Land Cover Monitoring Research Articles

A hybrid colour model based land cover classification using random forest and support vector machine classifiers

Land cover monitoring using remotely sensed data requires robust classification methods for the accurate mapping of complex land cover and land use categories. Classification is a supervised learning method which maps a data item into predefined classes. Colour is an important feature used in image classification since humans tend to distinguish images mostly based on colour feature. This paper proposes a hybrid colour model for land cover classification in which colour features are extracted by combining the hue (H) values of HSV colour space and luminance (L) values of LUV colour space. The extracted features are trained and tested with random forest (RF) and support vector machine (SVM) classifiers. The performance of the proposed hybrid colour model is compared with the existing HSV colour space model using RF and SVM classifiers based on several metrics such as accuracy, sensitivity, specificity and f-score. Hyper spectral dataset of Pavia University and an IRS LISS IV orthorectified dataset are chosen as the input image for this experiment.

Comparison of hyperspectral and multi-spectral imagery to building a spectral library and land cover classification performance

The main aim of this research work is to compare k -nearest neighbor algorithm (KNN) supervised classification with migrating means clustering unsupervised classification (MMC) method on the performance of hyperspectral and multispectral data for spectral land cover classes and develop their spectral library in Samara, Russia. Accuracy assessment of the derived thematic maps was based on the analysis of the classification confusion matrix statistics computed for each classified map, using for consistency the same set of validation points. We were analyzed and compared Earth Observing-1 (EO-1) Hyperion hyperspectral data to Landsat 8 Operational Land Imager (OLI) and Advance Land Imager (ALI) multispectral data. Hyperspectral imagers, currently available on airborne platforms, provide increased spectral resolution over existing space based sensors that can document detailed information on the distribution of land cover classes, sometimes species level. Results indicate that KNN (95, 94, 88 overall accuracy and .91, .89, .85 kappa coefficient for Hyp, ALI, OLI respectively) shows better results than unsupervised classification (93, 90, 84 overall accuracy and .89, .87, .81 kappa coefficient for Hyp, ALI, OLI respectively). Development of spectral library for land cover classes is a key component needed to facilitate advance analytical techniques to monitor land cover changes. Different land cover classes in Samara were sampled to create a common spectral library for mapping landscape from remotely sensed data. The development of these libraries provides a physical basis for interpretation that is less subject to conditions of specific data sets, to facilitate a global approach to the application of hyperspectral imagers to mapping landscape. In addition, it is demonstrated that the hyperspectral satellite image provides more accurate classification results than those extracted from the multispectral satellite image. The higher classification accuracy by KNN supervised was attributed principally to the ability of this classifier to identify optimal separating classes with low generalization error, thus producing the best possible classes’ separation.

Monitoring Land Cover Change Integrated Remote Sensing And GIS Techniques In Karbala Governorate

Monitoring Land Cover Change Integrated Remote Sensing And GIS Techniques In Karbala Governorate

A hybrid colour model based land cover classification using random forest and support vector machine classifiers

Land cover monitoring using remotely sensed data requires robust classification methods for the accurate mapping of complex land cover and land use categories. Classification is a supervised learning method which maps a data item into predefined classes. Colour is an important feature used in image classification since humans tend to distinguish images mostly based on colour feature. This paper proposes a hybrid colour model for land cover classification in which colour features are extracted by combining the hue (H) values of HSV colour space and luminance (L) values of LUV colour space. The extracted features are trained and tested with random forest (RF) and support vector machine (SVM) classifiers. The performance of the proposed hybrid colour model is compared with the existing HSV colour space model using RF and SVM classifiers based on several metrics such as accuracy, sensitivity, specificity and f-score. Hyper spectral dataset of Pavia University and an IRS LISS IV orthorectified dataset are chosen as the input image for this experiment.

Monitoring Land Cover Changes in the Disaster-Prone Area: A Case Study of Cangkringan Sub-District, the Flanks of Mount Merapi, Indonesia

Volcanic eruption is one of the natural factors that affect land cover changes. This study aimed to monitor land cover changes using a remote sensing approach in Cangkringan Sub-district, Yogyakarta, Indonesia, one of the areas most vulnerable to Mount Merapi eruption. Three satellite images, dating from 2001, 2006 and 2011, were used as main data for land cover classification based on a supervised classification approach. The land cover detection analysis was undertaken by overlaying the classification results from those images. The results show that the dominant land cover class is annual crops, covering 40% of the study area, while the remaining 60% consists of forest cover types, dryland farming, paddy fields, settlements, and bare land. The forests were distributed in the north, and the annual crops in the middle of the study area, while the villages and the rice fields were generally located in the south. In the 2001–2011 period, forests were the most increased land cover type, while annual crops decreased the most, as a result of the eruption of Mount Merapi in 2010. Such data and information are important for the local government or related institutions to formulate Detailed Spatial Plans (RDTR) in the Disaster-Prone Areas (KRB).

