Accuracy assessment using sub-pixel fractional error matrices of global land cover products derived from satellite data

Information on land cover distribution at regional and global scales has become fundamental for studying global changes affecting ecological and climatic systems. The remote sensing community has responded to this increased interest by improving data quality and methodologies for extracting land cover information. However, in addition to the advantages provided by satellite products, certain limitations exist that need to be objectively quantified and clearly communicated to users so that they can make informed decisions on whether and how land cover products should be used. Accuracy assessment is the procedure used to quantify product quality. Some aspects of accuracy assessment for evaluating four global land cover maps over Canada are discussed in this paper. Attempts are made to quantify limiting factors resulting from the coarse spatial resolution of data used for generating land cover information at regional and global levels. Sub-pixel fractional error matrices are introduced as a more appropriate way for assessing the accuracy of mixed pixels. For classification with coarse spatial resolution data, limitations of the classification method produce a maximum achievable accuracy defined as the average percent fraction of dominant land cover of all pixels in the mapped area. Relationships among spatial resolution, landscape heterogeneity and thematic resolution were studied and reported. Other factors that can affect accuracy, such as misregistration and legend conversion, are also discussed.     

The influence of urban structures on impervious surface maps from airborne hyperspectral data

Airborne hyperspectral data fulfills the high spectral and spatial resolution requirements of urban remote sensing applications. Its high spectral information content enables delineating impervious areas including the separation of built-up and non built-up surfaces, thus being of high relevance for many urban environmental applications. However, two phenomena related to surface structure negatively impact the accuracy of maps from such airborne data sets: (1) displaced buildings that lead to confusion between the class built-up and adjacent non built-up areas as a function of building height and view-angle; (2) urban street trees obscuring impervious surface underneath. Both effects have so far not been investigated from airborne hyperspectral data and potential sources of inaccuracy are usually not differentiated in analysis utilizing such data. Thus, the positive influence of hyperspectral information might have been undervalued in many cases. We set up an analysis scheme that allows for separately quantifying sources of error when producing land cover maps from urban areas. Given reliable cadastral information on building extent and street network, a detailed analysis for a relatively large Hyperspectral Mapper data set acquired over Berlin, Germany, was performed. Results show that both building displacement and impervious surface obscured by tree crowns are of great impact: at large view-angles, building displacement adds up to 16% error compared to nadir regions; more than 30% of the street area is classified as vegetation. Moreover, both effects show irregularities that prohibit empirical correction: misclassification due to building displacement also depends on view-direction, i.e. illumination properties and shadow, while the influence of trees differs significantly along streets and inside residential areas. Results from this work underline the necessity to consider all image processing steps when evaluating the accuracy and reliability of remote sensing products and they depict directions for future methodological development.     

Sources of error in accuracy assessment of thematic land-cover maps in the Brazilian Amazon

Valid measures of map accuracy are critical, yet can be inaccurate even when following well-established procedures. Accuracy assessment is particularly problematic when thematic classes lie along a land-cover continuum, and boundaries between classes are ambiguous. In this study, we examined error sources introduced during accuracy assessment of a regional land-cover map generated from Landsat Thematic Mapper (TM) data in Rondônia, southwestern Brazil. In this dynamic, highly fragmented landscape, the dominant land-cover classes represent a continuum from pasture to second growth to primary forest. We used high spatial resolution, geocoded videography as a reference, and focused on second-growth forest because of its potential contribution to the regional carbon balance. To quantify subjectivity in reference data labeling, we compared reference data produced by five trained interpreters. We also quantified the impact of other error sources, including geolocation errors between the map and reference data, land-cover changes between dates of data collection, heterogeneous reference samples, and edge pixels.Interpreters disagreed on classification of almost 30% of the samples; mixed reference samples and samples located in transitional classes accounted for a majority of disagreements. Agreement on second-growth forest labels between any two interpreters averaged below 50%, while agreement on primary forest was over 90%. Greater than 30% of disagreement between map and reference data was attributed to geolocation error, and 2.4% of disagreement was attributed to change in land cover between dates. After geocorrection, 24% of remaining disagreements corresponded to reference samples with mixed land cover, and 47% corresponded to edge pixels on the classified map. These findings suggest that: (1) labels of continuous land-cover types are more subjective and variable than commonly assumed, especially for transitional classes; however, using multiple interpreters to produce the reference data classification increases reference data accuracy; and (2) validation data sets that include only non-mixed, non-edge samples are likely to result in overly optimistic accuracy estimates, not representative of the map as a whole. These results suggest that different regional estimates of second-growth extent may be inaccurate and difficult to compare.     

