding areas' 1m-resolution land-cover mapping product. While significantly improving the product's resolution, L2HNet makes time- and labor-saving periodic mapping a reality.

Temporal-Domain Adaptation for Satellite Image Time-Series Land-Cover Mapping With Adversarial Learning and Spatially Aware Self-Training

Nowadays, Satellite Image Time Series (SITS) are commonly employed to derive land cover maps (LCM) to support decision makers in a variety of land management applications. In the most general workflow, the production of LCM strongly relies on available ground truth data to train supervised machine learning models. Unfortunately, this data is not always available due to time-consuming and costly field campaigns. In this scenario, the possibility to transfer a model learnt on a particular year ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">source domain</i> ) to a successive period of time ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">target domain</i> ), over the same study area, can save time and money. Such a kind of model transfer is challenging due to different acquisition conditions affecting each time period thus, resulting in possible distribution shifts between <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">source</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">target</i> domains. In the general field of machine learning, Unsupervised Domain Adaptation (UDA) approaches are well suited to cope with the learning of models under distribution shifts between <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">source</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">target</i> domains. While widely explored in the general computer vision field, they are still under investigated for SITS-based land cover mapping, especially for the temporal transfer scenario. With the aim to cope with this scenario in the context of SITS-based land cover mapping, here we propose Spatially Aligned Domain-Adversarial Neural Network, a framework that combines both adversarial learning and self-training to transfer a classification model from a time period (year) to a successive one on a specific study area. Experimental assessment on a study area located in Burkina Faso characterized by challenging operational constraints demonstrates the significance of our proposal. The obtained results have shown that our proposal outperforms all the UDA competing methods by 7 to 12 points of F1-score across three different transfer tasks.

The degradation of forest areas in Morocco: Case of Benslimane province

This article aims to shed light on the process of known degradation of the forest area of Benslimane province during the period 1990–2020 and to specify the most important human causes which contributed to it (quarries, extension of the built-up area, the impact of agricultural activities, grazing and collection of firewood), by using remote sensing techniques (spatial images for the years 1990–2000–2010–2020) to produce Land Cover maps. The following satellite images were used, Landsat 5 TM, Landsat 7 ETM+ and Landsat 8 OLI, with a spatial precision of 30 m, the Semi-Automatic Classification Plugin (SCP) in QGIS was used for atmospheric correction, and the Spectral Angle Mapping algorithm for the images’ classification. The rating evaluation of the Kappa coefficient shows the following ratios for the years 1990–2000–2010–2020 respectively ; 0.89–0.90–0.90–0.93. The results showed that the forest area of Benslimane province has declined by 11.4% or about 6,027.7 ha between 1990–2020 at the rate of 200 ha/year, which has been turned into matorral land or bare land. This forest also lost 35.2% of its vegetative density and has become much sparser, while the original grazing areas surrounding it have been reduced by 50.4%. Moreover, the area of quarries increased by 1,097.4%, the percentage of built-up area increased by 328.2%, and the agricultural area expanded by 32.7%. These results can be used as preliminary data for future studies and can help policymakers focus on the real drivers of forest degradation, in order to develop interventions to ensure the sustainability of natural resources.

Evaluation of effectiveness of supervised classification algorithms in land cover classification using ASTER images-A case study from the Mankweng (Turfloop) Area and its environs, Limpopo Province, South Africa

