Remote sensing alone is insufficient for quantifying changes in forest cover

A recent PNAS article (1) estimated global gross forest cover loss. Figure 1 in ref. 1 and the conclusions drawn from both refs. 1 and 2 leave the impression that tree cover losses in the southern United States are no different from deforestation activities elsewhere around the globe. Results from the Forest Service’s Forest Inventory and Analysis (FIA) program depict a different, and we think more accurate, assessment of forest conditions in the southern United States, because field plots are used to validate satellite interpretations and distinguish between land cover and land use changes.

Since the 1940s, FIA has combined information from remotely sensed imagery (in the early days, aerial photography; today, LANDSAT ETM+ and MODIS imagery) with spatially coregistered data collected on ground plots (125,000 across the United States) to track changes in land use and land cover over time. In the south, fresh imagery is interpreted, and the same ground plots are revisited every 5 y. This system of paired remote sensing and ground observation confers four distinct benefits: (i) field visits validate imagery interpretations, significantly improving the accuracy of inventory statistics for both land cover and land use changes; (ii) field visits help identify where young tree seedlings are established before they are discernible on satellite imagery; (iii) repetition provides solid trend data; and (iv) changes in tree cover vs. forest land use are clarified.

A major shortcoming in ref. 1 is that the analysis is one-sided, reporting only on gross loss of tree cover over a single 5-y period. Because of absent equivalent data on gross gain in tree cover over the same time period, the net change in land cover cannot be estimated. Contrary to the authors’ assertion that net changes are relatively unimportant, our experience, both domestically and internationally, is that the trend of net change over time is particularly important to interest groups and policy makers (1). The study by Smith et al. (3) is the latest of nine detailed national FIA reports dating back to 1953 chronicling changes in forests—both as land cover and land use. Forest area in the southern United States was essentially unchanged from 1997 (214.1 million acres) to 2007 (214.644 million acres), bracketing the period used in ref. 1. Furthermore, net volume of all trees south-wide increased 12% between 1997 and 2007 (256.3–288.5 billion cubic feet), even after all losses from both natural and human-caused events are counted. These statistics present a more holistic picture of forest conditions in the southern United States than the picture emerging from refs. 1 and 2. Southern forests are being replenished and are growing well.

Although we use before and after LANDSAT data to estimate losses from disturbances, such as large fires (4), we recognize the importance of also characterizing recovery trends with remote sensing data (e.g., MODIS) (5) and field visits. This more thorough approach better characterizes disturbance and recovery than the approach used in ref. 1.

The United Nations Food and Agricultural Organization has endorsed the principle of validating satellite-derived estimates with ground-plot data or finer-resolution imagery. In the United States, we are leading this validation effort.