Land Use Classification Research Articles

Techniques and Challenges of the Machine Learning Method for Land Use/Land Cover (LU/LC) Classification in Remote Sensing Using the Google Earth Engine

In order to accurately observe the globe, land use and land cover are crucial. Due to the proliferation of several global modifications associated with the existence of the planet, land use/land cover (LU/LC) classification is now regarded as a topic of highest significance in the natural environment and an important field to be researched by researchers. Google Earth provides satellite image dataset which contains high-resolution images; these images are used to analyze the land area. In order to address the dearth of review articles throughout the land use/land cover classification phase, we proposed a full evaluation, which might help researchers continue their work. Therefore, the purpose of this study is to investigate the methodical steps involved in classifying land use and land cover utilizing the Google Earth platform. The most widely used techniques researchers employ to achieve LU/LC classification using Google Earth Engine are examined in this work. The classification of land use and land cover for a specific region using time series was covered in this study, along with the many types of land use and land cover classes and the approach employed by Google Earth. The limits of the GEE tool and difficulties encountered during the process of classifying land use and cover have also been covered in this survey document. The importance of this review rests in inspiring future scholars to tackle the LU/LC analysis problem successfully, and this study offers researchers a road map for assessing land use/land cover classification.

Land-use pattern and landscape structure impact butterfly diversity and abundance in organic agroecosystems

ContextAgricultural intensification and land-use patterns impact biodiversity and reduce ecosystem functions. Organic farming is often seen as a more biodiversity friendly way of farming. However, not all organic farming is equally supportive for biodiversity, structural diversity of the farms has a great impact and the often high intertwinement of organic and conventional farmland causes spill-over problems.ObjectiveOn organic farmland in central Italy, we analysed butterfly diversity and community structures in different land-use categories across a heterogeneous landscape. We aim to analyse the impact of the different land-use structures such as hedges and forest edges on this farmland, as well as the influence of adjoining conventionally farmed areas.MethodsWe studied butterflies on organic farmland including olive and hazelnut plantations, as well as meadows and fallow land, surrounded by hedges and forests. Hereby we considered the effects of land-use on the studied site and the adjoining land, e.g., semi-natural habitats (as hedges or forest edges), organic or conventional farmland. We counted butterflies along line transects and recorded habitat parameters for each transect. All encountered butterflies were classified according their ecological demands, behaviour, and life-history.ResultsWe found highest mean butterfly species richness in fallow land and lowest in hazelnut plantations. Species community structures differed hardly between plantations and meadows, but considerably compared to fallow land. Butterfly abundances are reduced for olive plantations adjacent to agricultural fields treated with pesticides, compared to such surrounded by organic fields. Habitat para-meters such as the number of flowering plant species and shade impact butterfly diversity and abundance. Forest edges producing considerable shadow to the transects had a significantly negative effect on diversity and abundance, while the influence of hedges was positive.ConclusionsBased on our findings, we conclude that also organic farmland has different degrees of biodiversity depending on a variety of land-use pattern. In particular, habitat heterogeneity and a high diversity of flowers positively influence butterfly diversity. The intensive mixing of organic and conventional agriculture must be seen critical for conservation as negative effects might spill over to organic fields, but might be compensable by landscape structures like hedges.

Mapping and spatiotemporal dynamics of land-use and land-cover change based on the Google Earth Engine cloud platform from Landsat imagery: A case study of Zhoushan Island, China