The production of land cover maps using supervised classification algorithms is one of the most common applications of remote sensing. In this study, the effectiveness of supervised classification algorithms in land cover classification using ASTER data was evaluated in the Mankweng Area and its environs. The false colour composite image generated from combination of band 1, 2 and 3 in red, green and blue, respectively, was used to generate training classes for six land cover types (waterbody, forest, vegetation, Duiwelskloof leucogranite, Turfloop granite and built-up land). These were used to construct land cover maps using eight supervised classification algorithms: Maximum Likelihood, Minimum Distance, Support Vector Machine, Mahalanobis Distance, Parallelepiped, Neural Network, Spectral Angle Mapper and Spectral Information Divergence. To evaluate the effectiveness of the algorithms, the land cover maps were subjected to accuracy assessment to determine precision of the algorithms in accurately classifying the land cover types and level of confidence that can be attributed to the land cover maps. Most algorithms poorly performed in classifying spatially overlapping land cover types without abrupt boundaries. This indicates that the environmental conditions and distribution of land cover types can affect the performance of certain classification algorithms, and thus need to be considered prior to selection of algorithms. However, Support Vector Machine and Minimum Distance proved to be the two most effective algorithms as they provided better producer’s and user’s accuracy in the range of 80-100% for all land cover types, which represent good classification.

Cashew expansion holds potential for carbon stocks enhancement in the forest-savannah transitional zone of Ghana

Despite the scientific contributions of earlier studies to the understanding of shifts in land-use to cashew in the forest-savannah transition zone, some questions remain unanswered: (1) what is the extent of landscape-level conversions associated with cashew? (2) could the expansion of cashew in the forest-savannah transitional zone potentially serve as a carbon mitigation strategy relative to other land uses at the landscape level? and (3) how does the land use dynamics associated with cashew development hold implications for Ghana’s climate change mitigation strategies? The Wenchi Municipality was used as a case study. Landsat images for 2000, 2015 and 2020 were classified to produce land cover maps using the random forest algorithm to overall accuracies of 72.6 %, 86.9 % and 91.2 %, respectively. The InVEST software was used to estimate the carbon sequestered by each land cover class in the landscape. Finally, 150 cashew farmers, four focus group discussions and key experts were interviewed. The results show that cashew plantations had increased in the Municipality by 31103.05 ha between 2000 and 2020. The shift in land use towards cashew increased the Above Ground Carbon from 599956.73 tons to 1039402.32 tons between 2000 and 2020, representing 696219.15 tons (883.5 %) increase in the Municipality. According to the farmers, the shift in land use to cashew was mainly due to the available ready market and relative higher income. This suggests that, there is great potential for mitigating climate change in the forest-savannah transitional zone using cashew. Hence, Ghana needs to explore the gains associated with cashew plantations, particularly as it transforms low-income crops with challenging post-harvest and marketing constraints into relatively lucrative income generation cashew, which has major implications for improvements in livelihood and co-benefits for carbon stocks enhancement. Ultimately, as Ghana rolls out its Nationally Determined Contributions implementation among other forestry sector climate change mitigation interventions, it is important to consider the role of cashew as a major contributor to carbon stocks enhancement within the Forest–Savannah transitional zone. This study provides major decision support for formulating policies, interventions, and measures to promote climate change and livelihood improvement opportunities while rolling out safeguard measures to prevent further conversion of relatively high carbon stocks landscapes, such as the savannah woodlands within the forest-savannah transitional zone.

Optimizing Crowdsourced Land Use and Land Cover Data Collection: A Two-Stage Approach

Citizen science has become an increasingly popular approach to scientific data collection, where classification tasks involving visual interpretation of images is one prominent area of application, e.g., to support the production of land cover and land-use maps. Achieving a minimum accuracy in these classification tasks at a minimum cost is the subject of this study. A Bayesian approach provides an intuitive and reasonably straightforward solution to achieve this objective. However, its application requires additional information, such as the relative frequency of the classes and the accuracy of each user. While the former is often available, the latter requires additional data collection. In this paper, we present a two-stage approach to gathering this additional information. We demonstrate its application using a hypothetical two-class example and then apply it to an actual crowdsourced dataset with five classes, which was taken from a previous Geo-Wiki crowdsourcing campaign on identifying the size of agricultural fields from very high-resolution satellite imagery. We also attach the R code for the implementation of the newly presented approach.