Comparison of burned area estimates derived from SPOT-VEGETATION and Landsat ETM+ data in Africa: Influence of spatial pattern and vegetation type

An algorithm for burned area mapping in Africa based on classification trees was developed using SPOT-VEGETATION (VGT) imagery. The derived 1 km spatial resolution burned area maps were compared with 30 m spatial resolution maps obtained with 13 Landsat ETM+ scenes, through linear regression analysis. The procedure quantifies the bias in burned area estimation present in the low spatial resolution burned area map. Good correspondence was observed for seven sites, with values of the coefficient of determination (R²) ranging from 0.787 to 0.983. Poorer agreement was observed in four sites (R² values between 0.257 and 0.417), and intermediate values of R² (0.670 and 0.613) were obtained for two sites. The observed variation in the level of agreement between the Landsat and VGT estimates of area burned results from differences in the spatial pattern and size distribution of burns in the different fire regimes encompassed by our analysis. Small and fragmented burned areas result in large underestimation at 1 km spatial resolution. When large and compact burned areas dominate the landscape, VGT estimates of burned area are accurate, although in certain situations there is some overestimation. Accuracy of VGT burned area estimates also depends on vegetation type. Results showed that in forest ecosystems VGT maps underestimate substantially the amount of burned area. The most accurate estimates were obtained for woodlands and grasslands. An overall linear regression fitted with the data from the 13 comparison sites revealed that there is a strong relationship between VGT and Landsat estimates of burned area, with a value of R² of 0.754 and a slope of 0.803. Our findings indicate that burned area mapping based on 1 km spatial resolution VGT data provides adequate regional information.     

Comparative assessment of the measures of thematic classification accuracy

Accuracy assessment of classified imagery is an important task in remote sensing. Various measures have been developed to describe and compare the accuracy of maps and the performance of different classifiers, but the extent to which these measures are consistent with each other is largely unknown. In this paper the consistency of fourteen category-level and twenty map-level accuracy measures was tested on 595 published error matrices using nonparametric correlation coefficients (Spearman's rho and Kendall's tau-b) as well as the probability of concordance. The results show that four groups can be identified for the category-level measures and three groups for map-level measures. The consistency among the measures within a group is generally higher than that among the measures from different groups though all the measures at the same level are highly consistent with each other. We recommend that user's accuracy and producer's accuracy and the overall accuracy should be provided as primary accuracy measures and the two relative entropy change measures and the mutual information normalized by the arithmetic mean of the entropies on map and ground truthing be provided as supplementary measures. The chance-corrected, error matrix-normalized and user's and producer's accuracy-combined measures were found to contain estimation and interpretation problems at both category- and map-levels and are therefore not recommended for general use.     

The consequences of urban land transformation on net primary productivity in the United States

We use data from two satellites and a terrestrial carbon model to quantify the impact of urbanization on the carbon cycle and food production in the US as a result of reduced net primary productivity (NPP). Our results show that urbanization is taking place on the most fertile lands and hence has a disproportionately large overall negative impact on NPP. Urban land transformation in the US has reduced the amount of carbon fixed through photosynthesis by 0.04 pg per year or 1.6% of the pre-urban input. The reduction is enough to offset the 1.8% gain made by the conversion of land to agricultural use, even though urbanization covers an area less than 3% of the land surface in the US and agricultural lands approach 29% of the total land area. At local and regional scales, urbanization increases NPP in resource-limited regions and through localized warming “urban heat” contributes to the extension of the growing season in cold regions. In terms of biologically available energy, the loss of NPP due to urbanization of agricultural lands alone is equivalent to the caloric requirement of 16.5 million people, or about 6% of the US population.     

