The Review of Land Use/Land Cover Mapping AI Methodology and Application in the Era of Remote Sensing Big Data

With the increasing number of remote sensing satellites,the diversification of observation modals,and the continuous advancement of artificial intelligence algorithms,historically opportunities have been brought to the applications of earth observation and information retrieval,including climate change monitoring,natural resource investigation,ecological environment protection,and territorial space planning.Over the past decade,artificial intelligence technology represented by deep learning has made significant contributions to the field of Earth observation.Therefore,this review will focus on the bottlenecks and development process of using deep learning methods for land use/land cover mapping of the Earth’s surface.Firstly,it introduces the basic framework of semantic segmentation network models for land use/land cover mapping.Then,we summarize the development of semantic segmentation models in geographical field,focusing on spatial and semantic feature extraction,context relationship perception,multi-scale effects modelling,and the transferability of models under geographical differences.Then,the application of semantic segmentation models in agricultural management,building boundary extraction,single tree segmentation and inter-species classification are reviewed.Finally,we discuss the future development prospects of deep learning technology in the context of remote sensing big data.

Spatial and temporal classification of synthetic satellite imagery:land cover mapping and accuracy validation

This study focused on land cover mapping based on synthetic images,especially using the method of spatial and temporal classification as well as the accuracy validation of their results.Our experimental results indicate that the accuracy of land cover map based on synthetic imagery and actual observation has a similar standard compared with actual land cover survey data.These findings facilitate land cover mapping with synthetic data in the area where actual observation is missing.Furthermore,in order to improve the quality of the land cover mapping,this research employed the spatial and temporal Markov random field classification approach.Test results show that overall mapping accuracy can be increased by approximately 5% after applying spatial and temporal classification.This finding contributes towards the achievement of higher quality land cover mapping of areas with missing data by using spatial and temporal information.

Comparison of Layer-stacking and Dempster-Shafer Theory-based Methods Using Sentinel-1 and Sentinel-2 Data Fusion in Urban Land Cover Mapping

Data fusion has shown potential to improve the accuracy of land cover mapping,and selection of the optimal fusion technique remains a challenge.This study investigated the performance of fusing Sentinel-1(S-1)and Sentinel-2(S-2)data,using layer-stacking method at the pixel level and Dempster-Shafer(D-S)theory-based approach at the decision level,for mapping six land cover classes in Thu Dau Mot City,Vietnam.At the pixel level,S-1 and S-2 bands and their extracted textures and indices were stacked into the different single-sensor and multi-sensor datasets(i.e.fused datasets).The datasets were categorized into two groups.One group included the datasets containing only spectral and backscattering bands,and the other group included the datasets consisting of these bands and their extracted features.The random forest(RF)classifier was then applied to the datasets within each group.At the decision level,the RF classification outputs of the single-sensor datasets within each group were fused together based on D-S theory.Finally,the accuracy of the mapping results at both levels within each group was compared.The results showed that fusion at the decision level provided the best mapping accuracy compared to the results from other products within each group.The highest overall accuracy(OA)and Kappa coefficient of the map using D-S theory were 92.67%and 0.91,respectively.The decision-level fusion helped increase the OA of the map by 0.75%to 2.07%compared to that of corresponding S-2 products in the groups.Meanwhile,the data fusion at the pixel level delivered the mapping results,which yielded an OA of 4.88%to 6.58%lower than that of corresponding S-2 products in the groups.

LAND-COVER DENSITY-BASED APPROACH TO URBAN LAND USE MAPPING USING HIGH-RESOLUTION IMAGERY

Nowadays, remote sensing imagery, especially with its high spatialresolution, has become an indispensable tool to provide timely up-gradation of urban land use andland cover information, which is a prerequisite for proper urban planning and management. Thepossible method described in the present paper to obtain urban land use types is based on theprinciple that land use can be derived from the land cover existing in a neighborhood. Here, movingwindow is used to represent the spatial pattern of land cover within a neighborhood and seven windowsizes (61mx61m, 68mx68m, 75mx75m, 87mx87m, 99mx99m, 110mx110m and 121mxl21m) are applied todetermining the most proper window size. Then, the unsupervised method of ISODATA is employed toclassify the layered land cover density maps obtained by the moving window. The results of accuracyevaluation show that the window size of 99mx99m is proper to infer urban land use categories and theproposed method has produced a land use map with a total accuracy of 85%.