Land resources are an essential foundation for socioeconomic development. Island land resources are limited, the type changes are particularly frequent, and the environment is fragile. Therefore, large-scale, long-term, and high-accuracy land-use classification and spatiotemporal characteristic analysis are of great significance for the sustainable development of islands. Based on the advantages of remote sensing indices and principal component analysis in accurate classification, and taking Zhoushan Archipelago, China, as the study area, in this work long-term satellite remote sensing data were used to perform land-use classification and spatiotemporal characteristic analysis. The classification results showed that the land-use types could be exactly classified, with the overall accuracy and Kappa coefficient greater than 94% and 0.93, respectively. The results of the spatiotemporal characteristic analysis showed that the built-up land and forest land areas increased by 90.00 km2 and 36.83 km2, respectively, while the area of the cropland/grassland decreased by 69.77 km2. The areas of the water bodies, tidal flats, and bare land exhibited slight change trends. The spatial coverage of Zhoushan Island continuously expanded toward the coast, encroaching on nearby sea areas and tidal flats. The cropland/grassland was the most transferred-out area, at up to 108.94 km2, and built-up land was the most transferred-in areas, at up to 73.31 km2. This study provides a data basis and technical support for the scientific management of land resources.

The Land-Use and Land-Cover Changes in the Este District, South Gondar Zone, Northwestern Ethiopia, in the Last Four Decades (the 1980s to 2020s)

Environmental transformations are the results of land-use and land-cover (LULC) changes. This study aims to investigate the LULC changes and associated factors in the Este District in northwestern Ethiopia, for the last four decades (the 1980s to 2020s). The land-use and land-cover classes were analyzed using supervised classification techniques in ERDASS IMAGINE 2015 and ArcGIS 10.3.1, categorizing the Landsat satellite images for 1984, 2000, and 2018 into six cover classes: settlement, forests-shrublands, cropland, grassland, bare land, and water body. We used a historical Google map, topo-sheets, and ancillary data to verify the classification accuracies for 1984, 2000, and 2018, respectively. The climate and demographic data were obtained from the Ethiopian Meteorological Station and Ethiopian Statistical Agency, respectively. In addition, data from key informant interviews and focus group discussions were also used to understand the local community experiences and perceptions toward LULC changes. The values of each LULC class were correlated with the demographic and climatic variables, using simple correlation analysis to evaluate the impact of demographic and climatic conditions on LULC changes. The analysis indicated that the least and largest classes of LULC in Este District were water bodies (mean cover = 1.9 km2) and croplands (mean cover = 791.7 km2), respectively. Cropland coverage increased by 2% in 2000, then decreased by 11% in 2018. Between 1984 and 2018, the grasslands and settlements increased by 22% and 0.5%, respectively. Half of the bare land and one-tenth of the forests-shrublands also decreased over the last four decades (the 1980s to 2020s). As a result, the original croplands and forests–shrublands classes (11% each) before 2000 were converted into new grasslands and croplands after 2018, respectively. The study indicated that precipitation, solar radiation, and population growth are the potential drivers, and the perceptions of local communities are nearly in line with the statistical analysis results. Alternative income sources, such as tourism and carbon trading and the participation in afforestation programs, could reverse the situation in the study area.

An Application of Machine-Learning Model for Analyzing the Impact of Land-Use Change on Surface Water Resources in Gauteng Province, South Africa

The change in land-use diversity is attributed to the anthropogenic factors sustaining life. The surface water bodies and other crucial natural resources in the study area are being depleted at an alarming rate. This study explored the implications of the changing land-use diversity on surface water resources by using a random forest (RF) classifier machine-learning algorithm and remote-sensing models in Gauteng Province, South Africa. Landsat datasets from 1993 to 2022 were used and processed in the Google Earth Engine (GEE) platform, using the RF classifier. The results indicate nine land-use diversity classes having increased and decreased tendencies, with high F-score values ranging from 72.3% to 100%. In GP, the spatial coverage of BL has shrunk by 100.4 km2 every year over the past three decades. Similarly, BuA exhibits an annual decreasing rate of 42.4 km2 due to the effect of dense vegetation coverage within the same land use type. Meanwhile, water bodies, marine quarries, arable lands, grasslands, shrublands, dense forests, and wetlands were expanded annually by 1.3, 2.3, 2.9, 5.6, 11.2, 29.6, and 89.5 km2, respectively. The surface water content level of the study area has been poor throughout the study years. The MNDWI and NDWI values have a stronger Pearson correlation at a radius of 5 km (r = 0.60, p = 0.000, n = 87,260) than at 10 and 15 km. This research is essential to improve current land-use planning and surface water management techniques to reduce the environmental impacts of land-use change.