Assessing the accuracy of land cover change with imperfect ground reference data

The ground data used as a reference in the validation of land cover change products are often not an ideal gold standard but degraded by error. The effects of ground reference data error on the accuracy of land cover change detection and the accuracy of estimates of the extent of change were evaluated. Twelve data sets were simulated to allow the exploration of the impacts of a spectrum of ground data imperfections on the estimation of the producer's and user's accuracy of change as well as of change extent. Simulated data were used since this ensured that the actual properties of the data were known and to exclude effects due to other sources of ground reference data error; although the impacts of simulated reference data error on two real confusion matrices are also illustrated. The imperfections evaluated ranged from the inclusion of small amounts of known error into the ground reference data through to the extreme situation in which ground data were absent. The results show that even small amounts of error in the ground reference data can introduce large error into studies of land cover change by remote sensing and reinforce the desire to avoid the expression ground truth as this might imply that the data are a gold standard reference. The effect of reference data imperfections was dependent on the degree of association between the errors in the cross-tabulated data sets. For example, in the scenarios investigated, a 10% error in the reference data set introduced an underestimation of the producer's accuracy of 18.5% if the errors were independent but an over-estimation of the producer's accuracy of 12.3% if the errors were correlated. The magnitude of the mis-estimation of the producer's accuracy was also a function of the amount of change and greatest at low levels of change. The amount of land cover change estimated also varied greatly as a function of ground reference data error. Some possible methods to reduce or even remove the impacts of ground reference data error were illustrated. These ranged from simple algebraic means to estimate the actual values of accuracy and change extent if the imperfections were known through to a latent class analysis that allowed the assessment of classification accuracy and estimation of change extent without the use of ground reference data if the underlying model is defined appropriately.     

Mapping global urban areas using MODIS 500-m data: New methods and datasets based on ‘urban ecoregions’

Although cities, towns and settlements cover only a tiny fraction (< 1%) of the world's surface, urban areas are the nexus of human activity with more than 50% of the population and 70-90% of economic activity. As such, material and energy consumption, air pollution, and expanding impervious surface are all concentrated in urban areas, with important environmental implications at local, regional and potentially global scales. New ways to measure and monitor the built environment over large areas are thus critical to answering a wide range of environmental research questions related to the role of urbanization in climate, biogeochemistry and hydrological cycles. This paper presents a new dataset depicting global urban land at 500-m spatial resolution based on MODIS data (available at http://sage.wisc.edu/urbanenvironment.html). The methodological approach exploits temporal and spectral information in one year of MODIS observations, classified using a global training database and an ensemble decision-tree classification algorithm. To overcome confusion between urban and built-up lands and other land cover types, a stratification based on climate, vegetation, and urban topology was developed that allowed region-specific processing. Using reference data from a sample of 140 cities stratified by region, population size, and level of economic development, results show a mean overall accuracy of 93% (k = 0.65) at the pixel level and a high level of agreement at the city scale (R² = 0.90).     

Mapping urban change in the UK using satellite radar interferometry

A multitemporal sequence of ERS interferometric coherence data acquired between 1993 and 1999 are utilised for automatically mapping urban change within South Wales, UK. Validation of the change map derived from the coherence data is performed using independent, multidate, digital survey data of the city of Cardiff, UK. All major building developments that have occurred within the study area are located. There is evidence to suggest that this approach is generalisable for a wide range of coherence data and to other regions with similar landscapes.     

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.     

Effects of urbanization on the aquatic fauna of the Line Creek watershed, Atlanta—a satellite perspective

Impervious surface area (ISA) was derived for a period from 1979 to 1997 from Landsat MSS and TM data for the Line Creek watershed that lies to the south of the city of Atlanta, GA. The change in ISA is presented as an ecological indicator to examine the cumulative water resource impacts on mussel population in three sub-watersheds of Line Creek—namely, Line, Flat, and Whitewater creeks. The satellite analysis shows that ISA expansion occurred substantially from 1987 to 1997 and is predominantly in industrial, commercial, and shopping center (ICS) complexes but also in smaller lot-size residential development. Evidence of mussel habitat degradation is indicated and loss of species (in the region of 50 to 70%) is present in areas where ISA expansion is observed—specifically in ICS complex development in and around Peachtree City that drains directly into the Line and Flat creeks. This is in marked contrast to Whitewater Creek where overall development of ISA is less and no major loss of mussel species is observed.     

