Regional Land Cover Research Articles

High-resolution wetness index mapping: A useful tool for regional scale wetland management

Wetland ecosystems are key habitats for carbon sequestration, biodiversity and ecosystem services, yet in many they localities have been subject to modification or damage. In recent years, there has been increasing focus on effective management and, where possible, restoration of wetlands. Whilst this is highly laudable, practical implementation is limited by the high costs and unpredictable rates of success. Accordingly, there is a need for spatial information to guide restoration, ideally at the regional scale that land managers operate. In this study, we use high-resolution Light Detection and Ranging (LiDAR)-derived elevation, in conjunction with regional soil and land cover maps, to model the wetness potential of an area of conservation importance in north-west England. We use the Compound Topographic Index (CTI) as a measure for the site-specific wetness and potential to be receptive to wetland restoration. The resulting model is in agreement with the regional-scale distribution of wetlands and is clearly influenced by the topographic and soil parameters. An assessment of three representative case studies highlights the small scale features that determine the potential wetness of an area. For each site, the model results conform to the expected patterns of wetness, highlighting restoration and management activity. Furthermore, areas showing high potential wetness that may be suitable for wetland habitat creation, are highlighted. The increasing availability of LiDAR data at regional and national scales will allow studies of this nature to be undertaken at previously unobtainable resolutions. Simple models, such as implemented here, benefit from explainability and relatability and have clear potential for use by managers and conservation agencies involved in wetland restoration.

Spatial downscaling of MODIS Chlorophyll-a using Landsat 8 images for complex coastal water monitoring

Chlorophyll-a (Chl-a) is an indicator for monitoring the trophic status of coastal water. However, its sampling and analysing in situ can be time consuming and spatially limited. With high temporal resolution and global coverage, Moderate Resolution Imaging Spectroradiometer (MODIS) Level 3 products are reliable and widely used for ocean ecosystem monitoring. The Landsat images can provide detailed spatial information for regional scale land cover and land change studies. The purpose of this study is to combine the 4 km resolution MODIS Chl-a data with 30 m resolution Landsat data to obtain high spatial resolution Chl-a maps for complex coastal water quality monitoring. A data fusion model of MODIS Chl-a and Landsat 8 OLI sensor data was built up for the east coast of the Laizhou Bay, China. The model is an optimal regression model of the band ratio OLI+2/OLI+5 and MODIS Chl-a. With residual correction, it managed to downscale the MODIS Chl-a data from resolution 4 km to 30 m. Validation were provided from in situ Chl-a measured in spring and autumn of 2013 and 2014. The model yields satisfactory results and the derived Chl-a map presents detailed information of coastal trophic status, which will be useful for regional water quality monitoring.

Mapping of forest alliances with simulated multi-seasonal hyperspectral satellite imagery

A consistent and hierarchical classification of vegetation, such as the U.S. National Vegetation Classification (NVC) system, supports comprehensive conservation and management of natural ecosystems. At a detailed level, the NVC alliance is defined by diagnostic species and composition. Maps at this level of classification are often produced at local to regional scales (areas <25,000 km2) with costly manual to semi-automated interpretation of high resolution imagery. The main objective of this study was to assess the effectiveness of machine learning for automated, per-pixel (30 m) mapping of forest alliances with multi-seasonal hyperspectral imagery from a future satellite mission (HyspIRI), as simulated from Airborne Visible/Infrared Imaging Spectrometer Classic (AVIRIS-C) data. The study area was the San Francisco (S.F.) Bay Area, California, where we mapped forest alliances at regional and county scales. We implemented the Support Vector Machine (SVM) classifier in a two-stage approach, first mapping regional land cover followed by forest alliances in closed-canopy tree pixels. Predictor variables were reflectance bands and hyperspectral metrics based on indices, derivatives and absorption-fitting techniques applied to reflectance spectra, with data grouped into summer and three-season (spring, summer, fall) sets. For forest alliances, hyperspectral metrics improved overall accuracy of classifications by 2.9 to 6.4% relative to classifications based on the original reflectance bands. Multi-seasonal data improved overall accuracy by 1.3 to 6.2% relative to summer-only data. Using multi-seasonal metrics, the S.F. Bay Area regional classification with 21 alliances had an overall accuracy of 65.7% (Kappa 0.63), while the Sonoma County classification with 16 alliances had an accuracy of 75.9% (Kappa 0.72). Most forest alliances had internal variation in lifeform, species and structural properties that increased within-class spectral-temporal variation and complicated discrimination. Despite this challenge, classification accuracies were similar to regional NVC alliance reference data. We conclude that a hyperspectral satellite, with its repeat and global image acquisitions, has strong potential for accurate and economical mapping and monitoring the Earth's vegetation communities.