Depletion of Forested Area: Geidam Perspective

Land cover maps provide best understanding of current landscape change over time. One can evaluate past land cover maps for several different years for management decisions as well as gain insight into the possible effects on decisions making. One of the key monitoring areas is how the environment keeps degrading resulting from increased anthropogenic activities such as the removal of the forest covers. This study monitors the pattern changes of the Geidam Yobe state Nigeria, using Landsat images of two different periods from Enhanced Thematic Mapper (ETM+) image of data of 1988 and 2018. The images were geometrically and atmospherically pre-processed then classified, using maximum likelihood (MLC) algorithm to produce land cover maps of the Geidam. The accuracy of the classification was assessed with confusion matrices giving results morethan the minimum 85% required. The results revealed that the built-up and tree area increase by (+30.97%), water body reduced by (-5.06%) and forest reduce by (-23.48%) within the study period. This shows a rapid decrease in the forest, which is partly attributed to deforestation activities and partly to climate change impact.

Land Cover Map for Multifunctional Landscapes of Taita Taveta County, Kenya, Based on Sentinel-1 Radar, Sentinel-2 Optical, and Topoclimatic Data

Taita Taveta County (TTC) is one of the world’s biodiversity hotspots in the highlands with some of the world’s megafaunas in the lowlands. Detailed mapping of the terrestrial ecosystem of the whole county is of global significance for biodiversity conservation. Here, we present a land cover map for 2020 based on satellite observations, a machine learning algorithm, and a reference database for accuracy assessment. For the land cover map production processing chain, temporal metrics from Sentinel-1 and Sentinel-2 (such as median, quantiles, and interquartile range), vegetation indices from Sentinel-2 (normalized difference vegetation index, tasseled cap greenness, and tasseled cap wetness), topographic metrics (elevation, slope, and aspect), and mean annual rainfall were used as predictors in the gradient tree boost classification model. Reference sample points which were collected in the field were used to guide the collection of additional reference sample points based on high spatial resolution imagery for training and validation of the model. The accuracy of the land cover map and uncertainty of area estimates at 95% confidence interval were assessed using sample-based statistical inference. The land cover map has an overall accuracy of 81 ± 2.3% and it is freely accessible for land use planners, conservation managers, and researchers.

Generating continuous fine-scale land cover mapping by edge-guided maximum a posteriori based spatiotemporal sub-pixel mapping

Global land cover mapping activities are of great important for retrieving the environment we are living in. However, the trade-off between spatial and temporal resolution makes it difficult to obtain the continuous fine-scale land cover product for detail and frequent land surface analysis. To overcome the difficulty, this study proposed a modified maximum a posteriori (MAP) based spatiotemporal sub-pixel mapping (SSM), for a long-term and fine-scale land cover mapping. The proposed approach consists of three parts, i.e., data fidelity term, spatial regularization term, and enhanced temporal regularization term, which can make use of historical fine-scale image for a better super-resolution of the current coarse image. One regional experiment was implemented for the algorithm validation, and one real experiment was investigated, by generating continuous land cover mapping products (2010–2015) at Landsat-like spatial resolution from time series of MODIS images. Compared with traditional methods, the proposed approach reconstructed more pleasant spatial details and had greater quantitative accuracy, outperforming the state-of-the-art spatiotemporal sub-pixel mapping method by averagely 2.58% for all the inspected dates in OA metric. This study not only provides a generalized research framework for generating continuous fine-scale land cover product from daily available MODIS images, but also validates the usability of the time series products for detail land cover monitoring, and reveals a finding that the study area of Wuhan is a well practitioner of the sustainable development policy during its urbanization process.

