Thematic accuracy of the NLCD 2001 land cover for the conterminous United States

The land-cover thematic accuracy of NLCD 2001 was assessed from a probability-sample of 15,000 pixels. Nationwide, NLCD 2001 overall Anderson Level II and Level I accuracies were 78.7% and 85.3%, respectively. By comparison, overall accuracies at Level II and Level I for the NLCD 1992 were 58% and 80%. Forest and cropland were two classes showing substantial improvements in accuracy in NLCD 2001 relative to NLCD 1992. NLCD 2001 forest and cropland user's accuracies were 87% and 82%, respectively, compared to 80% and 43% for NLCD 1992. Accuracy results are reported for 10 geographic regions of the United States, with regional overall accuracies ranging from 68% to 86% for Level II and from 79% to 91% at Level I. Geographic variation in class-specific accuracy was strongly associated with the phenomenon that regionally more abundant land-cover classes had higher accuracy. Accuracy estimates based on several definitions of agreement are reported to provide an indication of the potential impact of reference data error on accuracy. Drawing on our experience from two NLCD national accuracy assessments, we discuss the use of designs incorporating auxiliary data to more seamlessly quantify reference data quality as a means to further advance thematic map accuracy assessment.     

Assessment of forest cover in Russia by combining a wall-to-wall coarse-resolution land-cover map with a sample of 30 m resolution forest maps

The process of gathering land-cover information has evolved significantly over the last decade (2000–2010). In addition to this, current technical infrastructure allows for more rapid and efficient processing of large multi-temporal image databases at continental scale. But whereas the data availability and processing capabilities have increased, the production of dedicated land-cover products with adequate accuracy is still a prerequisite for most users. Indeed, spatially explicit land-cover information is important and does not exist for many regions. Our study focuses on the boreal Eurasia region for which limited land-cover information is available at regional level.

The main aim of this paper is to demonstrate that a coarse-resolution land-cover map of the Russian Federation, the ‘TerraNorte’ map at 230 m × 230 m resolution for the year 2010, can be used in combination with a sample of reference forest maps at 30 m resolution to correctly assess forest cover in the Russian federation.

First, an accuracy assessment of the TerraNorte map is carried out through the use of reference forest maps derived from finer-resolution satellite imagery (Landsat Thematic Mapper (TM) sensor). A sample of 32 sites was selected for the detailed identification of forest cover from Landsat TM imagery. A methodological approach is developed to process and analyse the Landsat imagery based on unsupervised classification and cluster-based visual labelling. The resulting forest maps over the 32 sites are then used to evaluate the accuracy of the forest classes of the TerraNorte land-cover map. A regression analysis shows that the TerraNorte map produces satisfactory results for areas south of 65° N, whereas several forest classes in more northern areas have lower accuracy. This might be explained by the strong reflectance of background (i.e. non-tree) cover.

A forest area estimate is then derived by calibration of the TerraNorte Russian map using a sample of Landsat-derived reference maps (using a regression estimator approach). This estimate compares very well with the FAO FRA exercise for 2010 (1% difference for total forested area). We conclude that the TerraNorte map combined with finer-resolution reference maps can be used as a reliable spatial information layer for forest resources assessment over the Russian Federation at national scale.     

Requirements for labelling forest polygons in an object-based image analysis classification

The ability to spatially quantify changes in the landscape and create land-cover maps is one of the most powerful uses of remote sensing. Recent advances in object-based image analysis (OBIA) have also improved classification techniques for developing land-cover maps. However, when using an OBIA technique, collecting ground data to label reference units may not be straightforward, since these segments generally contain a variable number of pixels as well as a variety of pixel values, which may reflect variation in land-cover composition. Accurate classification of reference units can be particularly difficult in forested land-cover types, since these classes can be quite variable on the ground. This study evaluates how many prism sample locations are needed to attain an acceptable level of accuracy within forested reference units in southeastern New Hampshire (NH). Typical forest inventory guidelines suggest at least 10 prism samples per stand, depending on the stand area and stand type. However, because OBIA segments group pixels based on the variance of the pixels, fewer prism samples may be necessary in a segment to properly estimate the stand composition. A bootstrapping statistical technique was used to find the necessary number of prism samples to limit the variance associated with estimating the species composition of a segment. Allowing for the lowest acceptable variance, a maximum of only six prism samples was necessary to label forested reference units. All polygons needed at least two prism samples for classification.     