Two contrasting years of continuous N[formula omitted]O and CO[formula omitted] fluxes on a shallow-peated drained agricultural boreal peatland

Drained agricultural boreal peatlands comprise a large source of nitrous oxide (N2O) and carbon dioxide (CO2) but a small sink or source of methane (CH4). N2O fluxes have high spatial and temporal variability and are often measured with the chamber technique. Therefore, continuous measurements of N2O fluxes are needed to better understand how N2O emissions are triggered and to reduce the uncertainty of annual N2O budget estimations. Here we present a two-year-long time series of continuous measurements of CO2 and N2O fluxes of a shallow-peated drained agricultural boreal peatland cultivated for grass silage. The fluxes were measured with the area-averaging eddy covariance technique. Several N2O peak events were observed throughout all seasons. The peaks were associated with meteorological or management events, such as soil thawing or freezing, precipitation, fertilization and glyphosate application. The annual N2O budget was 4.74 ±0.47 and 6.08 ±0.49 kg N2O-N ha−1 y−1 in 2020 and 2021, respectively. The annual CO2 budget, comprising the sum of net ecosystem exchange and biomass export, was 3.70 ±0.22 and 5.54 ±0.33 t CO2-C ha−1 y−1 in 2020 and 2021, respectively. The N2O budget during the first, warmer winter was 106% higher than during the second, meteorologically more typical winter, due to the higher frequency of soil freezing–thawing cycles. The average annual N2O budget was 36%–50% lower than the IPCC Emission Factor (EF) while the CO2 budget was in accordance with the IPCC EF. CO2 emissions dominated the total CO2-eq emissions of our site but N2O also had a significant contribution of 12%. Our results also suggest that glyphosate application enhanced N2O emissions in the last quarter of 2021. However, the full rotation should be measured to confirm whether there is a need to re-evaluate the N2O IPCC EF for ‘grassland drained boreal’ land-use class.

Land-Use Mapping with Multi-Temporal Sentinel Images Based on Google Earth Engine in Southern Xinjiang Uygur Autonomous Region, China

Land-use maps are thematic materials reflecting the current situation, geographical diversity, and classification of land use and are an important scientific foundation that can assist decision-makers in adjusting land-use structures, agricultural zoning, regional planning, and territorial improvement according to local conditions. Spectral reflectance and radar signatures of time series are important in distinguishing land-use types. However, their impact on the accuracy of land-use mapping and decision making remains unclear. Also, the many spatial and temporal heterogeneous landscapes in southern Xinjiang limit the accuracy of existing land-use classification products. Therefore, our objective herein is to develop reliable land-use products for the highly heterogeneous environment of the southern Xinjiang Uygur Autonomous Region using the freely available public Sentinel image datasets. Specifically, to determine the effect of temporal features on classification, several classification scenarios with different temporal features were developed using multi-temporal Sentinel-1, Sentinel-2, and terrain data in order to assess the importance, contribution, and impact of different temporal features (spectral and radar) on land-use classification models and determine the optimal time for land-use classification. Furthermore, to determine the optimal method and parameters suitable for local land-use classification research, we evaluated and compared the performance of three decision-tree-related classifiers (classification and regression tree, random forest, and gradient tree boost) with respect to classifying land use. Yielding the highest average overall accuracy (95%), kappa (95%), and F1 score (98%), we determined that the gradient tree boost model was the most suitable for land-use classification. Of the four individual periods, the image features in autumn (25 September to 5 November) were the most accurate for all three classifiers in relation to identifying land-use classes. The results also show that the inclusion of multi-temporal image features consistently improves the classification of land-use products, with pre-summer (28 May–20 June) images providing the most significant improvement (the average OA, kappa, and F1 score of all the classifiers were improved by 6%, 7%, and 3%, respectively) and fall images the least (the average OA, kappa, and F1 score of all the classifiers were improved by 2%, 3%, and 2%, respectively). Overall, these analyses of how classifiers and image features affect land-use maps provide a reference for similar land-use classifications in highly heterogeneous areas. Moreover, these products are designed to describe the highly heterogeneous environments in the study area, for example, identifying pear trees that affect local economic development, and allow for the accurate mapping of alpine wetlands in the northwest.