Assessment of degraded mattoral land using remote sensing im-agery in Guadalteba Area,Spain

Natural land cover information is important for analysing and understanding of the current terrestrial situation, especially in the study area that is facing the environmental deteriorating increasingly. The study combined the remote sensing Aster data and ground truth to improve 2001 land cover map of Guadalteba area in Spain, and increased the accuracy from 47% to 70%. The general land cover map produced about the Guadalteba study area outlines the distribution of the vegetation type and the current natural land cover in the area. Based on this improved general land cover map, the natural cover map gave an indication of the present location of nature and agriculture areas. The shrub land degradation map identified location of various shrub/matorral areas and different levels of degradation. The further analysis and discussion were done. The output maps indicated that much of the natural cover mostly dominated by formations of shrubs has been changed to agriculture and other land uses. It is observed that shrubland covers a small percentage, approximately 9% of the study area, due to land degradation in most parts caused by human interfere. Keywords Accuracy assessment - Aster - Land cover map - Matorral degradation map - Remote Sensing CLC number S757.3 Document code A Foundation item: This paper was partly sponsored by NFP (Netherlands Feliowship Program) and National Strategic Project “Environmentally Sound Forest Management Techniques and Models in Natural Forest in Northeast China” (2001BA510B0702) respectively.Biography: XING Yan-qiu (1970-), female, Lecturer, in College of Engi neering and technology Northeast Forestry University. Harbin 150040. P. R. ChinaResponsible editor: Song Funan

Monitoring Land Use and Infrastructure Changes in Industrial Complex Using Geo-Informatics Technology in Gujarat State, India

In the present study, detailed investigations have been carried out in Petroleum, Chemicals and Petrochemical Investment Region (PCPIR) area in Vygra and Bharuch Talukas in Bharuch district of Gujarat State. Indian Remote Sensing Satellite (IRS-P6) LISS-III, LISS-IV and CARTOSAT digital data covering PCPIR area in Bharuch district for the period of January & February of 2011, 2012 and 2013 was analyzed for land use/land cover mapping and monitoring the changes in land use. Various thematic land use/land cover maps were prepared and GIS database for various thematic layers have been generated using satellite and ground based information. The results indicate that the major land use in the PCPIR area is agriculture with crop lands ranging from 61 to 63 per cent of the total area. Crop land has decreased from 64.7% during 2011 to 62.7% during 2013 in the PCPIR region. Area under plantations in PCPIR area has also decreased from 5.5% during 2011 to 5.2% during 2012. The industrial area has increased from 6.0% to 7.6% of the total area of the PCPIR region. The total built-up area (industries & village area) has increased from 7.1% during 2011 to 8.7% during 2013. Tree plantations in the area of around 42 ha were carried out by GIDC during 2012 and 2013 to increase the green cover in the PCPIR area.

Assessing the suitability of GlobeLand30 for mapping land cover in Germany

Global land cover(LC)maps have been widely employed as the base layer for a number of applications including climate change,food security,water quality,biodiversity,change detection,and environmental planning.Due to the importance of LC,there is a pressing need to increase the temporal and spatial resolution of global LC maps.A recent advance in this direction has been the GlobeLand30 dataset derived from Landsat imagery,which has been developed by the National Geomatics Center of China(NGCC).Although overall accuracy is greater than 80%,the NGCC would like help in assessing the accuracy of the product in different regions of the world.To assist in this process,this study compares the GlobeLand30 product with existing public and online datasets,that is,CORINE,Urban Atlas(UA),OpenStreetMap,and ATKIS for Germany in order to assess overall and per class agreement.The results of the analysis reveal high agreement of up to 92%between these datasets and GlobeLand30 but that large disagreements for certain classes are evident,in particular wetlands.However,overall,GlobeLand30 is shown to be a useful product for characterizing LC in Germany,and paves the way for further regional and national validation efforts.