Studies of land-cover, land-use, and biophysical properties of vegetation in the Large Scale Biosphere Atmosphere experiment in Amazônia

We summarize early research on land-cover, land-use, and biophysical properties of vegetation from the Large Scale Biosphere Atmosphere (LBA) experiment in Amazônia. LBA is an international research program developed to evaluate regional function and to determine how land-use and climate modify biological, chemical and physical processes there. Remote sensing has played a fundamental role in LBA in research planning, land-cover mapping and in long-term monitoring of changes in land-cover and land-use at multiple scales. This special issue includes 12 papers that cover a range in spatial scales from regional mapping to local scales that cover only a portion of a Landsat scene. Several themes dominate, including land-cover mapping with an emphasis on wetlands and second-growth forest, evaluation of pasture sustainability and forest degradation and the impact of land-cover change on stream chemistry. New techniques introduced include automated Monte Carlo unmixing (AutoMCU) and several new approaches for mapping land-cover. A diversity of sensors are utilized, including ETM+, IKONOS, SPOT-4, Airborne P-band synthetic aperture radar (SAR), and L-band SAR. Census data are fused with an existing land-cover map to generate spatially explicit estimates of land-use from historical data. Several papers include important, new field measures of species composition, forest structure and biomass in mature forest and secondary succession.     

Thematic accuracy of the 1992 National Land-Cover Data for the eastern United States: Statistical methodology and regional results

The accuracy of the 1992 National Land-Cover Data (NLCD) map is assessed via a probability sampling design incorporating three levels of stratification and two stages of selection. Agreement between the map and reference land-cover labels is defined as a match between the primary or alternate reference label determined for a sample pixel and a mode class of the mapped 3×3 block of pixels centered on the sample pixel. Results are reported for each of the four regions comprising the eastern United States for both Anderson Level I and II classifications. Overall accuracies for Levels I and II are 80% and 46% for New England, 82% and 62% for New York/New Jersey (NY/NJ), 70% and 43% for the Mid-Atlantic, and 83% and 66% for the Southeast.     

Mapping moderate-scale land-cover over very large geographic areas within a collaborative framework: A case study of the Southwest Regional Gap Analysis Project (SWReGAP)

Land-cover mapping efforts within the USGS Gap Analysis Program have traditionally been state-centered; each state having the responsibility of implementing a project design for the geographic area within their state boundaries. The Southwest Regional Gap Analysis Project (SWReGAP) was the first formal GAP project designed at a regional, multi-state scale. The project area comprises the southwestern states of Arizona, Colorado, Nevada, New Mexico, and Utah. The land-cover map/dataset was generated using regionally consistent geospatial data (Landsat ETM+ imagery (1999-2001) and DEM derivatives), similar field data collection protocols, a standardized land-cover legend, and a common modeling approach (decision tree classifier). Partitioning of mapping responsibilities amongst the five collaborating states was organized around ecoregion-based “mapping zones”. Over the course of 2¹/₂ field seasons approximately 93,000 reference samples were collected directly, or obtained from other contemporary projects, for the land-cover modeling effort. The final map was made public in 2004 and contains 125 land-cover classes. An internal validation of 85 of the classes, representing 91% of the land area was performed. Agreement between withheld samples and the validated dataset was 61% (KHAT = .60, n = 17,030). This paper presents an overview of the methodologies used to create the regional land-cover dataset and highlights issues associated with large-area mapping within a coordinated, multi-institutional management framework.     