Analysis of the vegetation fragments connectivity

The habitat fragmentation is one of the most cited causes of species extinction and loss of biological diversity and it is defined as the process by a continuous area is reduced in size and divided into two or more fragments separated by an environment or matrix, different from the original. The analysis of landscape fragmentation is done through land use mapping, where it is possible to identify different spots and determine the degree of fragmentation or connectivity. This work aims to analyze the connectivity of the landscape fragments present in the Tietê-Jacaré Hydrographic Basin - SP, through the application of the metric indices of Connectivity and Fragmentation, with the purpose of evaluating the dynamic and temporal process in 2007 and 2017, through the use of Geographic Information Systems (GIS). Considering the interval of 10 years, there was an expansion of agricultural activities with an increase of 24,507.53ha of cultivated areas and a vegetation loss of 32,149ha. It was also evidenced the decrease in the number of fragments, as well as the increase in the average size of the fragments, derived from the growth of the agricultural matrix with consequent reduction of the fragment’s diversification. The vegetation configuration of the region has undergone few structural changes, the main connections are related to the end of each sub-basin and consequently the hydrographic basin as a whole. The fragmentation indexes, when expressing the state of configuration of the landscapes, were presented as important tools in the diagnosis aimed at the conservation of ecosystems. These indices allowed the comparison not only of natural fragments but of all classes of land use and land cover, which is essential in regional planning.

An Analysis of CMAQ Gas Phase Dry Deposition over North America Through Grid-Scale and Land-Use Specific Diagnostics in the Context of AQMEII4.

The fourth phase of the Air Quality Model Evaluation International Initiative (AQMEII4) is conducting a diagnostic intercomparison and evaluation of deposition simulated by regional-scale air quality models over North America and Europe. In this study, we analyze annual AQMEII4 simulations performed with the Community Multiscale Air Quality Model (CMAQ) version 5.3.1 over North America. These simulations were configured with both the M3Dry and Surface Tiled Aerosol and Gas Exchange (STAGE) dry deposition schemes available in CMAQ. A comparison of observed and modeled concentrations and wet deposition fluxes shows that the AQMEII4 CMAQ simulations perform similarly to other contemporary regional-scale modeling studies. During summer, M3Dry has higher ozone (O3) deposition velocities (Vd) and lower mixing ratios than STAGE for much of the eastern U.S. while the reverse is the case over eastern Canada and along the West Coast. In contrast, during winter STAGE has higher O3 Vd and lower mixing ratios than M3Dry over most of the southern half of the modeling domain while the reverse is the case for much of the northern U.S. and southern Canada. Analysis of the diagnostic variables defined for the AQMEII4 project, i.e. grid-scale and land-use (LU) specific effective conductances and deposition fluxes for the major dry deposition pathways, reveals generally higher summertime stomatal and wintertime cuticular grid-scale effective conductances for M3Dry and generally higher soil grid-scale effective conductances (for both vegetated and bare soil) for STAGE in both summer and winter. On a domain-wide basis, the stomatal grid-scale effective conductances account for about half of the total O3 Vd during daytime hours in summer for both schemes. Employing LU-specific diagnostics, results show that daytime Vd varies by a factor of 2 between LU categories. Furthermore, M3Dry vs. STAGE differences are most pronounced for the stomatal and vegetated soil pathway for the forest LU categories, with M3Dry estimating larger effective conductances for the stomatal pathway and STAGE estimating larger effective conductances for the vegetated soil pathway for these LU categories. Annual domain total O3 deposition fluxes differ only slightly between M3Dry (74.4 Tg/year) and STAGE (76.2 Tg/yr), but pathway-specific fluxes to individual LU types can vary more substantially on both annual and seasonal scales which would affect estimates of O3 damages to sensitive vegetation. A comparison of two simulations differing only in their LU classification scheme shows that the differences in LU cause seasonal mean O3 mixing ratio differences on the order of 1 ppb across large portions of the domain, with the differences generally largest during summer and in areas characterized by the largest differences in the fractional coverages of the forest, planted/cultivated, and grassland LU categories. These differences are generally smaller than the M3Dry vs. STAGE differences outside the summer season but have a similar magnitude during summer. Results indicate that the deposition impacts of LU differences are caused both by differences in the fractional coverages and spatial distributions of different LU categories as well as the characterization of these categories through variables like surface roughness and vegetation fraction in look-up tables used in the land-surface model and deposition schemes. Overall, the analyses and results presented in this study illustrate how the diagnostic grid-scale and LU-specific dry deposition variables adopted for AQMEII4 can provide insights into similarities and differences between the CMAQ M3Dry and STAGE dry deposition schemes that affect simulated pollutant budgets and ecosystem impacts from atmospheric pollution.