The accuracy of sequential aerial photography and SAR data for observing urban flood dynamics, a case study of the UK summer 2007 floods

In this paper we examine, for the first time, the potential of remote sensing to monitor flood dynamics in urban areas and constrain mathematical models of these processes. This is achieved through the development of a unique data set consisting of a series of eight space-borne synthetic aperture radar (SAR) and aerial photographic images of flooding of the UK town of Tewkesbury acquired over an eight day period in summer 2007. Previous observations of urban flooding have used single image and wrack mark data and have therefore been unable to adequately chart the propagation and recession of flood waves through complex urban topography. By using a combination of space-borne radar and aerial imagery we are able to show that remotely sensed imagery, particularly from the new TerraSAR-X radar, can reproduce dynamics adequately and support flood modelling in urban areas. We illustrate that image data from different remote sensing platforms reveal sufficient information to distinguish between models with varying degrees of channel-floodplain connectivity, particularly toward the end of the recession phase of the event. For this test case, our results also show that high resolution SAR imagery even when acquired from satellites can reveal important hydraulic characteristics difficult to simulate with current dynamic flood models. Hence, it is established, at least for this test case and event, that SAR imagery from as far as several hundred kilometers from the Earth's surface can deliver important information about floodplain dynamics that can be used to identify and help build suitable models, even in built-up environments.     

Assessing the impact of urban land development on net primary productivity in the southeastern United States

The southeastern United States (SE-US) has undergone one of the highest rates of landscape changes in the country due to changing demographics and land use practices over the last few decades. Increasing evidence indicates that these changes have impacted mesoscale weather patterns, biodiversity and water resources. Since the Southeast has one of the highest rates of land productivity in the nation, it is important to monitor the effects of such changes regularly. Here, we propose a remote sensing based methodology to estimate regional impacts of urban land development on ecosystem structure and function. As an indicator of ecosystem functioning, we chose net primary productivity (NPP), which is now routinely estimated from the MODerate resolution Imaging Spectroradiometer (MODIS) data. We used the MODIS data, a 1992 Landsat-based land cover map and nighttime data derived from the Defense Meteorological Satellite Program's Operational Linescan System (DMSP/OLS) for the years 1992/1993 and 2000 to estimate the extent of urban development and its impact on NPP. The analysis based on the nighttime data indicated that in 1992/1993, urban areas amounted to 4.5% of the total land surface of the region. In the year 2000, the nighttime data showed an increase in urban development for the southeastern United States of 1.9%. Estimates derived from the MODIS data indicated that land cover changes due to urban development that took place during the 1992-2000 period reduced annual NPP of the southeastern United States by 0.4%. Despite the uncertainties in sensor fusion and the coarse resolution of the data used in this study, results show that the combination of MODIS products such as NPP with nighttime data could provide rapid assessment of urban land cover changes and their impacts on regional ecosystem resources.     

Within country accuracy of tree crown transparency estimates using the image analysis system CROCO: a case study from Switzerland

We evaluated the accuracy of tree crown transparency estimates made with the image analysis system CROCO compared to visual estimates by field observers in Switzerland. Data from Swiss field training courses in 1998 and 1999 each comprising four sessions were used and the median values of the different observers’ scores were assumed as the true values. First, Swiss reference photographs alone were used to obtain the calibration curves needed to predict crown transparency from the image analysis measure DSO. In this case, the CROCO estimates were biased for many species. Second, data from the first three field training sessions were added to the reference photos to derive the calibration curves. In this second case, CROCO provided unbiased estimates of both years, while the estimates by some of the field teams were significantly different from the true values. The result suggests that CROCO can be used as a more reliable reference to detect observer bias within a country than control observers. We conclude that few reference photographs alone are not sufficient, so additional field data should also be used to estimate crown transparency more accurately using CROCO.     

An analysis of urban thermal characteristics and associated land cover in Tampa Bay and Las Vegas using Landsat satellite data

Remote sensing data from both Landsat 5 and Landsat 7 systems were utilized to assess urban area thermal characteristics in Tampa Bay watershed of west-central Florida, and the Las Vegas valley of southern Nevada. To quantitatively determine urban land use extents and development densities, sub-pixel impervious surface areas were mapped for both areas. The urban-rural boundaries and urban development densities were defined by selecting certain imperviousness threshold values and Landsat thermal bands were used to investigate urban surface thermal patterns. Analysis results suggest that urban surface thermal characteristics and patterns can be identified through qualitatively based urban land use and development density data. Results show the urban area of the Tampa Bay watershed has a daytime heating effect (heat-source), whereas the urban surface in Las Vegas has a daytime cooling effect (heat-sink). These thermal effects strongly correlated with urban development densities where higher percent imperviousness is usually associated with higher surface temperature. Using vegetation canopy coverage information, the spatial and temporal distributions of urban impervious surface and associated thermal characteristics are demonstrated to be very useful sources in quantifying urban land use, development intensity, and urban thermal patterns.     

