Land Cover Dataset Research Articles

Global and regional impacts of land cover changes on isoprene emissions derived from spaceborne data and the MEGAN model

Abstract. Among the biogenic volatile organic compounds (BVOCs) emitted by plant foliage, isoprene is by far the most important in terms of both global emission and atmospheric impact. It is highly reactive in the air, and its degradation favours the generation of ozone (in the presence of NOx) and secondary organic aerosols. A critical aspect of BVOC emission modelling is the representation of land use and land cover (LULC). The current emission inventories are usually based on land cover maps that are either modelled and dynamic or satellite-based and static. In this study, we use the state-of-the-art Model of Emissions of Gases and Aerosols from Nature (MEGAN) model coupled with the canopy model MOHYCAN (Model for Hydrocarbon emissions by the CANopy) to generate and evaluate emission inventories relying on satellite-based LULC maps at annual time steps. To this purpose, we first intercompare the distribution and evolution (2001–2016) of tree coverage from three global satellite-based datasets, MODerate resolution Imaging Spectroradiometer (MODIS), ESA Climate Change Initiative Land Cover (ESA CCI-LC), and the Global Forest Watch (GFW), and from national inventories. Substantial differences are found between the datasets; e.g. the global areal coverage of trees ranges from 30 to 50×106 km2, with trends spanning from −0.26 to +0.03 % yr−1 between 2001 and 2016. At the national level, the increasing trends in forest cover reported by some national inventories (in particular for the US) are contradicted by all remotely sensed datasets. To a great extent, these discrepancies stem from the plurality of definitions of forest used. According to some local censuses, clear cut areas and seedling or young trees are classified as forest, while satellite-based mappings of trees rely on a minimum height. Three inventories of isoprene emissions are generated, differing only in their LULC datasets used as input: (i) the static distribution of the stand-alone version of MEGAN, (ii) the time-dependent MODIS land cover dataset, and (iii) the MODIS dataset modified to match the tree cover distribution from the GFW database. The mean annual isoprene emissions (350–520 Tg yr−1) span a wide range due to differences in tree distributions, especially in isoprene-rich regions. The impact of LULC changes is a mitigating effect ranging from 0.04 to 0.33 % yr−1 on the positive trends (0.94 % yr−1) mainly driven by temperature and solar radiation. This study highlights the uncertainty in spatial distributions of and temporal variability in isoprene associated with remotely sensed LULC datasets. The interannual variability in the emissions is evaluated against spaceborne observations of formaldehyde (HCHO), a major isoprene oxidation product, through simulations using the global chemistry transport model (CTM) IMAGESv2. A high correlation (R > 0.8) is found between the observed and simulated interannual variability in HCHO columns in most forested regions. The implementation of LULC change has little impact on this correlation due to the dominance of meteorology as a driver of short-term interannual variability. Nevertheless, the simulation accounting for the large tree cover declines of the GFW database over several regions, notably Indonesia and Mato Grosso in Brazil, provides the best agreement with the HCHO column trends observed by the Ozone Monitoring Instrument (OMI). Overall, our study indicates that the continuous tree cover fields at fine resolution provided by the GFW database are our preferred choice for constraining LULC (in combination with discrete LULC maps such as those of MODIS) in biogenic isoprene emission models.

Effect of Using Multi-Year Land Use Land Cover and Monthly LAI Inputs on the Calibration of a Distributed Hydrologic Model