THE CONTRIBUTION OF GRASSLANDS IN THE MANNER OF LAND USE FROM THE HILL AREA OF BANAT IN RELATION TO LIVESTOCK

The meadows in the hill area of Banat can be defined as resources with multidimensional significance. In another context, grasslands can be considered one of the defining elements in livestock activities in rural areas. In this respect, the purpose of the research was to quantify the grassland resources in the area of interest and assess their potential in forage cover of local animals. The analysis of the area of interest was done by geospatial methods, applied with ArcGIS 10.4 software. Geospatial data, statistical and bibliographic data from the literature were used. The area of interest is characterized by complex and variable environmental conditions, which is also reflected in the use of the territory. In the five analyzed administrative units, grasslands occupy surfaces between 16.65% (UAT Balinţ) and 52.14% (UAT Secaş), these being interspersed with other categories of land use, respectively with arable land in the low areas and with forests in the areas located at higher altitudes. As far as animals are concerned, it was found that the largest herds were in ATU Brestovăţ, where per hectare of pasture has 0.72 LSU. Through the economic benefits offered, the meadows in the analyzed hill area constitute important natural resources for local and regional communities.

Exploration of Land-Use Change and Drivers In Qixia City Based on Multi-Temporal Landsat8 Remote Sensing Image

The research results on land use and cover change (LUCC) are constantly updated, which is of great significance to correctly handle the relationship between human activities and ecological environment. In this article, Landsat8 remote sensing images, DEM data and statistical yearbook data of Qixia city in Shandong province are selected in 2013, 2018 and 2021 to explore the changes in this city’s land use. Through supervised classification, land use dynamic degree (LUDD) and land use transfer matrix, LUCC has been analyzed, and through the CORREL and DEM slope analysis, the main driving force has been investigated. The results indicate that the area of orchards in Qixia city is showing an increasing trend and the area of farmland, water and residential land shows a decreasing trend. Single land use dynamic degree of industrial land is the largest. The main driving forces of the industrial land and the orchard area are the social economic factors, which are mainly manifested as the influence of the regional GDP, the gross product of primary and tertiary sectors as well as the gross value of agricultural produce. In addition, the population is what is causing changes in the amount of farmland and residential land, and the terrain slope also impact on the variation of land-use classification significantly.