Land Cover and Land Use Mapping of the East Asian Summer Monsoon Region from 1982 to 2015

Owing to the recent intensification of the East Asian summer monsoon, the frequency of floods and dry spells, which commonly affect more than one billion people, is continuously increasing. Thus, understanding the causes of changes in the EASM is paramount. Land cover and land use change can perturb a regional climate system through biogeophysical and biogeochemical processes. However, due to the scarcity of temporally continuous land cover and land use maps, the impact of land cover and land use change on the EASM is still not thoroughly explored. In the present study, this limitation was addressed via the production of annual land cover and land use maps of the East Asian summer monsoon region covering a period of 34 years (1982–2015). This was achieved through a random forest classification of phenological information derived from the Advanced Very High-Resolution Radiometer Global Inventory Modeling and Mapping Studies Normalized Difference Vegetation Index dataset and terrain information from the Advanced Land Observing Satellite World 3D—30 m Digital Surface Model data. Nine ecological zones were involved in the random forest classification and the classified map in 2015 was validated using very high-resolution images obtained from Google Earth. The overall accuracy (73%) of the classification map surpasses the Moderate Resolution Imaging Spectroradiometer and Global Land Surface Satellite land cover products for the same year by ~7% and 4%, respectively. According to our classified maps, croplands and forests significantly increased in the East Asian summer monsoon region from 1982 to 2015. The dominant transition in these three decades was from croplands to forests.

Spatially Balanced Sampling for Validation of GlobeLand30 Using Landscape Pattern-Based Inclusion Probability

Global and local land-cover mapping products provide important data on land surface. However, the accuracy of land-cover products is the key issue for their further scientific application. There has been neglect of the relationship between inclusion probability and spatial heterogeneity in traditional spatially balanced sampling. The aim of this paper was to propose an improved spatially balanced sampling method using landscape pattern-based inclusion probability. Compared with other global land-cover datasets, Globeland30 has the advantages of high resolution and high classification accuracy. A two-stage stratified spatially balanced sampling scheme was designed and applied to the regional validation of GlobeLand30 in China. In this paper, the whole area was divided into three parts: the Tibetan Plateau region, the Northwest China region, and the East China region. The results show that 7242 sample points were selected, and the overall accuracy of GlobeLand30-2010 in China was found to be 80.46%, which is close to the third-party assessment accuracy of GlobeLand30. This method improves the representativeness of samples, reduces the classification error of remote sensing, and provides better guidance for biodiversity and sustainable development of environment.

Wild Plant Habitat Characterization in the Last Two Decades in the Nile Delta Coastal Region of Egypt

Environmental and land-use changes put severe pressure on wild plant habitats. The present study aims to assess the biodiversity of wild plant habitats and the associated spatiotemporal environmental changes in the coastal region of Dakahlia Governorate following an integrated approach of remote sensing, GIS, and samples analysis. Thirty-seven stands were spatially identified and studied to represent the different habitats of wild plants in the Deltaic Mediterranean coastline region. Physical and chemical characteristics of soil samples were examined, while TWINSPAN classification was used to identify plant communities. Two free Landsat images (TM and OLI) acquired in 1999 and 2019 were processed to assess changes via the production of land use and cover maps (LULC). Moreover, NDSI, NDMI, and NDSI indices were used to identify wild plant habitats. The floristic composition indicated the existence of 57 species, belonging to 51 genera of 20 families. The largest families were Asteraceae, Poaceae, and Chenopodiaceae. The classification of vegetation led to the identification of four groups. Canonical Correspondence Analysis (CCA) revealed that electrical conductivity, cations, organic carbon, porosity, chlorides, and bicarbonates are the most effective soil variables influencing vegetation. The results of the spectral analysis indicated an annual coverage of bare lands (3.56 km2), which is strongly related to the annual increase in vegetation (1.91 km2), water bodies (1.22 km2), and urban areas (0.43 km2). The expansion of urban and agricultural regions subsequently increased water bodies and caused occupancy of bare land, resulting in the development of wild plant habitats, which are mostly represented by the sparse vegetation class as evaluated by NDVI. The increase in mean moisture values (NDMI) from 0.03 in 1999 to 0.15 in 2019 might be explained by the increase in total areas of wild plant habitats throughout the study period (1999–2019). This may improve the adequacy of environments for wild habitats, causing natural plant proliferation.