Effective management of water resources entails the understanding of spatiotemporal changes in hydrologic fluxes with variation in land use, especially with a growing trend of urbanization, agricultural lands and non-stationarity of climate. This study explores the use of satellite-based Land Use Land Cover (LULC) data while simultaneously correcting potential evapotranspiration (PET) input with Leaf Area Index (LAI) to increase the performance of a physically distributed hydrologic model. The mesoscale hydrologic model (mHM) was selected for this purpose due to its unique features. Since LAI input informs the model about vegetation dynamics, we incorporated the LAI based PET correction option together with multi-year LULC data. The Globcover land cover data was selected for the single land cover cases, and hybrid of CORINE (coordination of information on the environment) and MODIS (Moderate Resolution Imaging Spectroradiometer) land cover datasets were chosen for the cases with multiple land cover datasets. These two datasets complement each other since MODIS has no separate forest class but more frequent (yearly) observations than CORINE. Calibration period spans from 1990 to 2006 and corresponding NSE (Nash-Sutcliffe Efficiency) values varies between 0.23 and 0.42, while the validation period spans from 2007 to 2010 and corresponding NSE values are between 0.13 and 0.39. The results revealed that the best performance is obtained when multiple land cover datasets are provided to the model and LAI data is used to correct PET, instead of default aspect-based PET correction in mHM. This study suggests that to minimize errors due to parameter uncertainties in physically distributed hydrologic models, adequate information can be supplied to the model with care taken to avoid over-parameterizing the model.

Prototyping a Methodology for Long-Term (1680–2100) Historical-to-Future Landscape Modeling for the Conterminous United States

Land system change has been identified as one of four major Earth system processes where change has passed a destabilizing threshold. A historical record of landscape change is required to understand the impacts change has had on human and natural systems, while scenarios of future landscape change are required to facilitate planning and mitigation efforts. A methodology for modeling long-term historical and future landscape change was applied in the Delaware River Basin of the United States. A parcel-based modeling framework was used to reconstruct historical landscapes back to 1680, parameterized with a variety of spatial and nonspatial historical datasets. Similarly, scenarios of future landscape change were modeled for multiple scenarios out to 2100. Results demonstrate the ability to represent historical land cover proportions and general patterns at broad spatial scales and model multiple potential future landscape trajectories. The resulting land cover collection provides consistent data from 1680 through 2100, at a 30-m spatial resolution, 10-year intervals, and high thematic resolution. The data are consistent with the spatial and thematic characteristics of widely used national-scale land cover datasets, facilitating use within existing land management and research workflows. The methodology demonstrated in the Delaware River Basin is extensible and scalable, with potential applications at national scales for the United States.

Impacts of Infrastructure Developments on Ecosystem Services Potential in Istanbul

Istanbul has been subject to drastic spatial changes in the last decade due to major investments in the northern section of the city. The primary purpose of the research was to interpret the ecological outcomes of spatial use changes due to such urban development activities and enhance the environmental impact assessment process by adopting an analytical and holistic approach. The approach used in the analysis is based on a model called “matrix model” developed specifically to estimate ecosystem services (ES) capacity of a given area depending on the spatial use types. With the use of this method, one can estimate how spatial use influences ES capacity in positive or negative ways, which ES type is affected most, and which part of the spatial change causes most impact on ecological assets. The methodology is applied for three different investment projects in Istanbul, respectively: North Marmara Motorway (NMM), Istanbul Airport, and Canal Istanbul. While the first two have been mostly completed, the third is in the planning phase. Corine Land Cover datasets for 2012 and 2018 are used as basis datasets within the methodology, and spatial plans of Canal Istanbul are used for future projections. Based on these datasets, ES capacity analysis is carried out for each year for the area of investment, and estimation results are compared. The analysis results indicate that these investments have caused and will cause severe ES losses as expected. The outputs prove that the “matrix model” can be used to evaluate ecological impacts in a straightforward, efficient, and inclusive way for evaluating spatial changes. The authors suggest that environmental impact assessments for such major investments must represent a comprehensive outlook based on more simplistic but yet informative approaches that highlight the potential losses of not only ecological assets but also their functions and benefits. In this context, it is strongly recommended that there is a paradigm shift in the understanding of the “environmental impact” in a direction where environment is not acted upon as a solid, rigid, and stable “land cover” but instead a living organism that produces benefits and services for the whole components of the ecosystem on earth, including humankind.

