Stephen V. Stehman

High-resolution global maps of 21st-century forest cover change.

Forests in Flux Forests worldwide are in a state of flux, with accelerating losses in some regions and gains in others. Hansen et al. (p. 850) examined global Landsat data at a 30-meter spatial resolution to characterize forest extent, loss, and gain from 2000 to 2012. Globally, 2.3 million square kilometers of forest were lost during the 12-year study period and 0.8 million square kilometers of new forest were gained. The tropics exhibited both the greatest losses and the greatest gains (through regrowth and plantation), with losses outstripping gains. Landsat data reveals details of forest losses and gains across the globe on an annual basis from 2000 to 2012. Quantification of global forest change has been lacking despite the recognized importance of forest ecosystem services. In this study, Earth observation satellite data were used to map global forest loss (2.3 million square kilometers) and gain (0.8 million square kilometers) from 2000 to 2012 at a spatial resolution of 30 meters. The tropics were the only climate domain to exhibit a trend, with forest loss increasing by 2101 square kilometers per year. Brazil’s well-documented reduction in deforestation was offset by increasing forest loss in Indonesia, Malaysia, Paraguay, Bolivia, Zambia, Angola, and elsewhere. Intensive forestry practiced within subtropical forests resulted in the highest rates of forest change globally. Boreal forest loss due largely to fire and forestry was second to that in the tropics in absolute and proportional terms. These results depict a globally consistent and locally relevant record of forest change.

Selecting and interpreting measures of thematic classification accuracy

Abstract An error matrix is frequently employed to organize and display information used to assess the thematic accuracy of a land-cover map, and numerous accuracy measures have been proposed for summarizing the information contained in this error matrix. No one measure is universally best for all accuracy assessment objectives, and different accuracy measures may lead to conflicting conclusions because the measures do not represent accuracy in the same tray. Choosing appropriate accuracy measures that address objectives of the mapping project is critical. Characteristics of some commonly used accuracy measures are described, and relationships among these measures are provided to aid the user in choosing an appropriate measure. Accuracy measures that are directly interpretable as probabilities of encountering certain types of misclassification errors or correct classifications should be selected in. preference to measures not interpretable as such. Users and producers accuracy and the overall proportion OF area correctly classified are examples of accuracy measures possessing the desired probabilistic interpretation. The kappa coefficient of agreement does not possess such a probabilistic interpretation because of the adjustment for hypothetical chance agreement incorporated into this measure, and the strong dependence of kappa on the marginal proportions of the error matrix makes the utility of kappa for comparisons suspect. Normalizing an error matrix results in estimates that are not consistent for accuracy paramteters of the map being assessed, so that this procedure is generally not warranted for most applications.

Humid tropical forest clearing from 2000 to 2005 quantified by using multitemporal and multiresolution remotely sensed data

Forest cover is an important input variable for assessing changes to carbon stocks, climate and hydrological systems, biodiversity richness, and other sustainability science disciplines. Despite incremental improvements in our ability to quantify rates of forest clearing, there is still no definitive understanding on global trends. Without timely and accurate forest monitoring methods, policy responses will be uninformed concerning the most basic facts of forest cover change. Results of a feasible and cost-effective monitoring strategy are presented that enable timely, precise, and internally consistent estimates of forest clearing within the humid tropics. A probability-based sampling approach that synergistically employs low and high spatial resolution satellite datasets was used to quantify humid tropical forest clearing from 2000 to 2005. Forest clearing is estimated to be 1.39% (SE 0.084%) of the total biome area. This translates to an estimated forest area cleared of 27.2 million hectares (SE 2.28 million hectares), and represents a 2.36% reduction in area of humid tropical forest. Fifty-five percent of total biome clearing occurs within only 6% of the biome area, emphasizing the presence of forest clearing “hotspots.” Forest loss in Brazil accounts for 47.8% of total biome clearing, nearly four times that of the next highest country, Indonesia, which accounts for 12.8%. Over three-fifths of clearing occurs in Latin America and over one-third in Asia. Africa contributes 5.4% to the estimated loss of humid tropical forest cover, reflecting the absence of current agro-industrial scale clearing in humid tropical Africa.

Design and Analysis for Thematic Map Accuracy Assessment: Fundamental Principles

Abstract Before being used in scientific investigations and policy decisions, thematic maps constructed from remotely sensed data should be subjected to a statistically rigorous accuracy assessment. The three basic components of an accuracy assessment are: 1) the sampling design used to select the reference sample; 2) the response design used to obtain the reference land-cover classification for each sampling unit; and 3) the estimation and analysis procedures. We discuss options available for each of these components. A statistically rigorous assessment requires both a probability sampling design and statistically consistent estimators of accuracy parameters, along with a response design determined in accordance with features of the mapping and classification process such as the land-cover classification scheme, minimum mapping unit, and spatial scale of the mapping.

Quantification of global gross forest cover loss

A globally consistent methodology using satellite imagery was implemented to quantify gross forest cover loss (GFCL) from 2000 to 2005 and to compare GFCL among biomes, continents, and countries. GFCL is defined as the area of forest cover removed because of any disturbance, including both natural and human-induced causes. GFCL was estimated to be 1,011,000 km2 from 2000 to 2005, representing 3.1% (0.6% per year) of the year 2000 estimated total forest area of 32,688,000 km2. The boreal biome experienced the largest area of GFCL, followed by the humid tropical, dry tropical, and temperate biomes. GFCL expressed as the proportion of year 2000 forest cover was highest in the boreal biome and lowest in the humid tropics. Among continents, North America had the largest total area and largest proportion of year 2000 GFCL. At national scales, Brazil experienced the largest area of GFCL over the study period, 165,000 km2, followed by Canada at 160,000 km2. Of the countries with >1,000,000 km2 of forest cover, the United States exhibited the greatest proportional GFCL and the Democratic Republic of Congo the least. Our results illustrate a pervasive global GFCL dynamic. However, GFCL represents only one component of net change, and the processes driving GFCL and rates of recovery from GFCL differ regionally. For example, the majority of estimated GFCL for the boreal biome is due to a naturally induced fire dynamic. To fully characterize global forest change dynamics, remote sensing efforts must extend beyond estimating GFCL to identify proximate causes of forest cover loss and to estimate recovery rates from GFCL.