Identification and modelling of forest fire severity and risk zones in the Cross – Niger transition forest with remotely sensed satellite data

Forest fires are a serious environmental hazard within the forest ecosystem, which can be studied with Remote Sensing and GIS. The aim of the study is to identify and model forest fires severity and risk zones within the Cross – Niger transition forest. To achieve this aim, remotely sensed data, such as Landsat – 8 OLI (2020) and ASTER DEM were used to produce land cover maps and topography parameters such as aspect, elevation and slope. The topographic maps and the Google Earth imagery were used to extract human settlements and road networks. The final forest fire risk zone (FFRZ) map was prepared by integrating the different parameters such as Land cover, aspect, elevation, slope, proximities to roads and settlements in the ArcGIS environment. The FFRZ was categorized into three categories as low, moderate and high risk zones, based on their fire susceptibility. The category of low, moderate and high FFRZ were represented as 2731.7 km2 (12.5%), 17997.69 km2 (82.59 %) and 1061.63 km2 (4.87%) respectively. The study shows that Remote Sensing and GIS are excellent tools for modelling forest fire risk zones, hence proving that fires are anthropogenic in origin.

BENCHMARKING OF CONVOLUTIONAL NEURAL NETWORK APPROACHES FOR VEGETATION LAND COVER MAPPING

Abstract. Satellite Image Time Series (SITS) are becoming available at high spatial, spectral and temporal resolutions across the globe by the latest remote sensing sensors. These series of images can be highly valuable when exploited by classification systems to produce frequently updated and accurate land cover maps. The richness of spectral, spatial and temporal features in SITS is a promising source of data for developing better classification algorithms. However, machine learning methods such as Random Forests (RF), despite their fruitful application to SITS to produce land cover maps, are structurally unable to properly handle intertwined spatial, spectral and temporal dynamics without breaking the structure of the data. Therefore, the present work proposes a comparative study of various deep learning algorithms from the Convolutional Neural Network (CNN) family and evaluate their performance on SITS classification. They are compared to the processing chain coined iota2, developed by the CESBIO and based on a RF model. Experiments are carried out in an operational context using with sparse annotations from 290 labeled polygons. Less than 80 000 pixel time series belonging to 8 land cover classes from a year of Sentinel-2 monthly syntheses are used. Results show on a test set of 131 polygons that CNNs using 3D convolutions in space and time are more accurate than 1D temporal, stacked 2D and RF approaches. Best-performing models are CNNs using spatio-temporal features, namely 3D-CNN, 2D-CNN and SpatioTempCNN, a two-stream model using both 1D and 3D convolutions.

Integrating Convolutional Neural Network and Multiresolution Segmentation for Land Cover and Land Use Mapping Using Satellite Imagery

Depletion of natural resources, population growth, urban migration, and expanding drought conditions are some of the reasons why environmental monitoring programs are required and regularly produced and updated. Additionally, the usage of artificial intelligence in the geospatial field of Earth observation (EO) and regional land monitoring missions is a challenging issue. In this study, land cover and land use mapping was performed using the proposed CNN–MRS model. The CNN–MRS model consisted of two main steps: CNN-based land cover classification and enhancing the classification with spatial filter and multiresolution segmentation (MRS). Different band numbers of Sentinel-2A imagery and multiple patch sizes (32 × 32, 64 × 64, and 128 × 128 pixels) were used in the first experiment. The algorithms were evaluated in terms of overall accuracy, precision, recall, F1-score, and kappa coefficient. The highest overall accuracy was obtained with the proposed approach as 97.31% in Istanbul test site area and 98.44% in Kocaeli test site area. The accuracies revealed the efficiency of the CNN–MRS model for land cover map production in large areas. The McNemar test measured the significance of the models used. In the second experiment, with the Zurich Summer dataset, the overall accuracy of the proposed approach was obtained as 92.03%. The results are compared quantitatively with state-of-the-art CNN model results and related works.