Using high-resolution multitemporal imagery to highlight severe land management changes in Mediterranean vineyards

Mediterranean vineyards are subjected to drastic changes due to socio-economic, physical and environmental drivers. Whereas global trends affecting the evolution of Mediterranean land cover are well-known, their descriptions at a fine spatial scale, i.e., a plot level of several decametres, and over long time periods are still underexplored. This lack of exploration persists despite the increase in multitemporal imagery at fine resolutions. As scale and stakeholders are often correlated, we hypothesize that monitoring land cover at fine spatial and temporal scale would help understanding the local drivers shaping landscapes. To that end, we exploited a database of aerial pictures to obtain classified and vectorized land uses at the field level during the past five decades of a large watershed. The land uses were analyzed to detect changes at the field level considering the evolution of vineyard management. Changes in land management were synthesized in transition matrices expressing gains, losses and swaps for each land use category and illustrated using chord and Sankey diagrams. The results showed both a change in land management through a severe transformation of vineyards areas, from goblet to trellised vineyards, during the 1980s and a progressive evolution of land uses from vineyards to urban areas, arboriculture areas and arable lands. This transformation resulted from the local policies on vineyard replanting and the arrival of mechanization.

Capabilities of Multispectral Satellite Data in an Assessment of the Status of Abandoned Fire Hazardous and Rewetting Peat Extraction Lands

The capabilities of several multispectral satellite data types to identify the status of peatlands affected by peat extraction and abandoned deposits are examined to assess potential fire dangers and rewetting effectiveness. The available level of detail of describing land/vegetation cover for monitoring abandoned peat extraction sites using Spot-5 HRG, Spot-6 HRG, and Landsat-7 ETM+ satellite images has been demonstrated using the example of peatlands in the Meschera National Park (Vladimir oblast). The results reflect the pros and cons of using different data types to analyze the status of abandoned peat-extraction lands for purposes of peatland inventory, land-cover monitoring, and the prioritization of sites subject to rewetting and mire restoration, as well as for an evaluation of the effectiveness of these measures.

Effects of high spatial and temporal resolution Earth observations on simulated hydrometeorological variables in a cropland (southwestern France)

Abstract. Agricultural landscapes are often constituted by a patchwork of crop fields whose seasonal evolution is dependent on specific crop rotation patterns and phenologies. This temporal and spatial heterogeneity affects surface hydrometeorological processes and must be taken into account in simulations of land surface and distributed hydrological models. The Sentinel-2 mission allows for the monitoring of land cover and vegetation dynamics at unprecedented spatial resolutions and revisit frequencies (20 m and 5 days, respectively) that are fully compatible with such heterogeneous agricultural landscapes. Here, we evaluate the impact of Sentinel-2-like remote sensing data on the simulation of surface water and energy fluxes via the Interactions between the Surface Biosphere Atmosphere (ISBA) land surface model included in the EXternalized SURface (SURFEX) modeling platform. The study focuses on the effect of the leaf area index (LAI) spatial and temporal variability on these fluxes. We compare the use of the LAI climatology from ECOCLIMAP-II, used by default in SURFEX-ISBA, and time series of LAI derived from the high-resolution Formosat-2 satellite data (8 m). The study area is an agricultural zone in southwestern France covering 576 km2 (24 km × 24 km). An innovative plot-scale approach is used, in which each computational unit has a homogeneous vegetation type. Evaluation of the simulations quality is done by comparing model outputs with in situ eddy covariance measurements of latent heat flux (LE). Our results show that the use of LAI derived from high-resolution remote sensing significantly improves simulated evapotranspiration with respect to ECOCLIMAP-II, especially when the surface is covered with summer crops. The comparison with in situ measurements shows an improvement of roughly 0.3 in the correlation coefficient and a decrease of around 30 % of the root mean square error (RMSE) in the simulated evapotranspiration. This finding is attributable to a better description of LAI evolution processes with Formosat-2 data, which further modify soil water content and drainage of soil reservoirs. Effects on annual drainage patterns remain small but significant, i.e., an increase roughly equivalent to 4 % of annual precipitation levels with simulations using Formosat-2 data in comparison to the reference simulation values. This study illustrates the potential for the Sentinel-2 mission to better represent effects of crop management on water budgeting for large, anthropized river basins.