Thematic accuracy of the National Land Cover Database (NLCD) 2001 land cover for Alaska

The National Land Cover Database (NLCD) 2001 Alaska land cover classification is the first 30-m resolution land cover product available covering the entire state of Alaska. The accuracy assessment of the NLCD 2001 Alaska land cover classification employed a geographically stratified three-stage sampling design to select the reference sample of pixels. Reference land cover class labels were determined via fixed wing aircraft, as the high resolution imagery used for determining the reference land cover classification in the conterminous U.S. was not available for most of Alaska. Overall thematic accuracy for the Alaska NLCD was 76.2% (s.e. 2.8%) at Level II (12 classes evaluated) and 83.9% (s.e. 2.1%) at Level I (6 classes evaluated) when agreement was defined as a match between the map class and either the primary or alternate reference class label. When agreement was defined as a match between the map class and primary reference label only, overall accuracy was 59.4% at Level II and 69.3% at Level I. The majority of classification errors occurred at Level I of the classification hierarchy (i.e., misclassifications were generally to a different Level I class, not to a Level II class within the same Level I class). Classification accuracy was higher for more abundant land cover classes and for pixels located in the interior of homogeneous land cover patches.     

Combining per-pixel and object-based classifications for mapping land cover over large areas

A plethora of national and regional applications need land-cover information covering large areas. Manual classification based on visual interpretation and digital per-pixel classification are the two most commonly applied methods for land-cover mapping over large areas using remote-sensing images, but both present several drawbacks. This paper tests a method with moderate spatial resolution images for deriving a product with a predefined minimum mapping unit (MMU) unconstrained by spatial resolution. The approach consists of a traditional supervised per-pixel classification followed by a post-classification processing that includes image segmentation and semantic map generalization. The approach was tested with AWiFS data collected over a region in Portugal to map 15 land-cover classes with 10 ha MMU. The map presents a thematic accuracy of 72.6 ± 3.7% at the 95% confidence level, which is approximately 10% higher than the per-pixel classification accuracy. The results show that segmentation of moderate-spatial resolution images and semantic map generalization can be used in an operational context to automatically produce land-cover maps with a predefined MMU over large areas.     

Combining medium and coarse spatial resolution satellite data to improve the estimation of sub-pixel NDVI time series

The potential of multitemporal coarse spatial resolution remotely sensed images for vegetation monitoring is reduced in fragmented landscapes, where most of the pixels are composed of a mixture of different surfaces. Several approaches have been proposed for the estimation of reflectance or NDVI values of the different land-cover classes included in a low resolution mixed pixel. In this paper, we propose a novel approach for the estimation of sub-pixel NDVI values from multitemporal coarse resolution satellite data. Sub-pixel NDVIs for the different land-cover classes are calculated by solving a weighted linear system of equations for each pixel of a coarse resolution image, exploiting information about within-pixel fractional cover derived from a high resolution land-use map. The weights assigned to the different pixels of the image for the estimation of sub-pixel NDVIs of a target pixel i are calculated taking into account both the spatial distance between each pixel and the target and their spectral dissimilarity estimated on medium-resolution remote-sensing images acquired in different periods of the year. The algorithm was applied to daily and 16-day composite MODIS NDVI images, using Landsat-5 TM images for calculation of weights and accuracy evaluation.Results showed that application of the algorithm provided good estimates of sub-pixel NDVIs even for poorly represented land-cover classes (i.e., with a low total cover in the test area). No significant accuracy differences were found between results obtained on daily and composite MODIS images. The main advantage of the proposed technique with respect to others is that the inclusion of the spectral term in weight calculation allows an accurate estimate of sub-pixel NDVI time series even for land-cover classes characterized by large and rapid spatial variations in their spectral properties.     