Fine Resolution Probabilistic Land Cover Classification of Landscapes in the Southeastern United States

Land cover classification provides valuable information for prioritizing management and conservation operations across large landscapes. Current regional scale land cover geospatial products within the United States have a spatial resolution that is too coarse to provide the necessary information for operations at the local and project scales. This paper describes a methodology that uses recent advances in spatial analysis software to create a land cover classification over a large region in the southeastern United States at a fine (1 m) spatial resolution. This methodology used image texture metrics and principle components derived from National Agriculture Imagery Program (NAIP) aerial photographic imagery, visually classified locations, and a softmax neural network model. The model efficiently produced classification surfaces at 1 m resolution across roughly 11.6 million hectares (28.8 million acres) with less than 10% average error in modeled probability. The classification surfaces consist of probability estimates of 13 visually distinct classes for each 1 m cell across the study area. This methodology and the tools used in this study constitute a highly flexible fine resolution land cover classification that can be applied across large extents using standard computer hardware, common and open source software and publicly available imagery.

Predicting risk of trace element pollution from municipal roads using site-specific soil samples and remotely sensed data

Studies of environmental processes exhibit spatial variation within data sets. The ability to derive predictions of risk from field data is a critical path forward in understanding the data and applying the information to land and resource management. Thanks to recent advances in predictive modeling, open source software, and computing, the power to do this is within grasp. This article provides an example of how we predicted relative trace element pollution risk from roads across a region by combining site specific trace element data in soils with regional land cover and planning information in a predictive model framework. In the Kenai Peninsula of Alaska, we sampled 36 sites (191 soil samples) adjacent to roads for trace elements. We then combined this site specific data with freely-available land cover and urban planning data to derive a predictive model of landscape scale environmental risk. We used six different model algorithms to analyze the dataset, comparing these in terms of their predictive abilities and the variables identified as important. Based on comparable predictive abilities (mean R2 from 30 to 35% and mean root mean square error from 65 to 68%), we averaged all six model outputs to predict relative levels of trace element deposition in soils—given the road surface, traffic volume, sample distance from the road, land cover category, and impervious surface percentage. Mapped predictions of environmental risk from toxic trace element pollution can show land managers and transportation planners where to prioritize road renewal or maintenance by each road segment's relative environmental and human health risk.

A geography of ecotourism in Cambodia: regions, patterns, and potentials

ABSTRACTTourism visits to Cambodia have expanded rapidly, with over 5 million visitors arriving in the country by 2016. Angkor Wat remains the primary tourist attraction, but increasing numbers of tourists also visit ecotourism destinations. This study draws on Weaver’s [(2002). Asian ecotourism: Patterns and themes. Tourism Geographies, 4(2), 153–172] spatial and structural analysis of ecotourism in Asia as an organizing framework to develop a geography of ecotourism in Cambodia. In a comprehensive review of academic publications, reports, online ecotourism clearinghouses, and project websites, 30 Cambodian ecotourism sites were identified based on 5 ecotourism criteria. A content analysis of ecotourism project websites was then conducted to group together ecotourism sites with common characteristics in particular geographic regions. Each site was plotted on a Google Terrain map, and local and regional terrain, land cover, transport, and other physical features were analyzed, supplemented by detailed geographical data from Google Panorama, Earth, and Satellite map functions. Three relatively distinct ecotourism regions were identified: the (a) mangrove and rainforest region, (b) highland trekking region, and (c) wetlands and waterways region. For each region, tourism gateways and ecotourism development challenges and potentials are elaborated. The paper concludes with a discussion of study findings and their implications for ecotourism development in Cambodia.

