Land Cover Dataset Research Articles

Comparisons of two global built area land cover datasets in methods to disaggregate human population in eleven countries from the global South

ABSTRACTMapping built land cover at unprecedented detail has been facilitated by increasing availability of global high-resolution imagery and image processing methods. These advances in urban feature extraction and built-area detection can refine the mapping of human population densities, especially in lower income countries where rapid urbanization and changing population is accompanied by frequently out-of-date or inaccurate census data. However, in these contexts it is unclear how best to use built-area data to disaggregate areal, count-based census data. Here we tested two methods using remotely sensed, built-area land cover data to disaggregate population data. These included simple, areal weighting and more complex statistical models with other ancillary information. Outcomes were assessed across eleven countries, representing different world regions varying in population densities, types of built infrastructure, and environmental characteristics. We found that for seven of 11 countries a Random Forest-based, machine learning approach outperforms simple, binary dasymetric disaggregation into remotely-sensed built areas. For these more complex models there was little evidence to support using any single built land cover input over the rest, and in most cases using more than one built-area data product resulted in higher predictive capacity. We discuss these results and implications for future population modeling approaches.

Assessment of Water Balance for Russian Subcatchment of Western Dvina River Using SWAT Model

The study provides a new assessment of the water balance components of the catchment (evapotranspiration, surface and lateral flow etc. and its spatial distribution and temporal variability) for the transboundary catchment of Western Dvina river within the poorly gauged Russian part of the catchment.The study focuses on modeling the inland flow generation processes using open source data and the SWAT (Soil Water Assessment Tool) hydrological model. The high interannual variability of river flow and impact of snowmelt processes were especially taken into account when setting up the model and processing the calibration. The database of daily meteorological data for the period 1981 – 2016 was prepared using global atmospheric reanalysis ERA-Interim data and observed station data from the GSOD NCDC/NOAA and ECA&D datasets. The considered datasets were tested on plausibility and regionalized. The catchment model was built on the basis of open land use / land cover (LULC) data sets, topography and soil, so that the entire transboundary catchment area could be easily implemented in the next step. For the daily model calibration, 19 sensitive parameters were chosen manually. The most sensitive are the parameters which consider snow melting processes and flow recession curve number. The highest impact on water balance components has the area and distribution of wetlands. Lakes strongly affect the evapotranspiration rate. The study provides further research with uncertainty analysis and recommendations for model improvement and model limitations. The developed modelling approach can be used to assess water resources, climate change impacts, and water quality issues in comparable regions.

Extensive land cover change across Arctic-Boreal Northwestern North America from disturbance and climate forcing.

A multitude of disturbance agents, such as wildfires, land use, and climate-driven expansion of woody shrubs, is transforming the distribution of plant functional types across Arctic-Boreal ecosystems, which has significant implications for interactions and feedbacks between terrestrial ecosystems and climate in the northern high-latitude. However, because the spatial resolution of existing land cover datasets is too coarse, large-scale land cover changes in the Arctic-Boreal region (ABR) have been poorly characterized. Here, we use 31years (1984-2014) of moderate spatial resolution (30m) satellite imagery over a region spanning 4.7×106 km2 in Alaska and northwestern Canada to characterize regional-scale ABR land cover changes. We find that 13.6±1.3% of the domain has changed, primarily via two major modes of transformation: (a) simultaneous disturbance-driven decreases in Evergreen Forest area (-14.7±3.0% relative to 1984) and increases in Deciduous Forest area (+14.8±5.2%) in the Boreal biome; and (b) climate-driven expansion of Herbaceous and Shrub vegetation (+7.4±2.0%) in the Arctic biome. By using time series of 30m imagery, we characterize dynamics in forest and shrub cover occurring at relatively short spatial scales (hundreds of meters) due to fires, harvest, and climate-induced growth that are not observable in coarse spatial resolution (e.g., 500m or greater pixel size) imagery. Wildfires caused most of Evergreen Forest Loss and Evergreen Forest Gain and substantial areas of Deciduous Forest Gain. Extensive shifts in the distribution of plant functional types at multiple spatial scales are consistent with observations of increased atmospheric CO2 seasonality and ecosystem productivity at northern high-latitudes and signal continental-scale shifts in the structure and function of northern high-latitude ecosystems in response to climate change.