A global land-cover validation data set,part I: fundamental design principles

A number of land-cover products, both global and regional, have been produced and more are forthcoming. Assessing their accuracy would be greatly facilitated by a global validation database of reference sites that allows for comparative assessments of uncertainty for multiple land-cover data sets. We propose a stratified random sampling design for collecting reference data. Because the global validation database is intended to be applicable to a variety of land-cover products, the stratification should be implemented independently of any specific map to facilitate general utility of the data. The stratification implemented is based on the Köppen climate/vegetation classification and population density. A map of the Köppen classification was manually edited and intersected by two layers of population density and a land water mask. A total of 21 strata were defined and an initial global sample of 500 reference sites was selected, with each site being a 5?×?5 km block. The decision of how to allocate the sample size to strata was informed by examining the distribution of the sample area of land cover for two global products resulting from different sample size allocations to the 21 strata. The initial global sample of 500 sites selected from the Köppen-based stratification indicates that these strata can be used effectively to distribute sample sites among rarer land-cover classes of the two global maps examined, although the strata were not constructed using these maps. This is the first article of two, with the second paper presenting details of how the sampling design can be readily augmented to increase the sample size in targeted strata for the purpose of increasing the sample sizes for rare classes of a particular map being evaluated.     

Object-based image analysis for operational fine-scale regional mapping of land cover within river corridors from multispectral imagery and thematic data

Accurate mapping of land-cover diversity within riparian areas at a regional scale is a major challenge for better understanding the influence of riparian landscapes and related natural and anthropogenic pressures on river ecological status. As the structure (composition and spatial organization) of riparian area land cover (RALC) is generally not accessible using moderate-scale satellite imagery, finer spatial resolution imagery and specific mapping techniques are needed. For this purpose, we developed a classification procedure based on a specific multiscale object-based image analysis (OBIA) scheme dedicated to producing fine-scale and reliable RALC maps in different geographical contexts (relief, climate and geology). This OBIA scheme combines information from very high spatial resolution multispectral imagery (satellite or airborne) and available spatial thematic data using fuzzy expert knowledge classification rules. It was tested over the Hérault River watershed (southern France), which presents contrasting landscapes and a total stream length of 1150 km, using the combination of SPOT (Système Probatoire d'Observation de la Terre) 5 XS imagery (10 m pixels), aerial photography (0.5 m pixels) and several national spatial thematic data. A RALC map was produced (22 classes) with an overall accuracy of 89% and a kappa index of 83%, according to a targeted land-cover pressures typology (six categories of pressures). The results of this experimentation demonstrate that the application of OBIA to multisource spatial data provides an efficient approach for the mapping and monitoring of RALC that can be implemented operationally at a regional or national scale. We further analysed the influence of map resolution on the quantification of riparian spatial indicators to highlight the importance of such data for studying the influence of landscapes on river ecological status at the riparian scale.     

Estimating area and map accuracy for stratified random sampling when the strata are different from the map classes

The results of an accuracy assessment are typically organized using an error matrix that displays the proportion of area correctly mapped for each class and the proportion of area misclassified. Stratified random sampling is commonly implemented to obtain the reference data used to estimate the error matrix. When the strata correspond exactly to the map classes, the formulas for estimating accuracy and area are well known. Nevertheless, applications arise in which the strata are different from the map classes, as for example when the stratification is based on the map class labels of one map but the sample is subsequently used to assess the accuracy of other maps. In this paper, the estimators required when the stratum label and map label do not match for all pixels are presented for the proportion of area of each class based on the reference classification and for overall, user’s, and producer’s accuracies. Standard error formulas are also presented. A numerical example is provided to illustrate the computations.     

Mapping forest alliances and associations using fuzzy systems and nearest neighbor classifiers

The study and management of biological communities depends on systems of classification and mapping for the organization and communication of resource information. Recent advances in remote sensing technology may enable the mapping of forest plant associations using image classification techniques. But few areas outside Europe have alliances and associations described in detail sufficient to support remote sensing-based modeling. Northwestern Montana has one of the few completed plant association classifications in the United States compliant with the recently established National Vegetation Classification system. This project examined the feasibility of mapping forest plant associations using Landsat Enhanced Thematic Mapper Plus data and advanced remote sensing technology and image classification techniques.Suitable reference data were selected from an extensive regional database of plot records. Fifteen percent of the plot samples were reserved for validation of map products, the remainder of plots designated as training data for map modeling. Key differentiae for image classification were identified from a suite of spectral and biophysical variables. Fuzzy rules were formulated for partitioning physiognomic classes in the upper levels of our image classification hierarchy. Nearest neighbor classifiers were developed for classification of lower levels (alliances and associations), where spectral and biophysical contrasts are less distinct.Maps were produced to reflect nine forest alliances and 24 associations across the study area. Error matrices were constructed for each map based on stratified random selections of map validation samples. Accuracy for the alliance map was estimated at 60%. Association classifiers provide between 54 and 86% accuracy within their respective alliances. Alternative techniques are proposed for aggregating classes and enhancing decision tree classifiers to model alliances and associations for interior forest types.     