Analyzing spatiotemporal land use/cover dynamic using remote sensing imagery and GIS techniques case: Kan basin of Iran

Land use/land cover (LU/LC) that are significant elements for the interconnection of human activities and environment monitoring can be useful to find out the deviations of saving a maintainable environment. Remote sensing is a very useful tool for the affair of land use or land cover monitoring, which can be helpful to decide the allocation of land use and land cover. Supervised classification-maximum likelihood algorithm in GIS was applied in this study to detect land use/land cover changes observed in Kan basin using multispectral satellite data obtained from Landsat 5 (TM) and 8 (OLI) for the years 2000 and 2016, respectively. The main aim of this study was to gain a quantitative understanding of land use and land cover changes in Kan basin of Tehran over the period 2000–2016. For this purpose, firstly supervised classification technique was applied to Landsat images acquired in 2000 and 2016. The Kan basin was classified into five major LU/LC classes including: Built up areas, garden, pasture, water and bare-land. Change detection analysis was performed to compare the quantities of land cover class conversions between time intervals. The results revealed both increase and decrease of the different LU/LC classes from 2000 to 2016. The results indicate that during the study period, built-up land, and pastures have increased by 0.2% (76.4 km2) and 0.3% (86.03 km2) while water, garden and bare land have decreased by 0, 0.01% (3.62 km2) and 0.4% (117.168 km2), respectively. Information obtained from change detection of LU/LC can aid in providing optimal solutions for the selection, planning, implementation and monitoring of development schemes to meet the increasing demands of human needs in land management.

A service relation model for web-based land cover change detection

Change detection with remotely sensed imagery is a critical step in land cover monitoring and updating. Although a variety of algorithms or models have been developed, none of them can be universal for all cases. The selection of appropriate algorithms and construction of processing workflows depend largely on the expertise of experts about the “algorithm-data” relations among change detection algorithms and the imagery data used. This paper presents a service relation model for land cover change detection by integrating the experts’ knowledge about the “algorithm-data” relations into the web-based geo-processing. The “algorithm-data” relations are mapped into a set of web service relations with the analysis of functional and non-functional service semantics. These service relations are further classified into three different levels, i.e., interface, behavior and execution levels. A service relation model is then established using the Object and Relation Diagram (ORD) approach to represent the multi-granularity services and their relations for change detection. A set of semantic matching rules are built and used for deriving on-demand change detection service chains from the service relation model. A web-based prototype system is developed in .NET development environment, which encapsulates nine change detection and pre-processing algorithms and represents their service relations asan ORD. Three test areas from Shandong and Hebei provinces, China with different imagery conditions are selected for online change detection experiments, and the results indicate that on-demand service chains can be generated according to different users’ demands.

Monitoring Land Cover of Google Web Service Images through ECOC and ANFIS Classifiers

Monitoring Land Cover of Google Web Service Images through ECOC and ANFIS Classifiers

Global Estimates of Ecosystem Service Value and Change: Taking Into Account Uncertainties in Satellite-based Land Cover Data

Global estimates of ecosystem service value (ESV) and change are often produced using satellite-based land cover maps. However, uncertainties in global land cover data and their impacts on ESV estimation have not been fully recognized. Considerably inflated estimates of land cover change and ESV change could be derived using a direct map comparison approach when classification uncertainties are not explicitly taken into account. This study collected all available global land cover datasets and applied an ensemble approach to derive the range and central tendency of terrestrial ESV estimates. Different input data caused ESV estimate varying between 35.0 and 56.5trillionInt$/year. Wetland classes, albeit having the highest per unit value, were the most uncertain classes mapped using satellite data. To reduce uncertainty, a spatial data harmonization procedure was developed, which resulted in an improved ESV estimate at 49.4trillionInt$/year. The study further illustrated the quantification of changes in forest ESV using a high-resolution global forest cover change dataset. An ESV loss of 716.0billionInt$/year was estimated between 2000 and 2012—a result representing one fifth of previous estimates. These findings highlighted the importance of improving the characterization and monitoring of land cover for global ESV and change estimation.