Estimating Barcelona's metropolitan daytime hot and cold poles using Landsat-8 Land Surface Temperature

The Barcelona Metropolitan Area (BMA) is located in Catalonia, northeastern Spain. With a population of over 3 billion people, the BMA is one of the most populous metropolitan areas on the Mediterranean coast. A local climatic modification known as the urban heat island (UHI) occurs in the urban areas. The UHI is usually quantified by means of air temperature, although remote sensing can be used to extract a thermal image of the earth's surface to provide temperature values throughout the study area. Estimation of the land surface temperature (LST) for the BMA enabled us to establish the spatial patterns of LST and to detect the poles of heat and cold within the BMA on 24 dates during the 2013–2018 period, distributed among the 4 seasons of the year. To this end we performed a principal component analysis (PCA) and a cluster analysis (CA). Moreover, we employed the Random Forest (RF) regression method to quantify the influence and variation of diverse geographic covariates according to season and location in the study area. Finally, to determine the influence of land covers on temperature, the thermal values of the 4 land covers included in the Corine Land Cover dataset were analyzed: industrial units, continuous urban fabric, green urban areas, and forest areas. Results show that the heat poles are concentrated in industrial areas primarily, followed by urban fabric areas. On the contrary, the cold pole is found in green urban areas, as well as forested areas. The maximum temperature range between land covers was detected in spring and summer, while in winter this difference was negligible. Our study showed that green urban areas presented temperatures up to 2.5 °C lower than in urban areas. The results of the present research are intended to serve as a roadmap for enhancing thermal comfort in the BMA.

Slower vegetation greening faced faster social development on the landscape of the Belt and Road region

The thinking of landscape sustainability discussed the win-win goal of landscape management for both ecological conservation and social development. The China-proposed Belt and Road Initiative has a pathway of green development towards sustainability, which could be benefited from a corporation on transnational landscape management. Although previous remote sensing observations showed that the ecosystem improvement and social development can coexist in most of the Belt and Road region, whether the change tendencies can be sustained have been neglected. Based on the continuous vegetation index, nighttime light and landcover datasets between 1980s–2010s, a detailed observation of the landscape evolution was obtained in this study. Depending on the identifications of temporal breakpoints, the time series was separated to identify the former and latter tendencies, and the evolution score were evaluated. The results showed that the vegetation greening trend weakened in nearly half of the region, and the arid region mostly experienced vegetation browning lately. The evolution patterns of ecological landscape and social landscape were coincided distributed in China and India, and in Southeast Asia was not scored high for both ecological conservation and social development. Consequently, we should not be overly optimistic about the synergy between ecosystem improvement and social development.

Contrasting Trends of Forest Coverage between the Inland and Coastal Urban Groups of China over the Past Decades

China is building forest urban groups through reforestation and afforestation. However, the fast process of urbanization inevitably conflicts with multiple vegetated areas around cities. Hence, it is critical to evaluate the changes in regional vegetation cover and its spatial pattern due to complex natural and anthropogenic factors. Nevertheless, systematic studies to quantify and compare the development of forest urban agglomerations were rarely reported. Based on a remote sensing landcover dataset from 1992 to 2015, this study investigated forest cover changes and the impacts on landscape pattern in several urban groups, and tried to explore their differences between the inland and coastal regions of China. The results showed that over the past 24 years, the forest coverage in the coastal urban agglomerations declined (103 km2/year) while it increased (26 km2/year) in the inland urban agglomerations. There was a certain conflict between forest and cropland for the coastal urban agglomerations where the forest area converted to cropland accounted for 61.9% of the total forest loss. The increase in forests coverage in inland urban agglomerations mainly came from grassland which nearly accounted for 66.47% of the total increase. The landscape diversity has also changed in areas where forests have changed significantly (e.g., Shanghai, Changzhi, and Jincheng).

GCN250, new global gridded curve numbers for hydrologic modeling and design

The USDA curve-number (CN) method is fundamental for rainfall-runoff modeling. A global CN database is not currently available for geospatial hydrologic analysis at a resolution higher than 0.1°. We developed a globally consistent, gridded dataset defining CNs at the 250 m spatial resolution from new global land cover (300 m) and soils data (250 m). The resulting data product – GCN250 – represents runoff for a combination of the European space agency global land cover dataset for 2015 (ESA CCI-LC) resampled to 250 m and geo-registered with the hydrologic soil group global data product (HYSOGs250m) released in 2018. Our analysis indicated that medium to high runoff potential currently dominates the globe, with curve numbers ranging between 75 and 85. Global curve numbers were 62, 78, and 90 for dry, average, and wet antecedent runoff conditions, respectively. Australia has the highest runoff potential, while Europe has the lowest. Runoff ratios compare well with GLDAS. The potential application of this data includes hydrologic design, land management applications, flood risk assessment, and groundwater recharge modeling.