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

Abstract 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.

Quantifying changes in the rates of forest clearing in Indonesia from 1990 to 2005 using remotely sensed data sets

Timely and accurate data on forest change within Indonesia is required to provide government, private and civil society interests with the information needed to improve forest management. The forest clearing rate in Indonesia is among the highest reported by the United Nations Food and Agriculture Organization (FAO), behind only Brazil in terms of forest area lost. While the rate of forest loss reported by FAO was constant from 1990 through 2005 (1.87 Mha yr −1 ), the political, economic, social and environmental drivers of forest clearing changed at the close of the last century. We employed a consistent methodology and data source to quantify forest clearing from 1990 to 2000 and from 2000 to 2005. Results show a dramatic reduction in clearing from a 1990s average of 1.78 Mha yr −1 to an average of 0.71 Mha yr −1 from 2000 to 2005. However, annual forest cover loss indicator maps reveal a near-monotonic increase in clearing from a low in 2000 to a high in 2005. Results illustrate a dramatic downturn in forest clearing at the turn of the century followed by a steady resurgence thereafter to levels estimated to exceed 1 Mha yr −1 by 2005. The lowlands of Sumatra and Kalimantan were the site of more than 70% of total forest clearing within Indonesia for both epochs; over 40% of the lowland forests of these island groups were cleared from 1990 to 2005. The method employed enables the derivation of internally consistent, national-scale changes in the rates of forest clearing, results that can inform carbon accounting programs such as the Reducing Emissions from Deforestation and Forest Degradation in Developing Countries (REDD) initiative.

Effects of landscape characteristics on land-cover class accuracy

The effects of patch size and land-cover heterogeneity on classification accuracy were evaluated using reference data collected for the National Land-Cover Data (NLCD) set accuracy assessment. Logistic regression models quantified the relationship between classification accuracy and these landscape variables for each land-cover class at both the Anderson Levels I and II classification schemes employed in the NLCD. The general relationships were consistent, with the odds of correctly classifying a pixel increasing as patch size increased and decreasing as heterogeneity increased. Specific characteristics of these relationships, however, showed considerable diversity among the various classes. Odds ratios are reported to document these relationships. Interaction between the two landscape variables was not a significant influence on classification accuracy, indicating that the effect of heterogeneity was not impacted by the sample being in a small or large patch. Landscape variables remained significant predictors of class-specific accuracy even when adjusted for regional differences in the mapping and assessment processes or landscape characteristics. The land-cover class-specific analyses provide insight into sources of classification error and a capacity for predicting error based on a pixels mapped land-cover class, patch size and surrounding land-cover heterogeneity.

Sampling designs for accuracy assessment of land cover.

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.

Time-series analysis of multi-resolution optical imagery for quantifying forest cover loss in Sumatra and Kalimantan, Indonesia

Monitoring loss of humid tropical forests via remotely sensed imagery is critical for a number of environmental monitoring objectives, including carbon accounting, biodiversity, and climate modeling science applications. Landsat imagery, provided free of charge by the U.S. Geological Survey Center for Earth Resources Observation and Science (USGS/EROS), enables consistent and timely forest cover loss updates from regional to biome scales. The Indonesian islands of Sumatra and Kalimantan are a center of significant forest cover change within the humid tropics with implications for carbon dynamics, biodiversity maintenance and local livelihoods. Sumatra and Kalimantan feature poor observational coverage compared to other centers of humid tropical forest change, such as Mato Grosso, Brazil, due to the lack of ongoing acquisitions from nearby ground stations and the persistence of cloud cover obscuring the land surface. At the same time, forest change in Indonesia is transient and does not always result in deforestation, as cleared forests are rapidly replaced by timber plantations and oil palm estates. Epochal composites, where single best observations are selected over a given time interval and used to quantify change, are one option for monitoring forest change in cloudy regions. However, the frequency of forest cover change in Indonesia confounds the ability of image composite pairs to quantify all change. Transient change occurring between composite periods is often missed and the length of time required for creating a cloud-free composite often obscures change occurring within the composite period itself. In this paper, we analyzed all Landsat 7 imagery with <50% cloud cover and data and products from the Moderate Resolution Imaging Spectroradiometer (MODIS) to quantify forest cover loss for Sumatra and Kalimantan from 2000 to 2005. We demonstrated that time-series approaches examining all good land observations are more accurate in mapping forest cover change in Indonesia than change maps based on image composites. Unlike other time-series analyses employing observations with a consistent periodicity, our study area was characterized by highly unequal observation counts and frequencies due to persistent cloud cover, scan line corrector off (SLC-off) gaps, and the absence of a complete archive. Our method accounts for this variation by generating a generic variable space. We evaluated our results against an independent probability sample-based estimate of gross forest cover loss and expert mapped gross forest cover loss at 64 sample sites. The mapped gross forest cover loss for Sumatra and Kalimantan was 2.86% of the land area, or 2.86 Mha from 2000 to 2005, with the highest concentration having occurred in Riau and Kalimantan Tengah provinces.