Estimating the effect of crop classification error on evapotranspiration derived from remote sensing in the lower Colorado River basin, USA

In the U.S. Bureau of Reclamation's Lower Colorado River Accounting System (LCRAS), crop classifications derived from remote sensing are used to calculate regional estimates of crop evapotranspiration for water monitoring and management activities on the lower Colorado River basin. The LCRAS accuracy assessment was designed to quantify the impact of crop classification error on annual total crop evapotranspiration (ETc), as calculated from the Penman-Monteith method using the map crop classification as input. The accuracy assessment data were also used to generate a sample-based estimate of total ETc using the crop type identified by direct ground observation of each sample field. A stratified random sampling design was implemented using field size as the stratification variable. The stratified design did not markedly improve precision for the accuracy assessment objective, but it was highly effective for the objective of estimating ETc derived from the ground-observed crop types. The sampling design and analysis methodology developed for LCRAS demonstrates the utility of a multi-purpose approach that satisfies the accuracy assessment objectives, but also allows for rigorous, sample-based estimates of other collective properties of a region (e.g., total ETc in this study). We discuss key elements of this multi-purpose sampling strategy and the planning process used to implement such a strategy.     

Mapping of riparian zone attributes using discrete return LiDAR, QuickBird and SPOT-5 imagery: Assessing accuracy and costs

Riparian zones in Australia are exposed to increasing pressures because of disturbance from agricultural and urban expansion, weed invasion, and overgrazing. Accurate and cost-effective mapping of riparian environments is important for assessing riparian zone functions associated with water quality, biodiversity, and wildlife habitats. The objective of this research was to compare the accuracy and costs of mapping riparian zone attributes from image data acquired by three different sensor types, i.e. Light Detection and Ranging (LiDAR) (0.5-2.4 m pixels), and multi-spectral QuickBird (2.4 m pixels) and SPOT-5 (10 m pixels). These attributes included streambed width, riparian zone width, plant projective cover, longitudinal continuity, vegetation overhang, and bank stability. The riparian zone attributes were mapped for a study area along Mimosa Creek in the Fitzroy Catchment, Central Queensland, Australia. Object-based image and regression analyses were used for mapping the riparian zone attributes. The validation of the LiDAR, QuickBird, and SPOT-5 derived maps of streambed width (R = 0.99, 0.71, and 0.44 respectively) and riparian zone width (R = 0.91, 0.87, and 0.74 respectively) against field derived measurements produced the highest accuracies for the LiDAR data and the lowest using the SPOT-5 image data. Cross-validation estimates of misclassification produced a root mean square error of 1.06, 1.35 and 1.51 from an ordinal scale from 0 to 4 of the bank stability score for the LiDAR, QuickBird and SPOT-5 image data, respectively. The validation and empirical modelling showed high correlations for all datasets for mapping plant projective cover (R > 0.93). The SPOT-5 image data were unsuitable for assessment of riparian zone attributes at the spatial scale of Mimosa Creek and associated riparian zones. Cost estimates of image and field data acquisition and processing of the LiDAR, QuickBird, and SPOT-5 image data showed that discrete return LiDAR can be used for costs lower than those for QuickBird image data over large spatial extents (e.g. 26,000 km of streams). With the higher level of vegetation structural and landform information, mapping accuracies, geometric precision, and lower overall costs at large spatial extents, LiDAR data are a feasible means for assessment of riparian zone attributes.     