Towards automated 10–30 m resolution land cover mapping in insular South-East Asia

In this study, we create and critically analyse an automated decision tree classification approach for regional level land cover mapping in insular South-East Asian conditions, using a combination of 10–30 m resolution optical and radar data. The resulting map contains 11 land cover classes and reveals a great deal of contextual information due to high spatial resolution. A limited accuracy assessment indicates 59–97% class wise accuracies. The unprecedented spatial detail of closed canopy oil palm mapping (with user’s accuracy of 90%) is seen as the most promising feature of the mapping approach. The incapability of separating primary forests from other tree cover, and the large variety of different landscapes (e.g. home gardens and tea plantations) classified as shrubland, are considered the main areas for future improvement. Overall, the study demonstrates the great potential of multi-source 10–30 m resolution high data volume land cover mapping approaches in insular South-East Asian conditions.

Evidence for regional nitrogen stress on chlorophyll a in lakes across large landscape and climate gradients

AbstractNitrogen (N) and phosphorus (P) commonly stimulate phytoplankton production in lakes, but recent observations from lakes from an agricultural region suggest that nitrate may have a subsidy‐stress effect on chlorophyll a (Chl a). It is unclear, however, how generalizable this effect might be. Here, we analyzed a large water quality dataset of 2385 lakes spanning 60 regions across 17 states in the Northeastern and Midwestern U.S. to determine if N subsidy‐stress effects on phytoplankton are common and to identify regional landscape characteristics promoting N stress effects in lakes. We used a Bayesian hierarchical modeling framework to test our hypothesis that Chl a–total N (TN) threshold relationships would be common across the central agricultural region of the U.S. (“the Corn Belt”), where lake N and P concentrations are high. Data aggregated across all regions indicated that high TN concentrations had a negative effect on Chl a in lakes with concurrent high total P. This large‐scale pattern was driven by relationships within only a subset of regions, however. Eight regions were identified as having Chl a–TN threshold relationships, but only two of these regions located within the Corn Belt clearly demonstrated this subsidy‐stress relationship. N stress effects were not consistent across other intense agricultural regions, as we hypothesized. These findings suggest that interactions among regional land use and land cover, climate, and hydrogeology may be important in determining the synergistic conditions leading to N subsidy‐stress effects on lake phytoplankton.

PEATMAP: Refining estimates of global peatland distribution based on a meta-analysis

Peatlands play important ecological, economic and cultural roles in human well-being. Although considered sensitive to climate change and anthropogenic pressures, the spatial extent of peatlands is poorly constrained. We report the development of an improved global peatland map, PEATMAP, based on a meta-analysis of geospatial information collated from a variety of sources at global, regional and national levels. We estimate total global peatland area to be 4.23millionkm2, approximately 2.84% of the world land area. Our results suggest that previous global peatland inventories are likely to underestimate peat extent in the tropics, and to overestimate it in parts of mid- and high-latitudes of the Northern Hemisphere. Global wetland and soil datasets are poorly suited to estimating peatland distribution. For instance, tropical peatland extents are overestimated by Global Lakes and Wetlands Database – Level 3 (GLWD-3) due to the lack of ground-truthing data; and underestimated by the use of histosols to represent peatlands in the Harmonized World Soil Database (HWSD) v1.2, as large areas of swamp forest peat in the humid tropics are omitted. PEATMAP and its underlying data are freely available as a potentially useful tool for scientists and policy makers with interests in peatlands or wetlands. PEATMAP's data format and file structure are intended to allow it to be readily updated when previously undocumented peatlands are found and mapped, and when regional or national land cover maps are updated and refined.