Long-Term Dynamics of Land Use in the Romanian Plain—The Central Bărăgan, Romania

Changes in land use and agricultural landscapes are primarily the result of socio-political and economic changes. This research is based on the analysis of old maps, pertaining to different historical periods, aiming to capture the dynamics of the landscape in the Central Bărăgan Plain. The cartographic materials used have the Map of Southern Romania from 1864, the Firing Master Plans and the Corine Land Cover dataset (1990, 2000, 2018) as reference. For the case studies, these sources are supplemented by the orthophotoplan from 2012 and the remotely-sensed image acquired by Corona satellites in 1974. The results highlight the fact that between 1864 and 2018, the Central Bărăgan Plain recorded important changes in land resources. The most significant transformations took place in the period between 1864 and 1959, when 58% of the plain area was purposed for another use in that large areas of land were introduced in the agricultural use. If in 1864 the agricultural lands represented less than half (196.896 ha) of the Central Bărăgan area, the natural areas being dominant (201.473 ha), in the first half of the 20th century, they increased exponentially (361.674 ha), the natural vegetation occupying much smaller areas (28.481 ha) mainly along the Ialomița and Călmățui rivers and near lakes. This trend is also expressed by the values of the index of naturalness which shows a drastic decrease, from 51.7% in 1864, to 10.6% in 1916–1959. There is also an increase of the area cover by settlements which should be explained by the occurrence of new villages and the increase in size of the existing villages. Another result of the approach concerns the changes that occurred in the agricultural landscape in the middle, respectively at the end of the 20th century, which is a fact quantified by computing the index of land fragmentation. The dynamics of land use and the changes in the features of the agricultural landscape, as far as the analyzed area is concerned, are due to the agrarian reforms pertaining to different historical periods, as well as to the forced relocations during the communist regime. Therefore, the study aims to highlight changes during historical, socio-economic and political time in land use and to reduce the degree of naturalness of the territory at the same time.

Integrating Sentinel Active and Passive Data to Map Land Cover in a National Park from GEE Platform

Objectives: National park is the main body of China's natural protected area system, and land cover mapping in national parks plays an important role in understanding the status of natural resources, identifying existing ecological security threats and responding to them quickly. This analysis attempts to develop an accurate and cost-effective model for mapping land cover dataset in national parks. Methods: Two land cover classification methods were developed based on Google Earth Engine (GEE) environment by combining Sentinel-1 and Sentinel-2 images, topographic and textural derivatives to map land cover types in the Qianjiangyuan National Park. One used pixel-based random forest (RF) classification algorithm, the other used object-oriented simple non-iterative clustering (SNIC) segmentation in partnership with RF algorithm, which indicated that the input featureswere first segmented into superpixel image objects then classified by RF algorithm. To optimize classification features, a method taking the advantages of cloud computing platform to design different groups of comparative experiments to finalize feature combination with the highest classification accuracy was proposed, followed by using a recursive feature elimination method to further screen the features. Results: Experimental results showed that both the pixel-based method and the object-oriented method had good performance in land cover classification, and the corresponding highest overall classification accuracies of the two methods were estimated at 92.37% and 93.98%, respectively. Furthermore, integration of synthetic aperture radar (SAR) data into the classifications could substantially improve the classification accuracy when using the pixel-based method, but there was no apparent escalating effect for the object-oriented classification. Conclusions: Experiments showed that the land cover classification map generated from the SNIC+RF algorithm in GEE platform was more complete and the algorithm requires fewer features (15 features, including multispectral bands and spectral indices) and runs quickly in the GEE platform. Thus, this algorithm deserves to be popularized in national park management practices.

Effects of Cropland Expansion on Temperature Extremes in Western India from 1982 to 2015

India has experienced extensive land cover and land use change (LCLUC). However, there is still limited empirical research regarding the impact of LCLUC on climate extremes in India. Here, we applied statistical methods to assess how cropland expansion has influenced temperature extremes in India from 1982 to 2015 using a new land cover and land use dataset and ECMWF Reanalysis V5 (ERA5) climate data. Our results show that during the last 34 years, croplands in western India increased by ~33.7 percentage points. This cropland expansion shows a significantly negative impact on the maxima of daily maximum temperature (TXx), while its impacts on the maxima of daily minimum temperature and the minima of daily maximum and minimum temperature are limited. It is estimated that if cropland expansion had not taken place in western India over the 1982 to 2015 period, TXx would likely have increased by 0.74 (±0.64) °C. The negative impact of croplands on reducing the TXx extreme is likely due to evaporative cooling from intensified evapotranspiration associated with croplands, resulting in increased latent heat flux and decreased sensible heat flux. This study underscores the important influences of cropland expansion on temperature extremes and can be applicable to other geographic regions experiencing LCLUC.