Lessons learned implementing an operational continuous United States national land change monitoring capability: The Land Change Monitoring, Assessment, and Projection (LCMAP) approach

Growing demands for temporally specific information on land surface change are fueling a new generation of maps and statistics that can contribute to understanding geographic and temporal patterns of change across large regions, provide input into a wide range of environmental modeling studies, clarify the drivers of change, and provide more timely information for land managers. To meet these needs, the U.S. Geological Survey has implemented a capability to monitor land surface change called the Land Change Monitoring, Assessment, and Projection (LCMAP) initiative. This paper describes the methodological foundations and lessons learned during development and testing of the LCMAP approach. Testing and evaluation of a suite of 10 annual land cover and land surface change data sets over six diverse study areas across the United States revealed good agreement with other published maps (overall agreement ranged from 73% to 87%) as well as several challenges that needed to be addressed to meet the goals of robust, repeatable, and geographically consistent monitoring results from the Continuous Change Detection and Classification (CCDC) algorithm. First, the high spatial and temporal variability of observational frequency led to differences in the number of changes identified, so CCDC was modified such that change detection is dependent on observational frequency. Second, the CCDC classification methodology was modified to improve its ability to characterize gradual land surface changes. Third, modifications were made to the classification element of CCDC to improve the representativeness of training data, which necessitated replacing the random forest algorithm with a boosted decision tree. Following these modifications, assessment of prototype Version 1 LCMAP results showed improvements in overall agreement (ranging from 85% to 90%).

Impact of the Accuracy of Land Cover Data sets on the Accuracy of Land Cover Change Scenarios in the Mono River Basin, Togo, West Africa

Knowledge about land use and land cover (LULC) dynamics is of high importance for a number of environmental studies including the development of water resources, land degradation and food security. Often, available global or regional data sets are used for impact studies, although they have not been validated for the area of interest. Validation is especially required if data are used to set up a land change model predicting future changes for management purposes. Therefore, three different LULC maps of the Mono River Basin in Togo were evaluated in this study. The analyzed maps were obtained from three sources: CILSS (2 km resolution), ESA (300 m), and Globeland (30m) datasets. Validation was performed using 1,000 reference points in the watershed derived from satellite images. The results reveal CILSS as the most accurate data set with a Kappa coefficient of 68% and an overall accuracy of 83%. CILSS data shows a decrease of savanna and forest whereas an increase of cropland over the period 1975 to 2013. The increase of cropland area of 30.97% from 1975 to 2013 can be related to the increase in population and their food demand, while the losses of forest area and the decrease of savanna are further amplified by using wood as energy sources and the lack of forest management. The three datasets were used to simulate future LULC changes using the Terrset Land Change Modeler. The validation of the model using CILSS data for 2013 showed a quality of 50.94%, it is only 40.04% for ESA and 20.13% for Globeland30. CILSS data was utilized to simulate the LULC distribution for the years 2020 and 2027 because of its satisfactory performances. The results show that a high spatial resolution is not a guarantee of high quality. The results of this study can be used for impact studies and to develop management strategies for mitigating negative effects of land use and land cover change.

Effectiveness of Natura 2000 areas for environmental protection in 21 European countries

To assess the dynamics of changes in CORINE Land Cover classes in areas of the Natura 2000 ecological network, three landscape metrics were examined. Traditional pixel-based classification, with majority rules aggregation based on the example of the CORINE Land Cover dataset, was applied. To identify the possible differences in the considered metrics, according to the year as well as to the country, statistical analysis between the linear mixed model and the variance model with repeated measurements was performed. The results of both tests are very similar. In the 1990–2012 period, the share of CORINE Land Cover class “Artificial areas” in all tested areas increased by 21.1% (the highest growth ratios were recorded in Poland, Portugal, Slovenia and in Spain). On the other hand, such countries as Slovakia, Romania, Germany, Lithuania and Estonia are characterised by the loss of artificial areas. At the same time, the share of “Forest and semi-natural areas” as well as “Water bodies” increased slightly. Negative trends that took place in the periods 1990–2000 and 2000–2006 were effectively stopped in the subsequent period, 2006–2012. Overall, for all the analysed countries, a minimal loss in environmental and landscape diversity was observed. Our results may be used as a basis for drawing conclusions on the effectiveness of environmental and landscape management systems in various countries. They might also constitute the starting point for detailed analysis of the management process.