Biophysical and socioeconomic determinants of tea expansion: Apportioning their relative importance for sustainable land use policy

Tea expansion, a typical process of regional land use and cover change (LUCC), has raised great concerns on regional sustainability. In this regard, exploring the determinants of tea expansion should provide critical implications for land use policy. It has been widely recognized that LUCC interacts nonlinearly with a set of determinants and their feedbacks should be rather complex. Policy makers are now facing the challenge to identify, apportion, and compare the determinants of regional tea expansion for designing more targeted political intervenes. Our paper utilizes a robust tool, the random forest (RF) regression in particular, to explore the determinants of tea expansion across two periods (1985–2007 and 2007–2016) in Anji County, a typical region of tea production in subtropical China. More specifically, tea is extracted from Landsat imageries and total tea cultivated area acts as the dependent variable. Exploratory variables include 38 potential determinants and these determinants are divided into two categories (biophysical and socioeconomic) at two levels (pixel and village). We obtain some similar findings, though the relative importance of determinants varies with the two periods. In general, biophysical determinants (e.g., topography, soil type, land use in the neighborhood) present greater relative importance than the socioeconomic determinants in both periods. In period 1985–2007, biophysical determinants at pixel level are more essential in governing tea expansion. In period 2007–2016, the relative importance of pixel level biophysical determinants is comparable with that of the village level determinants. Comparisons of the two periods indicate that relative importance of soil type and socioeconomic proximity becomes greater in period 2007–2016, while that of the total employees and non-agricultural population proportion becomes lower. Partial dependency plots are further drawn to visualize the marginal effect of each determinant. We finally propose three options for land use policy towards sustainability. Our study demonstrates that the RF regression is efficient for policy makers to understand the determinants of tea expansion with a nonlinear and complex nature.

Multi-scale 46-year remote sensing change detection of diamond mining and land cover in a conflict and post-conflict setting

The town of Tortiya was created in the rural northern region of Côte d′Ivoire in the late 1940s to house workers for a new diamond mine. Nearly three decades later, the closure of the industrial-scale diamond mine in 1975 did not diminish the importance of diamond profits to the region's economy, and resulted in the growth of artisanal and small-scale diamond mining (ASM) within the abandoned industrial-scale mining concession. In the early 2000s, the violent conflict that arose in Côte d′Ivoire highlighted the importance of ASM land use to the local economy, but also brought about international concerns that diamond profits were being used to fund the rebellion. In recent years, cashew plantations have expanded exponentially in the region, diversifying economic activity, but also creating the potential for conflict between diamond mining and agricultural land uses. As the government looks to address the future of Tortiya and this potential conflict, a detailed spatio-temporal understanding of the changes in these two land uses over time may assist in informing policymaking.Remotely sensed imagery presents an objective and detailed spatial record of land use/land cover (LULC), and change detection methods can provide quantitative insight regarding regional land cover trends. However, the vastly different scales of ASM and cashew orchards present a unique challenge to comprehensive understanding of land use change in the region. In this study, moderate-scale categories of LULC, including cashew orchards, uncultivated forest, urban space, mining/ bare, and mixed vegetation, were produced through supervised classification of Landsat multispectral imagery from 1984, 1991, 2000, 2007, and 2014. The fine-scale ASM land use was identified through manual interpretation of annually acquired high resolution satellite imagery. Corona imagery was also integrated into the study to extend the temporal duration of the remote sensing record back to the period of industrial-scale mining. These different-scale analyses were then integrated to create a record of 46 years of mining activity and land cover change in Tortiya. While similar in spatial extent, the mining/ bare class in the integrated analysis exhibits a substantially different spatial distribution than in the original classifications. This additional information regarding the locations of ASM activity in the Tortiya area is important from a policy and planning perspective. The results of this study also suggest that LULC classifications of Landsat imagery do not consistently capture areas of ASM in the Côte d′Ivoire landscape.