Extending coverage and thematic resolution of compositional land cover maps in a hierarchical Bayesian framework.

Ecological models are constrained by the availability of high-quality data at biologically appropriate resolutions and extents. Modeling a species' affinity or aversion with a particular land cover class requires data detailing that class across the full study area. Data sets with detailed legends (i.e., high thematic resolution) and/or high accuracy often sacrifice geographic extent, while large-area data sets often compromise on the number of classes and local accuracy. Consequently, ecologists must often restrict their study extent to match that of the more precise data set, or ignore potentially key land cover associations to study a larger area. We introduce a hierarchical Bayesian model to capitalize on the thematic resolution and accuracy of a regional land cover data set, and on the geographic breadth of a large area land cover data set. For the full extent (i.e., beyond the regional data set), the model predicts systematic discrepancies of the large-area data set with the regional data set, and divides an aggregated class into two more specific classes detailed by the regional data set. We illustrate the application of our model for mapping eastern white pine (Pinus strobus) forests, an important timber species that also provides habitat for an invasive shrub in the northeastern United States. We use the National Land Cover Database (NLCD), which covers the full study area but includes only generalized forest classes, and the NH GRANIT land cover data set, which maps White Pine Forest and has high accuracy, but only exists within New Hampshire. We evaluate the model at coarse (20km2 ) and fine (2km2 ) resolutions, with and without spatial random effects. The hierarchical model produced improved maps of compositional land cover for the full extent, reducing inaccuracy relative to NLCD while partitioning a White Pine Forest class out of the Evergreen Forest class. Accuracy was higher with spatial random effects and at the coarse resolution. All models improved upon simply partitioning Evergreen Forest in NLCD based on the predicted distribution of white pine. This flexible statistical method helps ecologists leverage localized mapping efforts to expand models of species distributions, population dynamics, and management strategies beyond the political boundaries that frequently delineate land cover data sets.

Long Time Series High-Quality and High-Consistency Land Cover Mapping Based on Machine Learning Method at Heihe River Basin

Long time series of land cover changes (LCCs) are critical in the analysis of long-term climate, environmental, and ecological changes. Although several moderate to fine resolution global land cover datasets have been publicly released and they show strong consistency at the global scale, they have large deviations at the regional scale; furthermore, high-quality land cover datasets from before 2000 are not available and the classification consistency among different datasets is not very good. Thus, long time series of land cover datasets with high quality and consistency are in great demand but they are still unavailable, even at the regional scale. The Landsat series of satellite imagery composed of eight successive satellites can be traced back to 1972 and it is, therefore, possible to produce a long time series land cover dataset. In addition, the newly available satellite data have the capability to construct time series satellite images and a time series analysis method such as LCMM can be employed for making high-quality land cover datasets. Therefore, by taking the advantages of the two categories of satellite data, we proposed a new time series land cover mapping method based on machine learning and it, thereafter, is applied to Heihe River Basin (HRB) for verification purposes. Firstly, the high-quality land cover datasets at HRB from 2011–2015, which were retrieved using the LCMM method, are used for quickly and accurately making training samples. Secondly, a strategy for transferring the training samples after 2011 to earlier years is established. Thirdly, the random forest model is employed to train the selected yearly samples and a land cover map for every year is subsequently made. Finally, comprehensive analysis and validation are carried out for evaluation. In this study, a long time series land cover dataset including 1986, 1990, 1995, 2000, 2005, 2010, 2011, 2012, 2013, 2014, and 2015 is finally made and an average precision of about 90% is achieved. It is the longest time series land cover map with 30 m resolution at HRB and the dataset has good time continuity and stability.

