The accuracy of a land cover classification is the degree to which the map land cover agrees with the reference land cover classification (i.e. ground condition). The basic sampling designs historically implemented for map accuracy assessment have served well for the error matrix based analyses traditionally used. But contemporary applications of land cover maps place greater demands on accuracy assessment, and sampling designs must be constructed to target objectives such as accuracy of land cover composition and landscape pattern. Sampling designs differ in their suitability to achieve different objectives, and trade-offs among desirable sampling design criteria must be recognized and accommodated when selecting a design. An overview is presented of the sampling designs used in accuracy assessment, and the status of these designs is appraised for meeting current needs. Sampling design features that facilitate multiple-objective accuracy assessments are described.

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.

Disturbance is a critical ecological process in forested systems, and disturbance maps are important for understanding forest dynamics. Landsat data are a key remote sensing dataset for monitoring forest disturbance and there recently has been major growth in the development of disturbance mapping algorithms. Many of these algorithms take advantage of the high temporal data volume to mine subtle signals in Landsat time series, but as those signals become subtler, they are more likely to be mixed with noise in Landsat data. This study examines the similarity among seven different algorithms in their ability to map the full range of magnitudes of forest disturbance over six different Landsat scenes distributed across the conterminous US. The maps agreed very well in terms of the amount of undisturbed forest over time; however, for the ~30% of forest mapped as disturbed in a given year by at least one algorithm, there was little agreement about which pixels were affected. Algorithms that targeted higher-magnitude disturbances exhibited higher omission errors but lower commission errors than those targeting a broader range of disturbance magnitudes. These results suggest that a user of any given forest disturbance map should understand the map’s strengths and weaknesses (in terms of omission and commission error rates), with respect to the disturbance targets of interest.

A global validation database that can be used to assess the accuracy of multiple global and regional land-cover maps would yield significant cost savings and enhance comparisons of accuracy of different maps. Because the global validation database should expand over time as new validation data are contributed, the sampling design must be constructed so that it is simple to increase the sample size from a specific region (e.g. a continent or country) or from targeted land-cover classes to improve standard errors of the accuracy estimates. Stratified random sampling provides the desired adaptability to augment a sample to address regional or class-specific accuracy objectives. The proposed global validation database will be initiated from a baseline global stratified sample and then this baseline sample will be augmented to address accuracy objectives related to a specific map or region. The strata are constructed from a modified Köppen climate classification and population density. The theory and formulas for estimating accuracy from the combined baseline and augmented stratified samples are presented, and an example application is provided in which the regional accuracy of a land-cover map is assessed. The stratification used for the baseline global sample is retained when the augmented sample is subsequently selected. Alternatively, it is possible to ‘restratify’ the design so that the land-cover classes of a particular map are used as strata when selecting an augmented sample. The protocol for restratifying the design is presented, but the complexity of this option makes it less practical than retaining the initial strata when selecting an augmented sample.

The ground reference data obtained to assess map accuracy can be simultaneously used to estimate area (extent). This dual-purpose use of ground reference data is examined for the special case of a two-class map of ‘change’ and ‘no change’. To assess the accuracy of a change map, stratified sampling is often implemented with a disproportionately larger sample size allocated to the map change stratum. But this allocation targeting user's accuracy of change is not necessarily effective for the competing objective of estimating the area of change. Sampling theory provides the basis for deciding a sample size allocation to strata when multiple, but competing estimation objectives are of interest. Neyman optimal allocation is preferred for estimating the area of change as well as overall accuracy, whereas equal allocation is effective for estimating user's accuracy. The results and recommendations developed in this article extend to any dichotomous classification in which one class is relatively rare.

Shifting cultivation has been shown to be the primary cause of land use change in the Democratic Republic of Congo (DRC). Traditionally, forested and fallow land are rotated in a slash and burn cycle that has created an agricultural mosaic, including secondary forest, known as the rural complex. This study investigates the land use context of new forest clearing (during 2000–2015) in primary forest areas outside of the established rural complex. These new forest clearings occur as either rural complex expansion (RCE) or isolated forest perforations (IFP), with consequent implications on the forest ecosystem and biodiversity habitat. During 2000–2015, subsistence agriculture was the dominant driver of forest clearing for both extension of settled areas and pioneer clearings removed from settled areas. Less than 1% of clearing was directly attributable to land uses such as mining, plantations, and logging, showing that the impact of commercial operations in the DRC is currently dwarfed by a reliance on small-holder shifting cultivation. However, analyzing the landscape context showed that large-scale agroindustry and resource extraction activities lead to increased forest loss and degradation beyond their previously-understood footprints. The worker populations drawn to these areas create communities that rely on shifting cultivation and non-timber forest products (NTFP) for food, energy, and building materials. An estimated 12% of forest loss within the RCE and 9% of the area of IFP was found to be within 5 km of mines, logging, or plantations. Given increasing demographic and commercial pressures on DRC’s forests, it will be crucial to factor in this landscape-level land use change dynamic in land use planning and sustainability-focused governance.