Flood risk assessment through rapid urbanization LULC change with destruction of urban green infrastructures based on NASA Landsat time series data: A case of study Kuala Lumpur between 1990–2021

Kuala Lumpur, as one of the fastest growing cities in the world, is facing significant negative impacts on its urban ecosystem services (UESs) due to rapid urbanization. Among the classes of land use and land cover (LULC) in Kuala Lumpur, urban green infrastructure (UGI) plays a crucial role in mitigating these impacts. The present study aimed to evaluate the amount of UGIs in Kuala Lumpur for a better understanding of hydro-meteorological hazards assessment. To this end, historical data from 1990 to 2021 were collected using NASA Landsat 4 and Landsat 8 to evaluate the past and current tendencies of LULC in Kuala Lumpur. Various techniques and software, including Markov Chain, ArcGIS, and TerrSet, were utilized to examine the interactions and interpret and create maps. Six categories of land were defined, including grass, waterbodies, soil, buildings, urban forest trees, and urban trees. The study utilized Landsat's Normalized Difference Vegetation Index (NDVI) to analyze the impact of modifications on UGIs. Additionally, the Landsat Normalized Difference Water Index (NDWI) was employed to illustrate changes in urban water values. The findings revealed a significant decline in urban forest trees from 15.99% in 1990 to 6.5% in 2021, resulting in an increase in urban water values and an increase in flooding. The NDWI values in 1990 ranged between 0.383635 and −0.593184, while by 2021, they had fallen to between 0.593288 and −0.44379. These changes in NDVI and NDWI have contributed significantly to the intensity and frequency of urban flooding in Kuala Lumpur.

Past and future land-use/land-cover change trends and its potential drivers in Koore’s agricultural landscape, Southern Ethiopia

Wise-use of land resources determines sustainability of environment and human welfare. But, the ongoing land-use change constrains environmental sustainability. Thus, this study aims to quantify past and future land-use change trend with its associated drivers in Koore’s agricultural landscape of Southern Ethiopia. To quantify the change trends, geospatial tools were applied. A CA–Markov model was used to predict land-use changes. Relative Importance Index was used to identify major drivers of the change. The result of the study showed agroforestry, built-up and cultivation rise by 192%, 152% and 18% at the expense of natural environment. Population growth and agricultural expansion are among the leading land-use change drivers. CA–Markov model revealed a continued increment of anthropogenically modified land-use classes which adversely affect the sustainability of the landscape. Information on spatiotemporal dynamics has been confirmed as effectual measure for nature conservation, agriculture, urban planners, and policymakers to plan sustainable development actions.

Machine Learning Classifier Evaluation for Different Input Combinations: A Case Study with Landsat 9 and Sentinel-2 Data

High-resolution multispectral remote sensing images offer valuable information about various land features, providing essential details and spatially accurate representations. In the complex urban environment, classification accuracy is not often adequate using the complete original multispectral bands for practical applications. To improve the classification accuracy of multispectral images, band reduction techniques are used, which can be categorized into feature extraction and feature selection techniques. The present study examined the use of multispectral satellite bands, spectral indices (including Normalized Difference Built-up Index, Normalized Difference Vegetation Index, and Normalized Difference Water Index) for feature extraction, and the principal component analysis technique for feature selection. These methods were analyzed both independently and in combination for the classification of multiple land use and land cover features. The classification was performed for Landsat 9 and Sentinel-2 satellite images in Delhi, India, using six machine learning techniques: Classification and Regression Tree, Minimum Distance, Naive Bayes, Random Forest, Gradient Tree Boosting, and Support Vector Machine on Google Earth Engine platform. The performance of the classifiers was evaluated quantitatively and qualitatively to analyze the classification results with whole image (comprehensive feature) and small subset (targeted feature). The RF and GTB classifiers were found to outperform all others in the quantitative analysis of all input combinations for both Landsat 9 and Sentinel-2 datasets. RF achieved a classification total accuracy of 96.19% for Landsat and 96.95% for Sentinel-2, whereas GTB achieved 91.62% for Landsat and 92.89% for Sentinel-2 in all band combinations. Furthermore, the RF classifier achieved the highest F1 score of 0.97 in both the Landsat and Sentinel datasets. The qualitative analysis revealed that the PCA bands were particularly useful to classifiers in distinguishing even the slightest differences among the feature class. The findings contribute to the understanding of feature extraction and selection techniques for land use and land cover classification, offering insights into their effectiveness in different scenarios.