An Integrated Land Cover Mapping Method Suitable for Low-Accuracy Areas in Global Land Cover Maps

In land cover mapping, an area with complex topography or heterogeneous land covers is usually poorly classified and therefore defined as a low-accuracy area. The low-accuracy areas are important because they restrict the overall accuracy (OA) of global land cover classification (LCC) data generated. In this paper, low-accuracy areas in China (extracted from the MODIS global LCC maps) were taken as examples, identified as the regions having lower accuracy than the average OA of China. An integrated land cover mapping method targeting low-accuracy regions was developed and tested in eight representative low-accuracy regions of China. The method optimized procedures of image choosing and sample selection based on an existent visually-interpreted regional LCC dataset with high accuracies. Five algorithms and 16 groups of classification features were compared to achieve the highest OA. The support vector machine (SVM) achieved the highest mean OA (81.5%) when only spectral bands were classified. Aspect tended to attenuate OA as a classification feature. The optimal classification features for different regions largely depends on the topographic feature of vegetation. The mean OA for eight low-accuracy regions was 84.4% by the proposed method in this study, which exceeded the mean OA of most precedent global land cover datasets. The new method can be applied worldwide to improve land cover mapping of low-accuracy areas in global land cover maps.

Changes in the extent and distribution of urban land cover in the Democratic People's Republic of Korea (North Korea) between 1987 and 2010

AbstractLargely due to data unavailability, the spatial pattern of urban growth in North Korea has been rarely studied. This study explored urban changes in North Korea and provided their plausible causes. The present study used satellite‐based land cover data sets produced by the government of South Korea to examine the extent and distribution of urban land cover in North Korea between the late 1980s (1987–1989) and late 2000s (2008–2010) at the municipal level. The urban expansion intensity index (UEII) was calculated for two 10‐year intervals, and the spatial autocorrelation of UEII values was examined. Major findings from the study are summarized as follows: (a) the capital city Pyongyang's dominance continued without any obvious sign of slowing down, (b) economic development districts do not appear to have much influence on urban land cover changes, and (c) the extent of urban land cover slightly decreased in much of the country between the late 1990s (1997–1999) and late 2000s (2008–2010). The results and discussion in this study suggest internal migration and stagnant economy as probably important causes for the phenomena.

Visualisation and analysis of multi-resolution raster geodatasets in the cloud

Abstract. Raster geodatasets are an important data source for map-related web applications. However, existing differences in how those datasets are georeferenced, formatted and made available complicate their integration for analysis and visualisation purposes. Furthermore, the inconsistent cell resolutions make it difficult to perform efficient and reliable display of the datasets over a range of scales. In particular, it is challenging to achieve satisfactory user experience in visual exploration of interactive analysis operations carried out on these datasets.Dataset harmonisation carried out as a pre-processing procedure seems to be necessary for resolving the raster geodata inconsistency challenge. After the pre-processing step, the harmonized datasets have to be stored into an easily accessible data storage, preferably in close connection with the computing platform where the analysis operations are to be performed.Multidimensional datacube has emerged as a conceptual framework for organising a repository of harmonised raster geodatasets. Satellite images have so far been the most popular geospatial application area for datacubes. The set of dimensions of such geodatacube typically consist of two or three coordinate axes, time, and the content theme. While being ingested into the datacube, the involved datasets are brought into a common georeferencing frame, series of cell resolutions, and storage format. The selected harmonised resolution levels can actually be seen as a set of individual geodatacubes.An initiative has launched in Finland to build a harmonised multi-resolution geodatacube, called GeoCubes Finland, containing some of the most important national geodatasets. The datasets currently incorporated into the geodatacube include digital elevation models together with land cover, superficial deposit, forest inventory and administrative unit datasets. GeoCubes Finland is being built as part of the Finnish Open Geospatial Information Infrastructure for Research (oGIIR) programme, aimed at developing the geospatial research infrastructure of the country. The data storage and the related computing facilities have been implemented on a cloud service platform provided by CSC - IT Center for Science that serves Finnish universities and research institutes.