Positional accuracy of the Wide Area Augmentation System in consumer-grade GPS units

Global Positioning System devices are increasingly being used for data collection in many fields. Consumer-grade GPS units without differential correction have a published horizontal positional accuracy of approximately 10-15 m (average positional accuracy). An attractive option for differential correction for these GPS units is the Wide Area Augmentation System (WAAS). Most consumer-grade GPS units on the market are WAAS capable. According to the Federal Aviation Authority (FAA), the WAAS broadcast message provides integrity information about the GPS signal as well as accuracy improvements, which are reported to improve accuracy to 3-5 m. Limited empirical evidence has been published on the accuracy of WAAS-enabled GPS compared to autonomous GPS. An empirical study was conducted comparing the horizontal and vertical accuracy of WAAS-corrected GPS and autonomous GPS under ideal conditions using consumer-grade receivers. Data were collected for 30-min time spans over accurately surveyed control points. Metrics of median, 68th and 95th percentile, Root Mean Squared Error (RMSE), and average positional accuracy in the horizontal and vertical dimensions were computed and statistically compared. No statistically significant difference was found between WAAS and autonomous position fixes when using two different consumer-grade units. When using WAAS, a third unit type exhibited a statistically significant improvement in positional accuracy. Analysis of data collected for a 27-h time span indicates that while WAAS is altering the estimated position of a point compared to an autonomous position estimate, WAAS augmentation actually appears to decrease the positional accuracy.     

Errors in landmarking and the evaluation of the accuracy of traditional and 3D anthropometry

Body dimensions are based on landmarks of the body, but the magnitude of error in landmark determination is not well known. Therefore, a study was performed in which 40 subjects were marked five times in total by one highly skilled marker and a novice marker. Immediately after marking, a skilled measurer determined 34 body dimensions that were based on the mark locations. Intra- and inter-observer errors in landmarking of 35 landmarks, as well as those in 34 body dimensions were quantified. The error in landmarking was defined as the distance between two marks made on the same landmark by the same marker (intra-observer error) or by two different markers (inter-observer error). To make the first mark invisible when the second mark was made, the first mark was made using an invisible ink pen under black light. Landmarks with large intra-observer errors also had large inter-observer errors. Errors in body dimensions were smaller than landmarking errors in 23 measurements, which suggested that the magnitude of landmarking error would be underestimated from errors in body dimensions. In 15 body dimensions, measurements based on marks made by two different markers were not comparable according to the ISO 20685 criterion. Examination of body dimensions and landmarks with large inter-observer errors suggested that reducing inter-observer landmarking errors was necessary to reduce inter-observer measurement errors, and that a possible solution was to explicitly define landmarks with large errors in more detail so that anthropometrists can pinpoint them on the skin. Quantitative data on the intra- and inter-observer landmarking errors in the present study may be useful as a reference when evaluating and comparing the performance of software for calculating landmark locations for 3D anthropometry.     

Using the spatial and spectral precision of satellite imagery to predict wildlife occurrence patterns

We investigated the potential of using unclassified spectral data for predicting the distribution of three bird species over a ∼400,000 ha region of Michigan's Upper Peninsula using Landsat ETM+ imagery and 433 locations sampled for birds through point count surveys. These species, Black-throated Green Warbler, Nashville Warbler, and Ovenbird, were known to be associated with forest understory features during breeding. We examined the influences of varying two spatially explicit classification parameters on prediction accuracy: 1) the window size used to average spectral values in signature creation and 2) the threshold distance required for bird detections to be counted as present. Two accuracy measurements, proportion correctly classified (PCC) and Kappa, of maps predicting species' occurrences were calculated with ground data not used during classification. Maps were validated for all three species with Kappa values > 0.3 and PCC > 0.6. However, PCC provided little information other than a summary of sample plot frequencies used to classify species' presence and absence. Comparisons with rule-based maps created using the approach of Gap Analysis showed that spectral information predicted the occurrence of these species that use forest subcanopy components better than could be done using known land cover associations (Kappa values 0.1 to 0.3 higher than Gap Analysis maps). Accuracy statistics for each species were affected in different ways by the detection distance of point count surveys used to stratify plots into presence and absence classes. Moderate-to-large detection distances (100 m and 180 m) best classified maps of Black-throated Green Warbler and Nashville Warbler occurrences, while moderate detection distances (50 m and 100 m), which ignored remote observations, provided the best source of information for classification of Ovenbird occurrence. Window sizes used in signature creation also influenced accuracy statistics but to a lesser extent. Highest Kappa values of majority maps were typically obtained using moderate window sizes of 9 to 13 pixels (0.8 to 1.2 ha), which are representative of the study species territory sizes. The accuracy of wildlife occurrence maps classified from spectral data will therefore differ given the species of interest, the spatial precision of occurrence records used as ground references and the number of pixels included in spectral signatures. For these reasons, a quantitative examination is warranted to determine how subjective decisions made during image classifications affect prediction accuracies.