Measuring land-use and land-cover change using the U.S. department of agriculture’s cropland data layer: Cautions and recommendations

Monitoring agricultural land is important for understanding and managing food production, environmental conservation efforts, and climate change. The United States Department of Agriculture's Cropland Data Layer (CDL), an annual satellite imagery-derived land cover map, has been increasingly used for this application since complete coverage of the conterminous United States became available in 2008. However, the CDL is designed and produced with the intent of mapping annual land cover rather than tracking changes over time, and as a result certain precautions are needed in multi-year change analyses to minimize error and misapplication. We highlight scenarios that require special considerations, suggest solutions to key challenges, and propose a set of recommended good practices and general guidelines for CDL-based land change estimation. We also characterize a problematic issue of crop area underestimation bias within the CDL that needs to be accounted for and corrected when calculating changes to crop and cropland areas. When used appropriately and in conjunction with related information, the CDL is a valuable and effective tool for detecting diverse trends in agriculture. By explicitly discussing the methods and techniques for post-classification measurement of land-cover and land-use change using the CDL, we aim to further stimulate the discourse and continued development of suitable methodologies. Recommendations generated here are intended specifically for the CDL but may be broadly applicable to additional remotely-sensed land cover datasets including the National Land Cover Database (NLCD), Moderate Resolution Imaging Spectroradiometer (MODIS)-based land cover products, and other regional, national, and global land cover classification maps.

Production of Global Land Cover Data – GLCNMO2013

Global land cover products have been created for global environmental studies by several institutions and organizations. The Global Mapping Project coordinated by the International Steering Committee for Global Mapping (ISCGM) has been periodically producing global land cover datasets asone of the eight basic global datasets. It has produced a new fifteen-second (approximately 500 m resolution at the equator) global land cover dataset – GLCNMO2013 (or GLCNMO version 3). This paper describes the method of producing GLCNMO2013. GLCNMO2013 has 20 land cover classes, and they were mapped by improved methods from GLCNMO version 2. In GLCNMO2013, five classes,which are urban, mangrove, wetland, snow/ice, and waterwere independently classified. The remaining 15 classes were divided into 4 groups and mapped individually by supervised classification. 2006 polygons of training data collected for GLCNMO2008 were used for supervised classification. In addition, about 3000 polygons of new training data were collected globally using Google Earth, MODIS Normalized Difference Vegetation Index (NDVI) seasonal change patterns, existing regional land cover maps, and existing four global land cover products. The primary data of this product were Moderate Resolution Imaging Spectroradiometer (MODIS) data of 2013. GLCNMO2013 was validated at 1006 sampled points. The overall accuracy of GLCNMO2013 was 74.8%, and the overall accuracy for eight aggregated classes was 90.2%. The accuracy of the GLCNMO2013 was not improved compared with the GLCNMO2008 at heterogeneous land covers. It is necessary to prepare the training data for mosaic classes and heterogeneous land covers for improving the accuracy.

The Standardization and Harmonization of Land Cover Classification Systems towards Harmonized Datasets: A Review

A number of national, regional and global land cover classification systems have been developed to meet specific user requirements for land cover mapping exercises, independent of scale, nomenclature and quality. However, this variety of land-cover classification systems limits the compatibility and comparability of land cover data. Furthermore, the current lack of interoperability between different land cover datasets, often stemming from incompatible land cover classification systems, makes analysis of multi-source, heterogeneous land cover data for various applications a very difficult task. This paper provides a critical review of the harmonization of land cover classification systems, which facilitates the generation, use and analysis of land cover maps consistently. Harmonization of existing land cover classification systems is essential to improve their cross-comparison and validation for understanding landscape patterns and changes. The paper reviews major land cover classification standards according to different scales, summarizes studies on harmonizing land cover mapping, and discusses some research problems that need to be solved and some future research directions.

Carbon lost and carbon gained: a study of vegetation and carbon trade-offs among diverse land uses in Phoenix, Arizona.