Response of net primary productivity to vegetation restoration in Chinese Loess Plateau during 1986-2015.

Land use and land cover change induced by large scale ecological restoration programs has a significant impact on the terrestrial ecosystem carbon cycle, especially on the net primary productivity (NPP) in arid and semi-arid regions. This study investigated the change in NPP caused by the large-scale ecological restoration in the Chinese Loess Plateau (LPR) region from 1986 to 2015 based on land cover datasets and NPP calculated using the Carnegie-Ames-Stanford Approach model. The results indicated that the annual total NPP exhibited a significant uptrend (P < 0.01) throughout the whole vegetation restoration region during the last 30 years, with an annual increase of 0.137 Tg C. A significant abrupt change was detected in 2006 for the annual total NPP series. Over half of the restoration region showed an increase in NPP in the past three decades, however, about 30~40% of the vegetation restoration region exhibited NPP loss before 2006, but subsequently NPP loss was found in only approximately 20% of the study region. Overall, the increase in NPP attributed to the vegetation restoration reached 51.14 Tg C in the past three decades, indicating that these large-scale vegetation restoration programs increased the carbon sequestration capacity of terrestrial ecosystems in the Loess Plateau. The findings of this study improve our understanding of the effects of the green campaign on terrestrial ecosystems.

A global land cover map produced through integrating multi-source datasets

ABSTRACTIn the past decades, global land cover datasets have been produced but also been criticized for their low accuracies, which have been affecting the applications of these datasets. Producing a new global dataset requires a tremendous amount of efforts; however, it is also possible to improve the accuracy of global land cover mapping by fusing the existing datasets. A decision-fuse method was developed based on fuzzy logic to quantify the consistencies and uncertainties of the existing datasets and then aggregated to provide the most certain estimation. The method was applied to produce a 1-km global land cover map (SYNLCover) by integrating five global land cover datasets and three global datasets of tree cover and croplands. Efforts were carried out to assess the quality: 1) inter-comparison of the datasets revealed that the SYNLCover dataset had higher consistency than these input global land cover datasets, suggesting that the data fusion method reduced the disagreement among the input datasets; 2) quality assessment using the human-interpreted reference dataset reported the highest accuracy in the fused SYNLCover dataset, which had an overall accuracy of 71.1%, in contrast to the overall accuracy between 48.6% and 68.9% for the other global land cover datasets.

Lyme borreliosis incidence in Lombardy, Italy (2000–2015): Spatiotemporal analysis and environmental risk factors

Lyme borreliosis cases have been reported from Lombardy in northern Italy, where Ixodes ricinus is the main vector of Borrelia burgdorferi sensu lato. However, spatial and temporal variation in the incidence of Lyme borreliosis is not well understood. In the present study, based on new notified cases of Lyme borreliosis from 2000 to 2015, an average of 1.24 new cases per million residents per year was documented. New cases, georeferenced at the municipal level, were analyzed by retrospective space-time analysis (using SaTScan v. 9.3.1); and land cover, extrapolated from a Corine Land Cover dataset (using QGIS 2.8.1), was used to implement an environmental risk factor analysis. Firstly, a temporal high-risk cluster was detected in Lombardy: the relative risk of Lyme borreliosis was 3.73 times higher during 2008–2015 compared with the entire study period. Moreover, in a spatiotemporal high-risk cluster with a circular base, land cover consisting of wildland-urban interface, meadow, forest and meadow-forest transition were significantly higher compared to low-risk areas. Results of the present study demonstrate that the incidence of Lyme borreliosis is increasing in Lombardy and that environmental conditions are suitable for I. ricinus ticks infected with B. burgdorferi s.l.: citizens and health systems should be aware of Lyme borreliosis to reduce tick bites with personal protective behaviors and to avoid misdiagnosis, particularly within the area including the observed high-risk cluster. Economic resources should be invested to inform about methods to prevent tick bites, how to check people and pets after frequenting risk areas, and ways of removing the biting ticks when they are found.