Coastal wetland classification with multiseasonal high-spatial resolution satellite imagery

ABSTRACT

Accurate mapping of wetland distribution is required for wetland conservation, management, and restoration, but remains a challenge due to the complexity of wetland landscapes. This research employed four seasons of multispectral images from Gaofen-1 satellite to map wetland land-cover distribution in Hangzhou bay coastal wetland (245 km²) in China. Maximum likelihood classifier (MLC), random forest (RF), and the expert-based approach were examined based on spectral, spatial, and phenological features. The results showed that land-cover classification accuracies of 83.9% using RF and 90.3% using the expert-based approach were obtained, and they had higher accuracy than MLC, which had an overall accuracy of only 63.3%. The high classification accuracy for nine land-cover classes using the expert-based approach indicated the important role of expert knowledge from the phenological features in improving wetland classification accuracy. As high spatial resolution satellite images become more easily obtainable, effective use of temporal information of different sensor data will be valuable for detailed land-cover classification with higher accuracy. The approach to establish expert rules from multitemporal images provides a new way to improve land-cover classification in different terrestrial ecosystems.     

Classification of boreal forest cover types using SAR images

Mapping forest cover types in the boreal ecosystem is important for understanding the processes governing the interaction of the surface with the atmosphere. In this paper, we report the results of the land-cover classification of the SAR (synthetic aperture radar) data acquired during the Boreal Ecosystem Atmospheric Study's intensive field campaigns over the southern study area near Prince Albert, Canada. A Bayesian maximum a posteriori classifier was applied on the national Aeronautics and Space Administration/Jet Propulsion Laboratory airborne SAR images covering the region during the peak of the growing season in July 1994. The approach is supervised in the sense that a combination of field data and existing land-cover maps are used to develop training areas for the desired classes. The images acquired were first radiometrically and absolutely calibrated, the incidence angle effect in airborne images was corrected to an acceptable accuracy, and the images were used in a mosaic form and geocoded and georeferenced with an existing land-cover map for validation purposes. The results show that SAR images can be classified into dominant forest types such as jack pine, black spruce, trembling aspen, clearing, open water, and three categories of mixed strands with better than 90% accuracy. The unispecies stands such as jack pine and black spruce are separated with 98% accuracy, but the accuracy of mixed coniferous and deciduous stands suffers from confusing factors such as varying species composition, surface moisture, and understory effects. To satisfy the requirements of process models, the number of cover types was reduced from eight to five general classes of conifer wet, conifer dry, mixed deciduous, disturbed, and open water. Reduction of classes improved the overall accuracy of the classification over the entire region from 77% to 92%.     

Incorporating the uncertainty of linguistic-scale reference data to assess accuracy of land-cover maps using fuzzy intervals

The reference classifications that serve as the fundamental basis of accuracy assessment of land-cover maps are subject to uncertainty. A fuzzy interval approach is proposed in which linguistic-scale labels assigned to each land-cover class at each sample observation are converted to fuzzy intervals. These fuzzy intervals are then used to produce a fuzzy confusion matrix from which fuzzy thematic accuracy measures analogous to overall, user's, and producer's accuracy are produced. An advantage of this methodology is that it employs a practical and relatively simple reference labelling protocol (the linguistic scale) that accounts for reference database uncertainty and information on the percentage of the pixel covered by a land-cover class incorporating this uncertainty in fuzzy accuracy measures, providing an analysis that is readily interpretable because of similarity to the familiar confusion matrix approach. The fuzzy accuracy measures can be defuzzified to provide simplified accuracy measures analogous to the thematic accuracy measures derived from traditional (i.e. crisp classification) confusion matrices. The proposed methodology is illustrated using a case study in which the accuracy assessment of a land-cover map for Continental Portugal derived from Medium Resolution Imaging Spectrometer (MERIS) images is made.     