Soft Computing Techniques for Land Use and Land Cover Monitoring with Multispectral Remote Sensing Images: A Review

Multispectral remote sensing images are the primary source in the land use and land cover (LULC) monitoring. This is achieved by LULC classification and LULC change detection. The change detection in LULC includes the detection of water bodies, forest fire, forest degradation, agriculture areas monitoring, etc. Various change detection and LULC classification methods have their own advantages and disadvantages, and no single method is optimal and finds applicability for all cases. This paper summarizes and analyses the various soft computing and feature extraction techniques used for LULC classification and change detection. Based on the average error rate, performances of the different soft computing techniques are evaluated. The broad usage of multispectral remote sensing images, object-based change detection, neural networks and various levels of image fusion methods offer more potential in LULC monitoring.

Mapping finer‐resolution land surface emissivity using Landsat images in China

AbstractLand surface emissivity is a crucial parameter for obtaining the land surface temperature and estimating the land surface energy budget from remote sensing data. The current emissivity products always have a coarser spatial resolution than the products from the visible and near‐infrared data. This study focused on the generation of an emissivity product at a spatial resolution of 30 m using a new global land cover product called Finer‐Resolution Observation and Monitoring of Global Land Cover and Landsat images. Summer‐average emissivity products in four narrowbands (Landsat 5/Thematic Mapper Band 6, Landsat 7/Enhanced Thematic Mapper Plus Band 6, and Landsat 8 Thermal Infrared Sensor bands 1 and 2) and two broadbands (3–14 μm and 8–13.5 μm) were produced in China. Results illustrated that the narrowband emissivities ranged from 0.95 to 0.99, whereas the broadband emissivities ranged from 0.93 to 0.99 in the study area. Intercomparisons in different places showed that the new emissivity was close to Advanced Spaceborne Thermal Emission and Reflection Radiometer emissivity with a difference of about 0.015 for narrowband emissivity and about 0.02 for broadband emissivity on a regional scale. For application purposes, the emissivities were released in the Worldwide Reference System 2 and geographic coordinate systems with several spatial resolutions resampled from its original scale of 30 m.

COMPARING ATMOSPHERIC CORRECTION METHODS FOR LANDSAT OLI DATA

Landsat data used for monitoring activities to land cover because it has spatial resolution and high temporal. To monitor land cover changes in an area, atmospheric correction is needed to be performed in order to obtain data with precise digital value picturing current condition. This study compared atmospheric correction methods namely Quick Atmospheric Correction (QUAC), Dark Object Subtraction (DOS) and Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH). The correction results then were compared to Surface Reflectance (SR) imagery data obtained from the United States Geological Survey (USGS) satelite. The three atmospheric correction methods were applied to Landsat OLI data path/row126/62 for 3 particular dates. Then, sample on vegetation, soil and bodies of water (waterbody) were retrieved from the image. Atmospheric correction results were visually observed and compared with SR sample on the absolute value, object spectral patterns, as well as location and time consistency. Visual observation indicates that there was a contrast change on images that had been corrected by using FLAASH method compared to SR, which mean that the atmospheric correction method was quite effective. Analysis on the object spectral pattern, soil, vegetation and waterbody of images corrected by using FLAASH method showed that it was not good enough eventhough the reflectant value differed greatly to SR image. This might be caused by certain variables of aerosol and atmospheric models used in Indonesia. QUAC and DOS made more appropriate spectral pattern of vegetation and water body than spectral library. In terms of average value and deviation difference, spectral patterns of soil corrected by using DOS was more compatible than QUAC.

Landsat-based classification in the cloud: An opportunity for a paradigm shift in land cover monitoring

Satellite-derived land cover maps play an important role in many applications, including monitoring of smallholder-dominated agricultural landscapes. New cloud-based computing platforms and satellite sensors offer opportunities for generating land cover maps designed to meet the spatial and temporal requirements of specific applications. Such maps can be a significant improvement compared to existing products, which tend to be coarser than 300m, are often not representative of areas with fast-paced land use change, and have a fixed set of cover classes. Here, we present two approaches for land cover classification using the Landsat archive within Google Earth Engine. Random forest classification was performed with (1) season-based composites, where median values of individual bands and vegetation indices were generated from four years for each of four seasons, and (2) metric-based composites, where different quantiles were computed for the entire four-year period. These approaches were tested for six land cover types spanning over 18,000 locations in Zambia, with ground “truth” determined by visual inspection of high-resolution imagery from Google Earth. The methods were trained on 30% of these points and tested on the remaining 70%, and results were also compared with existing land cover products. Overall accuracies of about 89% were achieved for the season- and metric-based approaches for individual classes, with 93% and 94% accuracy for distinguishing cropland from non-cropland. For the latter task, the existing Globeland30 dataset based on Landsat had much lower accuracies (around 77% on average), as did existing cover maps at coarser resolutions. Overall, the results support the use of either season or metric-based classification approaches. Both produce better results than those obtained from previous classifiers, which supports a general paradigm shift away from dependence on standard static products and towards custom generation of on-demand cover maps designed to fulfill the needs of each specific application.