Evaluation of diverse classification approaches for land use/cover mapping in a Mediterranean region utilizing Hyperion data

Information on Earth’s land surface cover is commonly obtained through digital image analysis of data acquired from remote sensing sensors.In this study,we evaluated the use of diverse classification techniques in discriminating land use/cover types in a typical Mediterranean setting using Hyperion imagery.For this purpose,the spectral angle mapper(SAM),the object-based and the non-linear spectral unmixing based on artificial neural networks(ANNs)techniques were applied.A further objective had been to investigate the effect of two approaches for training sites selection in the SAM classification,namely of the pixel purity index(PPI)and of the direct selection of training points from the Hyperion imagery assisted by a QuickBird imagery and field-based training sites.Objectbased classification outperformed the other techniques with an overall accuracy of 83%.Sub-pixel classification based on the ANN showed an overall accuracy of 52%,very close to that of SAM(48%).SAM applied using the training sites selected directly from the Hyperion imagery supported by the QuickBird image and the field visits returned an increase accuracy by 16%.Yet,all techniques appeared to suffer from the relatively low spatial resolution of the Hyperion imagery,which affected the spectral separation among the land use/cover classes.

Producing consistent visually interpreted land cover reference data: learning from feedback

Reference data for large-scale land cover map are commonly acquired by visual interpretation of remotely sensed data.To assure consistency,multiple images are used,interpreters are trained,sites are interpreted by several individuals,or the procedure includes a review.But little is known about important factors influencing the quality of visually interpreted data.We assessed the effect of multiple variables on land cover class agreement between interpreters and reviewers.Our analyses concerned data collected for validation of a global land cover map within the Copernicus Global Land Service project.Four cycles of visual interpretation were conducted,each was followed by review and feedback.Each interpreted site element was labelled according to dominant land cover type.We assessed relationships between the number of interpretation updates following feedback and the variables grouped in personal,training,and environmental categories.Variable importance was assessed using random forest regression.Personal variable interpreter identifier and training variable timestamp were found the strongest predictors of update counts,while the environmental variables complexity and image availability had least impact.Feedback loops reduced updating and hence improved consistency of the interpretations.Implementing feedback loops into the visually interpreted data collection increases the consistency of acquired land cover reference data.

Developing spatially and thematically detailed backdated maps for land cover studies

Global or regional land cover change on a decadal time scale can be studied at a high level of detail using the availability of remote sensing data such as that provided by Landsat.However,there are three main technical challenges in this goal.First,the generation of land cover maps without reference data is problematic(backdating).Second,it is important to maintain high accuracies in land cover change map products,requiring a reasonably rich legend within each map.Third,a high level of automation is necessary to aid the management of large volumes of data.This paper describes a robust methodology for processing time series of satellite data over large spatial areas.The methodology includes a retrospective analysis used for the generation of training and test data for historical periods lacking reference information.This methodology was developed in the context of research on global change in the Iberian Peninsula.In this study we selected two scenes covering geographic regions that are representative of the Iberian Peninsula.For each scene,we present the results of two classifications(1985-1989 and 2000-2004 quinquennia),each with a legend of 13 categories.An overall accuracy of over 92%was obtained for all 4 maps.