Major trends in the land surface phenology (LSP) of Africa,controlling for land-cover change

ABSTRACT

Monitoring land surface phenology (LSP) trends is important in understanding how both climatic and non-climatic factors influence vegetation growth and dynamics. Controlling for land-cover changes in these analyses has been undertaken only rarely, especially in poorly studied regions like Africa. Using regression models and controlling for land-cover changes, this study estimated LSP trends for Africa from the enhanced vegetation index (EVI) derived from 500 m surface reflectance Moderate-Resolution Imaging Spectroradiometer (MOD09A1), for the period from 2001 to 2015. Overall end of season showed slightly more pixels with significant trends (12.9% of pixels) than start of season (11.56% of pixels) and length of season (LOS) (5.72% of pixels), leading generally to more ‘longer season’ LOS trends. Importantly, LSP trends that were not affected by land-cover changes were distinguished from those that were influenced by land-cover changes such as to map LSP changes that have occurred within stable land-cover classes and which might, therefore, be reasonably associated with climate changes through time. As expected, greater slope magnitudes were observed more frequently for pixels with land-cover changes compared to those without, indicating the importance of controlling for land cover. Consequently, we suggest that future analyses of LSP trends should control for land-cover changes such as to isolate LSP trends that are solely climate-driven and/or those influenced by other anthropogenic activities or a combination of both.     

Estimating standard errors of accuracy assessment statistics under cluster sampling

Cluster sampling is a viable sampling design for collecting reference data for the purpose of conducting an accuracy assessment of land-cover classifications obtained from remotely sensed data. The formulas for estimating various accuracy parameters such as the overall proportion of pixels correctly classified, the kappa coefficient of agreement, and user's and producer's accuracy are the same under cluster sampling and simple random sampling, but the formulas for estimating standard errors differ between the two designs. If standard error formulas appropriate for cluster sampling are not employed in an accuracy assessment based on this design, the reported variability of map accuracy statistics is likely to be grossly misleading. The proper standard error formulas for common map accuracy statistics are derived for one-stage cluster sampling. The validity of these standard error formulas is verified by a small simulation study, and the standard errors computed according to the usual simple random sampling formulas are shown to underestimate the true cluster sampling standard errors by 20–70% if the intracluster correlation is moderate.     

Pixels, blocks of pixels, and polygons: Choosing a spatial unit for thematic accuracy assessment

Pixels, blocks of pixels, and polygons are all potentially viable spatial assessment units for conducting an accuracy assessment. We develop a population-based statistical framework to examine how the spatial unit chosen affects the outcome of an accuracy assessment. The population is conceptualized as a difference map created by overlaying a complete coverage reference classification and the target map being evaluated. The per-class areas of agreement and disagreement derived from this population are summarized by a population error matrix and accuracy parameters (e.g., overall, user's and producer's accuracies). The population and values of the accuracy parameters are strongly affected by the protocols implemented for the response design which include the choice of spatial unit, how within-unit homogeneity is addressed when assigning class labels, and the definition of agreement between the reference and map classification. Several complete coverage populations are used to illustrate how accuracy results are affected by the spatial unit chosen for the assessment and also to evaluate how spatial misregistration of the map and reference locations impacts accuracy results for different spatial units. The sampling design implemented for accuracy assessment does not change the population or values of the accuracy parameters, but the choice of spatial unit will influence decisions regarding use of strata and clusters in the design. A universally best spatial assessment unit does not exist, so it is critical to recognize how the population, values of the accuracy parameters, and sampling design are impacted by the choice of spatial unit.