Influence of urban land cover data uncertainties on the numerical simulations of urbanization effects in the 2013 high-temperature episode in Eastern China

Updating the urban land cover information has been proved a necessary method for the numerical studies of urban climate and urban atmospheric environment in China, a fast urbanizing country. However, there are uncertainties in the urban land use/cover (ULUC) information in different datasets due to the uncertainties in raw data sources and produce methods. In this study, the Weather Research and Forecasting model is used to simulate the summer climate over the Yangtze River Delta in July and August 2013, when a heatwave episode occurred. Five numerical experiments are designed; one is the baseline experiment (non-urban land cover (NOURB)) in which all urban land covers are removed, and different urban land use/cover (ULUC) experiments from four different global land cover datasets are used in other four experiments. The differences between the ULUC experiments and the NOURB experiment are used to observe the urbanization effects. The results show a remarkable urbanization effect in all the ULUC experiments, and the urbanization influences are most different over the union-urban area, which defined as the area marked as urban area by at least one land cover dataset, because the uncertainties in urban land cover information are greatest in this area. All the ULUC experiments show the regional influence of urbanization as well, which means urbanization effects are not limited in cities. The Climate Change Initiative (CCI) experiment using the European Space Agency CCI 2013 dataset showed the greatest urbanization influence and the highest effect index in all analyzed variables, because of the largest urban land cover area in the analysis domain. The urban land cover area in the Global Land Cover experiment is the smallest among the ULUC experiments, but it distributes in a more concentrated way than in other experiments; it indicates that both city size and city shape affect the regional influence.

Analysis of Land Use Land Cover in Annapurna Conservation Area in Gandaki Province, Nepal using Vegetation Index and Land Use Land Cover Datasets

Conservation areas are originally well-known for protecting landscape features and wildlife. They are playing key role in conserving and providing a wide range of ecosystem services, social, economic and cultural benefits as well as vital places for climate mitigation and adaptation. We have analyzed decadal changes in land cover and status of vegetation cover in the conservation area using both national level available data on land use land cover (LULC) changes (1990-2010) and normalized difference vegetation index (NDVI) (2010-2018) in Annapurna conservation area. LULC showed the barren land as the most dominant land cover types in all three different time series 1990, 2000 and 2010 with followed by snow cover, grassland, forest, agriculture and water body. The highest NDVI values were observed at Southern, Southwestern and Southeastern part of conservation area consisting of forest area, shrub land and grassland while toward low to negative in the upper middle to the Northern part of the conservation area.

Relationships between regional coastal land cover distributions and elevation reveal data uncertainty in a sea-level rise impacts model

Abstract. Understanding land loss or resilience in response to sea-level rise (SLR) requires spatially extensive and continuous datasets to capture landscape variability. We investigate the sensitivity and skill of a model that predicts dynamic response likelihood to SLR across the northeastern US by exploring several data inputs and outcomes. Using elevation and land cover datasets, we determine where data error is likely, quantify its effect on predictions, and evaluate its influence on prediction confidence. Results show data error is concentrated in low-lying areas with little impact on prediction skill, as the inherent correlation between the datasets can be exploited to reduce data uncertainty using Bayesian inference. This suggests the approach may be extended to regions with limited data availability and/or poor quality. Furthermore, we verify that model sensitivity in these first-order landscape change assessments is well-matched to larger coastal process uncertainties, for which process-based models are important complements to further reduce uncertainty.

Does Quantification of Ecosystem Services Depend Upon Scale (Resolution and Extent)? A Case Study Using the InVEST Nutrient Delivery Ratio Model in Georgia, United States

Modeling ecosystem services (ESs) intrinsically involves the use of spatial and temporal data. Correct estimates of ecosystem services are inherently dependent upon the scale (resolution and extent) of the input spatial data. Sensitivity of modeling platforms typically used for quantifying ESs to simultaneous changes in the resolution and extent of spatial data is not particularly clear at present. This study used the Nutrient Delivery Ratio (NDR) model embedded in InVEST (Integrated Valuation of Ecosystem Services and Tradeoffs) for ascertaining the sensitivity of the outputs to three digital elevation models (DEM), two land cover datasets, and three precipitation grids for 57 watersheds located in Georgia, United States. Multivariate regression models were developed to verify the influence of the spatial resolution of input data on the NDR model output at two spatial extents (the state of Georgia and six physiographical regions within the state). Discrepancies in nutrient exports up to 77.4% and 168.1% were found among scenarios at the state level for nitrogen and phosphorus, respectively. Land cover datasets differing in resolution were responsible for the highest differences in nutrient exports. Significance (at 5% level) of spatial variables on the model outputs were different for the two spatial extents, demonstrating the influence of scale when modeling nutrient runoff and its importance for better policy prescriptions.