A method characterizing urban expansion based on land cover map at 30 m resolution

Based on the global land cover data at 30 m resolution(Globe Land30) in the year 2000 and 2010, the urban expansion process of 320 cities in China was analyzed using lognormal regression, and the expansion model were established. Three metrics were presented for the models, including the peak position, the full width at half maximum, and the skewness. It was found that the three metrics could reveal different patterns of the urban expansion process of cities with different sizes. Specifically, cities with larger size tend to expand outward strongly, and their expansion intensity and influence are likely to be higher. Moreover, most cities' expansion occurs around the urban core with spatially limited influence. In addition, it was also found that the city's expansion intensity is related to the city size. These results showed that the lognormal regression model could describe the distribution of urban expansion with effectiveness and robustness.

Assessment of forest geospatial patterns over the three giant forest areas of China

Geospatial patterns of forest fragmentation over the three traditional giant forested areas of China （Northeastern, southwestern and Southern China） were analyzed comparatively and reported based on a 250-m resolution land cover dataset. Specifically, the spatial patterns of forest fragmentation were characterized by combining geospatial metrics and forest fragmentation models. The driving forces resulting in the differences of the forest spatial patterns were also investigated. Results suggested that forests in southwest China had the highest severity of forest fragmentation, followed by south region and northeast region. The driving forces of forest fragmentation in China were primarily the giant population and improper exploitation of forests. In conclusion, the generated information in the study provided valuable insights and implications as to the fragmentation patterns and the conservation of hiodiversity or genes, and the use of the chosen geospatial metrics and forest fragmentation models was quite useful for depicting forest fragmentation patterns.

Accuracies of Global Land Cover Maps Checked against Fluxnet Sites

Global land cover data products are key sources of information in understanding the complex interactions between human activities and global change. They play a critical role in improving performances of ecosystem, hydrological and atmospheric models. Three freely available global land cover products developed in the United States are popularly used by the scientific community. These include two global maps developed separately by the United States Geological Survey （USGS） and the University of Maryland （UMD） with NOAA Advanced Very High Resolution Radiometer （ AVHRR ） data, and one developed by Boston University with the EOS Moderate Resolution Imaging Spectroradiometer （ MODIS） data. They are compared with known land cover types at 250 available Fluxnet sites around the world. The overall accuracies are 37%, 36% and 42%, respectively for the USGS, UMD and Boston global land cover maps, Some future global land cover mapping strategies are suggested.

Characterizing and monitoring global landscapes using GlobeLand30 datasets:the first decade of the twenty-first century

Global land cover maps are important sources of information for a wide range of studies including land change analysis and climate change research.While the global land cover maps attempt to present a consistent and homogenous data in terms of the production process,the existing datasets offer coarse resolution data,e.g.1000 m for IGBP DISCover and 300 m for GlobeCover 2009 that is oftentimes challenging.Recently,GlobeLand30 data based on Landsat archive for two timestamps of 2000 and 2010 has been released.It presents a finer spatial resolution of 30 m,which provides numerous opportunities for a wide range of studies.The main objective of this study is to use this dataset for characterizing global land cover patterns,monitoring,and identifying extreme land change cases with their types and magnitude.The findings reveal massive land change patterns including deforestation,desertification,shrinkage of water bodies,and urbanization across the globe.The results and discussions of this research can help policy-makers,environmental planners,ecosystem services providers and climate change researchers to gain finer insights about the forms of global land change.Future research calls for further investigation of the underlying causes of the massive changes and their consequences on our ecosystems and human populations.

Validation of GIS layers in the EU:getting adapted to available reference data

An optimal validation of a thematic map would ideally require in-situ observations of a large sample of units specifically conceived for the map under validation.This is often not possible due to budget limitations.The alternative can be using photo-interpretation of high or very high resolution images instead of in-situ observations or using available data sets that do not fully comply with the ideal characteristics:unit size,reference date or sampling plan.This paper illustrates some examples of use of available data in the European Union.For land cover maps,the best existing data set is probably Land Use/Cover Areaframe Survey(LUCAS)that has been conducted by Eurostat on four occasions since 2001.Because LUCAS is based on systematic sampling,advantages and limitations of systematic sampling are discussed.A fine-scale population density map is presented as an example of a situation in which reference data on a statistical sample cannot be collected.