Effect of different SRFs on time series of spectral indices, between sentinel-2 and other sensors for the purpose of vegetation land cover monitoring

Effect of different SRFs on time series of spectral indices, between sentinel-2 and other sensors for the purpose of vegetation land cover monitoring

Acquisition probability differences in cloud coverage of the available Landsat observations over mainland Southeast Asia from 1986 to 2015

ABSTRACTLandsat data are the longest available records that consistently document global change. However, the extent and degree of cloud coverage typically determine its usability, especially in the tropics. In this study, scene-based metadata from the U.S. Geological Survey Landsat inventories, ten-day, monthly, seasonal, and annual acquisition probabilities (AP) of targeted images at various cloud coverage thresholds (10% to 100%) were statistically analyzed using available Landsat TM, ETM+, and OLI observations over mainland Southeast Asia (MSEA) from 1986 to 2015. Four significant results were found. First, the cumulative average acquisition probability of available Landsat observations over MSEA at the 30% cloud cover (CC) threshold was approximately 41.05%. Second, monthly and ten-day level probability statistics for the 30% CC threshold coincide with the temporal distribution of the dry and rainy seasons. This demonstrates that Landsat images acquired during the dry season satisfy the requirements needed for land cover monitoring. Third, differences in acquisition probabilities at the 30% CC threshold are different between the western and eastern regions of MSEA. Finally, the ability of TM, ETM+, and OLI to acquire high-quality imagery has gradually enhanced over time, especially during the dry season, along with consequently larger probabilities at lower CC thresholds.

Analytical review of European projects LUCAS and CORINE for monitoring and validation of land cover and landuse on the basis of satellite and ground observations and experience of land cover mapping in Ukraine

Analytical review of the European project CORINE (Coordination of Information on the Environment) Land Cover (CLC) and LUCAS (Land Use/Cover Area Frame Statistical Survey) and their use for the monitoring and validation of land cover and land use is fulfilled. The experience of land cover mapping in Ukraine on the basis of developed in the Space Research Institute of NAS of Ukraine and SSA of Ukraine neural network classification methods, heterogeneous satellite data fusion methods and implementation of appropriate automated information technologies within the SIGMA and Sen2Agri European projects are considered.

Scrubbing Up: Multi-Scale Investigation of Woody Encroachment in a Southern African Savannah

Changes in the extent of woody vegetation represent a major conservation question in many savannah systems around the globe. To address the problem of the current lack of broad-scale cost-effective tools for land cover monitoring in complex savannah environments, we use a multi-scale approach to quantifying vegetation change in Kruger National Park (KNP), South Africa. We test whether medium spatial resolution satellite data (Landsat, existing back to the 1970s), which have pixel sizes larger than typical vegetation patches, can nevertheless capture the thematic detail required to detect woody encroachment in savannahs. We quantify vegetation change over a 13-year period in KNP, examine the changes that have occurred, assess the drivers of these changes, and compare appropriate remote sensing data sources for monitoring change. We generate land cover maps for three areas of southern KNP using very high resolution (VHR) and medium resolution satellite sensor imagery from February 2001 to 2014. Considerable land cover change has occurred, with large increases in shrubs replacing both trees and grassland. Examination of exclosure areas and potential environmental driver data suggests two mechanisms: elephant herbivory removing trees and at least one separate mechanism responsible for conversion of grassland to shrubs, theorised to be increasing atmospheric CO2. Thus, the combination of these mechanisms causes the novel two-directional shrub encroachment that we observe (tree loss and grassland conversion). Multi-scale comparison of classifications indicates that although spatial detail is lost when using medium resolution rather than VHR imagery for land cover classification (e.g., Landsat imagery cannot readily distinguish between tree and shrub classes, while VHR imagery can), the thematic detail contained within both VHR and medium resolution classifications is remarkably congruent. This suggests that medium resolution imagery contains sufficient thematic information for most broad-scale land cover monitoring requirements in heterogeneous savannahs, while having the benefits of being cost-free and providing a longer historical archive of data than VHR sources. We conclude that monitoring of broad-scale land cover change using remote sensing has considerable potential as a cost-effective tool for both better informing land management practitioners, and for monitoring the future landscape-scale impacts of management policies in savannahs.