A landscape classification map of Ireland and its potential use in national land use monitoring

This study presents a novel landscape classification map of the Republic of Ireland and is the first to identify broad landscape classes by incorporating physiographic and land cover data. The landscape classification responds to commitments to identify and classify the Irish landscape as a signatory to the European Landscape Convention. The methodology applied a series of clustering iterations to determine an objective multivariate classification of physiographic landscape units and land cover datasets. The classification results determined nine statistically significant landscape classes and the development of a landscape classification map at a national scale. A statistical breakdown of land cover area and diversity of each class was interpreted, and a comparison was extended using independent descriptive variables including farmland use intensity, elevation, and dominant soil type. Each class depicts unique spatial and composition characteristics, from coastal, lowland and elevated, to distinct and dominating land cover types, further explained by the descriptive variables. The significance of individual classes and success of the classification is discussed with particular reference to the wider applicability of the map. The transferability of the methodology to other existing physiographic maps and environmental datasets to generate new landscape classifications is also considered. This novel work facilitates the development of a strategic framework to efficiently monitor, compare and analyse ecological and other land use data that is spatially representative of the distribution and extent of land cover in the Irish countryside.

Characterizing changes in land cover and forest fragmentation from multitemporal Landsat observations (1993-2018) in the Dhorpatan Hunting Reserve, Nepal

Natural forces and anthropogenic activities greatly alter land cover, deteriorate or alleviate forest fragmentation and affect biodiversity. Thus land cover and forest fragmentation dynamics have become a focus of concern for natural resource management agencies and biodiversity conservation communities. However, there are few land cover datasets and forest fragmentation information available for the Dhorpatan Hunting Reserve (DHR) of Nepal to develop targeted biodiversity conservation plans. In this study, these gaps were filled by characterizing land cover and forest fragmentation trends in the DHR. Using five Landsat images between 1993 and 2018, a support vector machine algorithm was applied to classify six land cover classes: forest, grasslands, barren lands, agricultural and built-up areas, water bodies, and snow and glaciers. Subsequently, two landscape process models and four landscape metrics were used to depict the forest fragmentation situations. Results showed that forest cover increased from 39.4% in 1993 to 39.8% in 2018. Conversely, grasslands decreased from 38.2% in 1993 to 36.9% in 2018. The forest shrinkage was responsible for forest loss during the period, suggesting that the loss of forest cover reduced the connectivity between forest and non-forested areas. Expansion was the dominant component of the forest restoration process, implying that it avoided the occurrence of isolated forests. The maximum value of edge density and perimeter area fractal dimension metrics and the minimum value of aggregation index were observed in 2011, revealing that forests in this year were most fragmented. These specific observations from the current analysis can help local authorities and local communities, who are highly dependent on forest resources, to better develop local forest management and biodiversity conservation plans.

Neighborhood Green Land Cover and Neighborhood-Based Walking in U.S. Older Adults

Greenspace exposure has been associated with physical activity, but few studies have investigated its association with physical activity in the residential neighborhood. This study investigates whether greater amounts of neighborhood open space and forest are associated with neighborhood-based walking in older adults. In 2020, cross-sectional analyses were conducted on those aged ≥65 years from the 2017 National Household Travel Survey. Minutes of neighborhood walking per day were derived from travel diaries. Green land cover measures from the 2011 National Land Cover Dataset were linked to respondent data by the U.S. census tract. Adjusted linear regression models, using weights accounting for survey sampling, tested the associations between the percentage of green land cover in the neighborhood (open space, forest) and minutes of neighborhood walking per day. Adjusted models were stratified to examine whether the associations varied by an individual- and neighborhood-level SES, sex, and race/ethnicity. Respondents (N=72,753) were aged 74 (SD=7) years on average. Greater percentage of open space was associated with more neighborhood walking in African Americans (estimate=0.069, 95% CI=0.005, 0.133). Greater percentage of forest was associated with more neighborhood walking in the overall sample (estimate=0.028, 95% CI=0.006, 0.050), women (estimate=0.025, 95% CI=0.005, 0.045), and Whites (estimate=0.034, 95% CI=0.004, 0.064). Type of neighborhood green land cover (open space versus forest) may be differentially associated with neighborhood walking depending on race/ethnicity. This study suggests a possible association between greater neighborhood open space and greater walking among African Americans that must be confirmed in future studies.