Human modification and management of urban landscapes drastically alters vegetation and soils, thereby altering carbon (C) storage and rates of net primary productivity (NPP). Complex social and ecological processes drive vegetation cover in cities, leading to heterogeneity in C dynamics depending on regional climate, land use, and land cover. Recent work has demonstrated homogenization in ecological processes within human-dominated landscapes (the urban convergence hypothesis) in soils and biotic communities. However, a lack of information on vegetation in arid land cities has hindered an understanding of potential C storage and NPP convergence across a diversity of ecosystem types. We estimated C storage and NPP of trees and shrubs for six different land-use types in the arid metropolis of Phoenix, Arizona, USA, and compared those results to native desert ecosystems, as well as other urban and natural systems around the world. Results from Phoenix do not support the convergence hypothesis. In particular, C storage in urban trees and shrubs was 42% of that found in desert vegetation, while NPP was only 20% of the total NPP estimated for comparable natural ecosystems. Furthermore, the overall estimates of C storage and NPP associated with urban trees in the CAP ecosystem were much lower (8-63%) than the other cities included in this analysis. We also found that C storage (175.25-388.94g/m2 ) and NPP (8.07-15.99g·m-2 ·yr-1 ) were dominated by trees in the urban residential land uses, while in the desert, shrubs were the primary source for pools (183.65g/m2 ) and fluxes (6.51g·m-2 ·yr-1 ). These results indicate a trade-off between shrubs and trees in arid ecosystems, with shrubs playing a major role in overall C storage and NPP in deserts and trees serving as the dominant C pool in cities. Our research supports current literature that calls for the development of spatially explicit and standardized methods for analyzing C dynamics associated with vegetation in urbanizing areas.

Spatial clusters and temporal trends of seasonal surface soil moisture across China in responses to regional climate and land cover changes

AbstractInvestigating the spatiotemporal changes of regional surface soil moisture in responses to climate and land cover changes is vital for understanding the underlying mechanisms of hydrological processes. While previous studies mainly attributed the causes of soil moisture changes to climatic factors, few took land cover into consideration. We analyzed the seasonal‐differentiated effects of climate and land cover changes on surface soil moisture of China's 80 drainage basins using the Essential Climate Variable Soil Moisture product spanning the 1979–2010 period. The low–low spatial clusters of annual (Jan–Dec), warm‐season (Apr–Sept) and cool‐season (Oct–Dec and Jan–Mar) surface soil moisture have spread from Northwest China to northeast and even Central China during the past three decades. In cool seasons, significant decreasing trends of surface soil moisture in most drainage basins of Northwest, Northeast, Central, and South China were detected. But in warm seasons, the surface soil moisture in Central and South China showed significant increasing trends. Precipitation, potential evapotranspiration and the ratio of the two are three driving forces for warm‐season surface soil moisture changes in Northeast and Northwest China. These climatic factors are main contributors to the declining trends of cool‐season surface soil moisture in Central and South China. Land cover changes showed to be the major factor driving the significant decreasing trends of cool‐season surface soil moisture in Northwest, Central, and South China. Due to vegetation's self‐shading effect, the large‐scale reforestation and vegetation growth are believed as main causes of the increasing trends of warm‐season surface soil moisture in Central and South China.

On the performance of remote sensing time series reconstruction methods – A spatial comparison