Spatiotemporal evaluation of waning grassland habitats for swamp deer conservation across the human-dominated upper Gangetic Plains, India

SummaryGrassland habitats currently face severe anthropogenic exploitation, thereby affecting the survival of grassland-dependent biodiversity globally. The biodiversity-rich grasslands of India lack quantitative spatiotemporal information on their status. We evaluated the status of upper Gangetic Plains grasslands in 2015 and compared it with those from 1985, 1995 and 2005. On-ground mapping and visual classifications revealed a 57% decline in these grasslands between 1985 (418 km2) and 2015 (178 km2), mostly driven by habitat conversion (74% contribution by cropland). Limited radiotelemetry data from endemic swamp deer indicated a possible grassland-dominated average home range size of 1.02 km2, and these patches were highly preferred (average Ivlev’s index = 0.85) over other land-use classes at both spatial and temporal scales. Camera-trapping within the core habitats suggests the critical use of these patches as fawning/breeding grounds. Habitat suitability analysis indicates only c. 17% of the area along the Ganges is suitable as swamp deer habitat. We recommend the protection of these critical grassland patches to maintain ‘dynamic corridors’, with restoration and other management approaches involving multiple stakeholders to ensure the survival of this critical ecosystem.

Difficult times for amphibians: Effects of land-use change at the local and landscape scales in the Iberá Wetlands

Land-use change and management practices have led to habitat loss, one of the greatest factors in biodiversity decline. Particularly, amphibians comprise the highest number of threatened vertebrate species at the global scale. In this work, amphibian communities were analysed in three differing landscapes: a protected wetland, a livestock-impaired rangeland and a pine afforestation. In each landscape, amphibian species were sampled. Land-cover type and composition, as well as primary vegetation types were characterised at the landscape and local scale, respectively. The relationship between these environmental variables and the amphibian communities was analysed. Twenty-one amphibian species were identified; the protected and afforested landscapes were the richest, whereas the rangeland showed the lowest species richness and diversity estimates. At the local scale, vegetation, water coverage and land-use category explained the higher presence of amphibian species and their abundance. These results show how different land-uses, especially livestock farming, modify the composition of amphibian communities. This work constitutes a foundation for the development of sustainable management practices for conserving amphibians in landscape-level altered habitats.

Mammalian Roadkill in a Semi-Arid Region of Brazil: Species, Landscape Patterns, Seasonality, and Hotspots

Roadkill is one of the principal causes of the loss of biodiversity around the world. The effects of roads on mammals are still poorly understood in regions with a semi-arid climate, where many knowledge gaps persist. The present study provides an inventory of the mammalian species affected on highways in northeastern Brazil, as well as identifying roadkill hotspots and contributing to the understanding of how seasonality and the landscape may influence the roadkill patterns of wild mammals. A total of 6192.52 km of road were sampled in 53 field surveys conducted between 2013 and 2017. Landsat 8 satellite images and data from the MapBiomas platform were used to classify land use and cover for analysis. Buffers of 1 km, 5 km, and 10 km were created around the study roads to identify the landscape variables associated with roadkill events. Ripley’s 2D K-Statistics and the 2D HotSpot test were used to identify roadkill aggregations and hotspots; GLMMs were generated for the landscape variables and evaluated using the Akaike Information Criterion. The Kruskal–Wallis test was applied to investigate the potential effects of seasonality. A total of 527 wild animal carcasses were recorded as a result of vehicular collision. The species with the highest roadkill records were Cerdocyon thous, Euphractus sexcinctus, and Procyon cancrivorus, while two species—Leopardus emiliae and Herpailurus yagouaroundi—are considered to be under threat of extinction. For mammals in general, the best GLMM indicated an increase in roadkills with increasing density of local vegetation areas, and a decrease as urban areas increased. The model also found that the mammals were less impacted in the vicinity of a protected area. In the specific case of C. thous, the roadkill rate was lower when urban infrastructure was more common than dense vegetation; the rate increased as areas of dense vegetation increased. In the case of P. cancrivorus and E. sexcinctus, the best models of roadkill patterns included an area of exposed soil and sparse vegetation, respectively. Roadkill rates were higher in the rainy season for all the mammals, with the exception of C. thous. These results reflect the ecological characteristics of the species with the highest roadkill rates. The findings of the present study raise concerns with regard to the impact of highways on the populations of C. thous, as well as the region’s most threatened species. They also indicate the potential functionality of the local protected area, as well as identifying roadkill hotspots, which will support the development of effective mitigation measures.