Fully convolutional DenseNet with adversarial training for semantic segmentation of high-resolution remote sensing images

Semantic segmentation is an important and foundational task in the application of high-resolution remote sensing images (HRRSIs). However, HRRSIs feature large differences within categories and minor variances across categories, posing a significant challenge to the high-accuracy semantic segmentation of HRRSIs. To address this issue and obtain powerful feature expressiveness, a deep conditional generative adversarial network (DCGAN), integrating fully convolutional DenseNet (FC-DenseNet) and Pix2pix, is proposed. The DCGAN is composed of a generator–discriminator pair, which is built on a modified downsampling unit of FC-DenseNet. The proposed method possesses strong feature expression ability because of its skip connections, the very deep network structure and multiscale supervision introduced by FC-DenseNet, and the supervision from the discriminator. Experiments on a Deep Globe Land Cover dataset demonstrate the feasibility and effectiveness of this approach for the semantic segmentation of HRRSIs. The results also reveal that our method can mitigate the influence of class imbalance. Our approach for precise semantic segmentation can effectively facilitate the application of HRRSIs.

Modelling the Impacts of Habitat Changes on the Population Density of Eurasian Skylark (Alauda arvensis) Based on Its Landscape Preferences

The dramatic decline of the abundance of farmland bird species can be related to the level of land-use intensity or the land-cover heterogeneity of rural landscapes. Our study area in central Europe (Hungary) included 3049 skylark observation points and their 600 m buffer zones. We used a very detailed map (20 × 20 m minimum mapping unit), the Hungarian Ecosystem Basemap, as a land-cover dataset for the calculation of three landscape indices: mean patch size (MPS), mean fractal dimension (MFRACT), and Shannon diversity index (SDI) to describe the landscape structure of the study areas. Generalized linear models were used to analyze the effect of land-cover types and landscape patterns on the abundance of the Eurasian skylark (Alauda arvensis). According to our findings, the proportions of arable land, open sand steppes, closed grassland patches, and shape complexity and size characteristics of these land cover patches have a positive effect on skylark abundance, while the SDI was negatively associated with the skylark population. On the basis of the used statistical model, the abundance density (individuals/km*) of skylarks could be estimated with 37.77% absolute percentage error and 2.12 mean absolute error. We predicted the skylark population density inside the Natura 2000 Special Protected Area of Hungary which is 0–6 individuals/km* and 23746 ± 8968 skylarks. The results can be implemented for the landscape management of rural landscapes, and the method used are adaptable for the density estimation of other farmland bird species in rural landscapes. According to our findings, inside the protected areas should increase the proportion, the average size and shape complexity of arable land, salt steppes and meadows, and closed grassland land cover patches.

Exploring Climate Change Effects on Vegetation Phenology by MOD13Q1 Data: The Piemonte Region Case Study in the Period 2001–2019

Rising temperature, rainfall, and wind regime changes, increasing of frequency and intensity of extreme events are only some of the effects of climate change affecting the agro-forestry sector. Earth Observation data from satellite missions (often available for free) can certainly support analysis of climate change effects on vegetation, making possible to improve land management in space and time. Within this context, the present work aims at investigating natural and agricultural vegetation, as mapped by Corine Land Cover (CLC) dataset, focusing on phenological metrics trends that can be possibly conditioned by the ongoing climate-change. The study area consists of the entire Piemonte region (NW-Italy). MOD13Q1-v6 dataset from TERRA MODIS mission was used to describe pluri-annual (2001–2019) phenological behavior of vegetation focusing on the following CLC classes: Non-irrigated arable land, Vineyards, Pastures, and Forests. After computing and mapping some phenological metrics as derivable from the interpretation of at-pixel level NDVI (Normalized Difference Vegetation Index) temporal profile, we found that the most significant one was the maximum annual NDVI (MaxNDVI). Consequently, its trend was analyzed at CLC class level for the whole Piemonte region. Natural and semi-natural vegetation classes (Pastures and Forests) were furtherly investigated testing significance of the Percent Total Variation (TV%) of MaxNDVI in the period 2001–2019 for different altitude classes. Results proved that Non-irrigated arable land showed a not significant trend of MaxNDVI; differently, vineyards and forests showed a significant increasing one. Concerning TV%, it was found that it increases with altitude for the Forests CLC class, while it decreases with altitude for the pastures class.