Abstract The satellite observed Normalized Difference Vegetation Index (NDVI) time series, which describe the temporal and spatial variability of global terrestrial vegetation, are inevitably contaminated by clouds, aerosol, snow and ice cover. In general all these conditions yield negative deviations in the time series of NDVI. Many time series reconstruction models have been developed to eliminate effect of the negative deviations and most of them perform differently in different applications and regions. The Harmonic Analysis (HA), Double logistic (DL), Asymmetric Gaussian (AG), Whittaker smoother (WS) and Savitzky–Golay filter (SG) are five of the most widely used time series reconstruction models owing to their simplicity of implementation or the capability to extract phenological metrics from the time series. The performance of these models varies with the NDVI signal and the noise distribution, however, and until now there is no consensus on which method outperforms all others under all situations. Since the NDVI signal and the noise distribution are highly dependent on regional climate and land cover, the reconstruction performance is expected to be spatially variable. Thus this study compared the five reconstruction models at pixel scale to provide practical and biome – specific recommendations for future time series reconstruction applications. Specifically, the 14 years raw daily reflectance data and ancillary Quality Assessment (QA) information from the MODIS sensor were used to generate pixel reference series and noisy series. Then the five candidate models were applied to both reference series and noisy series and three reconstruction performance metrics i.e. Overall Reconstruction Error (ORE), Fitting Related Error (FRE), and Normalized Noise Related Error (NNRE), were calculated. Finally, the performance of the five candidate reconstruction models was evaluated by applying the three metrics. The preliminary results showed that when considering ORE only, the Asymmetric Gaussian model outperforms other models over most areas of high latitude boreal region, while the Savitzky–Golay model gives the best reconstruction performance in tropical and subtropical regions. The FRE and the NNRE helped to reveal the main error sources in the reconstruction in different regions. The comparison method developed and applied in this study led to suggest adaptive selection of the best reconstruction model for specific NDVI signals and noise distribution.

Land’s complex role in climate change

To mitigate climate change at local, regional, and global scales, we must begin to think beyond greenhouse gases.

Rapid Land Cover Map Updates Using Change Detection and Robust Random Forest Classifiers

The paper evaluated the Landsat Automated Land Cover Update Mapping (LALCUM) system designed to rapidly update a land cover map to a desired nominal year using a pre-existing reference land cover map. The system uses the Iteratively Reweighted Multivariate Alteration Detection (IRMAD) to identify areas of change and no change. The system then automatically generates large amounts of training samples (n &gt; 1 million) in the no-change areas as input to an optimized Random Forest classifier. Experiments were conducted in the KwaZulu-Natal Province of South Africa using a reference land cover map from 2008, a change mask between 2008 and 2011 and Landsat ETM+ data for 2011. The entire system took 9.5 h to process. We expected that the use of the change mask would improve classification accuracy by reducing the number of mislabeled training data caused by land cover change between 2008 and 2011. However, this was not the case due to exceptional robustness of Random Forest classifier to mislabeled training samples. The system achieved an overall accuracy of 65%–67% using 22 detailed classes and 72%–74% using 12 aggregated national classes. “Water”, “Plantations”, “Plantations—clearfelled”, “Orchards—trees”, “Sugarcane”, “Built-up/dense settlement”, “Cultivation—Irrigated” and “Forest (indigenous)” had user’s accuracies above 70%. Other detailed classes (e.g., “Low density settlements”, “Mines and Quarries”, and “Cultivation, subsistence, drylands”) which are required for operational, provincial-scale land use planning and are usually mapped using manual image interpretation, could not be mapped using Landsat spectral data alone. However, the system was able to map the 12 national classes, at a sufficiently high level of accuracy for national scale land cover monitoring. This update approach and the highly automated, scalable LALCUM system can improve the efficiency and update rate of regional land cover mapping.

Temporal replication of the national land cover database using active machine learning

Regional and national level land cover datasets, such as the National Land Cover Database (NLCD) in the United States, have become an important resource in physical and social science research. Updates to the NLCD have been conducted every 5 years since 2001; however, the procedure for producing a new release is labor-intensive and time-consuming, taking 3 or 4 years to complete. Furthermore, in most countries very few, if any, such releases exist, and thus there is high demand for efficient production of land cover data at different points in time. In this paper, an active machine learning framework for temporal updating (or backcasting) of land cover data is proposed and tested for three study sites covered by the NLCD. The approach employs a maximum entropy classifier to extract information from one Landsat image using the NLCD, and then replicate the classification on a Landsat image for the same geographic extent from a different point in time to create land cover data of similar quality. Results show that this framework can effectively replicate the land cover database in the temporal domain with similar levels of overall and within class agreement when compared against high resolution reference land cover datasets. These results demonstrate that the land cover information encapsulated in the NLCD can effectively be extracted using solely Landsat imagery for replication purposes. The algorithm is fully automated and scalable for applications at landscape and regional scales for multiple points in time.