Cross-city Landuse classification of remote sensing images via deep transfer learning

This paper describes a deep transfer model which consists of multiple sub-networks which are independently optimized by a supervised task-oriented loss and an unsupervised consistency loss. The former loss function utilizes annotations to accomplish the designed goal. The latter loss prompts the network to learn the target domain data distribution and encourages multiple sub-networks to share learned knowledge. We utilize the proposed model to work with a global local climate zone classification. For the dataset, the source domain includes 352,366 training samples from 42 cities, and the target domain has 48,307 samples from 10 other cities. According to our experiments, the proposed model improves 3.46% overall accuracy and 2.97% average accuracy when compared with other state-of-the-art domain adaptation methodologies. Besides, the classification maps also visualize the outstanding performance of the proposed deep transfer network.

Semantic Identification of Urban Green Spaces: Forest

Abstract. Urban Green Spaces (UGSs) are recognized as crucial parts of the human-nature ecosystem in densely populated urban centers. Even though they have been intensively studied, an ultimate list of all types of UGSs in Europe still does not exist. This challenges decision making on whether an area should be considered an UGS or belong to another land-use class. Furthermore, the means of precise identification of UGSs are dependent, among others, on their type and semantics. Therefore, in this paper, we investigate forests as UGSs and automatically identify them using their distinct characteristics from Sentinel-2 images as well as descriptive properties derived from them, i.e., vegetation indices and texture metrics.We enrich these properties with forest relevant features such as minimum vegetation height and homogeneity. To assess the reliability of the proposed workflow, we test our approach in two German cities and compare the results with existing governmental land use data sets. With the implemented approach we precisely identify over 90% of the existing forests in the study areas. The main restriction of the approach is the transferability of the thresholds of predictor variables such as homogeneity and dissimilarity.

Impacts of Diffuse Land-Use on Plant Diversity Patterns in the Miombo Woodlands of Western Zambia

Land use is known to influence the diversity of vascular plants in the Miombo woodlands. However, little is known about the interaction between soil and land use in herbaceous and woody species. We compared the diversity of vascular plants at the plot level (20 m × 50 m) and site level for three sites in the Miombo woodlands of western Zambia subject to different levels of intensity classes of diffuse land use (e.g., livestock herbivory and selective timber harvesting). For each of the sites, twenty plots were randomly selected for assessment of species composition of vascular plant species, indicators of land-use intensity, and soil chemistry per plot. We hypothesized that the site with the lowest human impact would have the highest richness and diversity of woody and herbaceous species. At the site level, we found that richness and diversity of woody species were unaffected by land-use intensity, whereas herbaceous species richness was higher for the protected site (28 species on average per 1000 m2) than the two other sites (23 and 21 species on average per 1000 m2). At the plot level, herbaceous species richness was positively associated with woodcutting and soil pH. We interpret the positive effect of woodcutting on herbaceous species richness as the effect of lower competition by the woody component for resources such as water, nutrients, and light. With regard to the absence of any effect of land-use intensity on the richness of woody species, we conclude that in our study areas selective timber harvesting may be at a sustainable level and might even have a positive effect on the diversity of the herbaceous layer.