Production of global daily seamless data cubes and quantification of global land cover change from 1985 to 2020 - iMap World 1.0

Longer time high-resolution, high-frequency, consistent, and more detailed land cover data are urgently needed in order to achieve sustainable development goals on food security, high-quality habitat construction, biodiversity conservation and planetary health, and for the understanding, simulation and management of the Earth system. However, due to technological constraints, it is difficult to provide simultaneously high spatial resolution, high temporal frequency, and high quality observation data. Existing mapping solutions are limited by traditional remotely sensed data, that have shorter observation periods, poor spatio-temporal consistency and comparability. Therefore, a new mapping paradigm is needed. This paper develops a framework for intelligent mapping (iMap) of land cover based on state-of-the-art technologies such as cloud computing, artificial intelligence, virtual constellations, and spatio-temporal reconstruction and fusion. Under this framework, we built an automated, serverless, end-to-end data production chain and parallel mapping system based on Amazon Web Services (AWS) and produced the first 30 m global daily seamless data cubes (SDC), and annual to seasonal land cover maps for 1985–2020. The SDC was produced through a multi-source spatio-temporal data reconstruction and fusion workflow based on Landsat, MODIS, and AVHRR virtual constellations. Independent validation results show that the relative mean error of the SDC is less than 2.14%. As analysis ready data (ARD), it can lay a foundation for high-precision quantitative remote sensing information extraction. From this SDC, we produced 36-year long, 30 m resolution global land cover map data set by combining strategies of sample migration, machine learning, and spatio-temporal adjustment. The average overall accuracy of our annual land cover maps over multiple periods of time is 80% for level 1 classification and over 73% for level 2 classification (29 and 33 classes). Based on an objective validation sample consisting of FLUXNET sites, our map accuracy is 10% higher than that of existing global land cover datasets including Globeland30. Our results show that the average global land cover change rate is 0.36%/yr. Global forest decreased by 1.47 million km2 from 38.44 million km2, cropland increased by 0.84 million km2 from 12.49 million km2 and impervious surface increased by 0.48 million km2 from 0.57 million km2 during 1985– 2020.

Landscape Composition Encompassing Cooper's Hawk Nest Sites in Western North Dakota

The Cooper's Hawk (Accipiter cooperii) is often classified as a woodland species, but it exhibits marked plasticity by nesting successfully in what was previously thought to be atypical habitat such as urban areas and sparsely wooded grasslands, including the Little Missouri National Grassland in North Dakota. Forest expansion into North Dakota grasslands is likely due to the reduction of natural disturbances such as wildfires and grazing by bison (Bison bison). From the 1970s to 2001, forest cover increased in McKenzie County, North Dakota, by .1000%. From 1950 to 1972 there were two recorded instances of nesting Cooper's Hawks in McKenzie County. In 2001, we recorded 18 active nests, 13 of which were successful, four that failed, and one with unknown fate. Using the 2001 National Landcover Dataset (NLCD), we calculated that the mean proportion of forest cover in circular plots (18.1 km2) surrounding nests was 0.21. We found no relationship between nest fate and proportion of forest land cover. Additionally, we found no significant difference in proportion of forest cover between nest sites and randomly selected sites centered on forest land cover, indicating that Cooper's Hawks were using forest land cover in proportion to its availability. Understanding the relationship between the shifting landscape composition of North Dakota and its avifauna is important for monitoring and managing breeding Cooper's Hawks.

DFFAN: Dual Function Feature Aggregation Network for Semantic Segmentation of Land Cover

Analyzing land cover using remote sensing images has broad prospects, the precise segmentation of land cover is the key to the application of this technology. Nowadays, the Convolution Neural Network (CNN) is widely used in many image semantic segmentation tasks. However, existing CNN models often exhibit poor generalization ability and low segmentation accuracy when dealing with land cover segmentation tasks. To solve this problem, this paper proposes Dual Function Feature Aggregation Network (DFFAN). This method combines image context information, gathers image spatial information, and extracts and fuses features. DFFAN uses residual neural networks as backbone to obtain different dimensional feature information of remote sensing images through multiple downsamplings. This work designs Affinity Matrix Module (AMM) to obtain the context of each feature map and proposes Boundary Feature Fusion Module (BFF) to fuse the context information and spatial information of an image to determine the location distribution of each image’s category. Compared with existing methods, the proposed method is significantly improved in accuracy. Its mean intersection over union (MIoU) on the LandCover dataset reaches 84.81%.

On the runoff validation of ‘Global BROOK90’ automatic modeling framework

Abstract The recently presented Global BROOK90 automatic modeling framework combines a non-calibrated lumped hydrological model with ERA5 reanalysis data as the main driver, as well as with global elevation, land cover and soil datasets. The focus is to simulate the water fluxes within the soil–water–plant system of a single plot or of a small catchment especially in data-scarce regions. The comparison to runoff is an obvious choice for the validation of this approach. Thus, we choose for validation 190 small catchments (with a median size of 64 km2) with discharge observations available within a time period of 1979–2020 and located all over the globe. They represent a wide range of relief, land cover and soil types within all climate zones. The simulation performance was analyzed with standard skill-score criteria: Nash–Sutcliffe Efficiency, Kling–Gupta Efficiency, Kling–Gupta Efficiency Skill Score and Mean Absolute Error. Overall, the framework performed well (better than mean flow prediction) in more than 75% of the cases (KGESS > 0) and significantly better on a monthly rather than on a daily scale. Furthermore, it was found that Global BROOK90 outperforms GloFAS-ERA5 discharge reanalysis. Additionally, cluster analysis revealed that some of the catchment characteristics have a significant influence on the framework performance.

High agricultural water consumption led to the continued shrinkage of the Aral Sea during 1992-2015.

The shrinkage of the Aral Sea started in the 1960s, and it has been continued for decades due to arguably both human and natural causes. However, the change of the Aral Sea in the post-Soviet era and its correlations with other changes in the extent of the basin have yet to be fully investigated. Here, we studied the land cover dynamics of the entire Aral Sea basin during 1992-2015 from the perspective of the surrounding environment, in order to investigate the causes of the Aral Sea further shrinkage in recent years. We used the annual Climate Change Initiative (CCI) land cover dataset to provide a spatiotemporally consistent delineation of land cover throughout the period. We found that: (1) In recent years, the Aral Sea continued shrinkage, approximately 50.38% of its water area in 1992 had dried out and turned into bare land by 2015. (2) The cultivated land area remained stable with a slight increase during the period, suggesting that no large-scale abandonment or expansion of farming extent occurred in the post-Soviet era. (3) Among other land types, urban areas are small and slightly expand at a rate of 0.024×104km2/year, suggesting urbanization, and likely contribute to more water consumption. Our investigation also found that climate warming increased the upstream runoff, which has a positive effect on the water supply of the Aral Sea. The impact of human activity on the Aral Sea is more pronounced than climate change. Therefore, the continued shrinkage of the Aral Sea was likely due to high water consumption of agriculture continues to exert the influence that existed in the 1960s. Other factors, such as urbanization have exacerbated this effect. The study examined the continued shrinkage of the Aral Sea in post-Soviet era, to provide an insight into the driving factors of the complex and still controversial Aral Sea crisis.