Measuring the planet to fill terrestrial data gaps.

H ow a growing world population can sustain itself on Planet Earth has been the focus of much research. Norman Borlaug (1), the ‘‘father of the Green Revolution,’’ argued that one answer is to intensify agriculture to produce more food on the same area of land. But what he promoted as a policy strategy, others have called the ‘‘Borlaug hypothesis,’’ aiming to add it to sustainability science theory and test it empirically. In this issue of PNAS Rudel et al. (2) follow earlier national studies (3) by testing the hypothesis at global scale. They demonstrate the existence of the general phenomenon of ‘‘land sparing,’’ by showing that cropland area has increased more slowly than population since 1970. However, they find relatively little evidence that intensification has gone further, by shrinking cropland and generating surplus ‘‘spared land,’’ and much of this evidence is linked to changes in trade patterns. Yet their article will prove important for introducing the concept of spared land into the literature, inspiring more research, and stimulating debate about how land sparing relates to existing theory. This commentary focuses on a key challenge they identify (the measurement of land sparing) and wider monitoring issues raised by it. It refers mainly to developing countries. Rudel et al. (2) actually discuss two types of land sparing. The first type, which I shall call “sparing land,” reduces the rate of farmland expansion. To demonstrate it requires evidence that farmland has expanded more slowly than population and (in the tropics) that deforestation rates have fallen. It can be predicted by models, e.g., using a simple model showed that if mean farm yields rose by 1–2% a 1 (depending on the region and scenario) deforestation rates could fall by 44–88% between 1980 and 2020, saving 59–64 ha.106 of forest (4). The second type, which Rudel et al. (2) term “spared land,” is generated by farmland contraction. To identify it requires evidence for the latter, and that the surplus has been transferred to another land use via various routes (Fig. 1). When a time series of estimates of tropical moist forest area for 1973–2000 did not show the expected decline, it led to an hypothesis that continuing deforestation could be offset by natural reforestation elsewhere (5). Spared land is where this reforestation could occur, and so spared land could become a new intermediate item in carbon accounting. Both types of land sparing control deforestation, helping countries to pass through their ‘‘forest transitions’’ (6). But whereas sparing land depends on technological innovation, spared land can appear in any country once all highly fertile areas are identified and less fertile lands are abandoned or transferred to less intensive uses. Rudel et al. (2) infer the existence of both types from trends in the areas of 10 major crops. This approach has limitations, which they discuss. Whether they could confirm the sparing land phenomenon in the tropics is debatable, because uncertainty about national deforestation rates makes it difficult to find convincing evidence for their decline. The main source of estimates are the Forest Resources Assessments (FRAs) of the United Nations Food and Agriculture Organization (FAO), which rely largely on government statistics, not scientific measurements (5). Two forest area data points are needed to determine a trend. Yet only half of tropical countries have had two national forest surveys, of which the latest occurred after 1980 (Table 1). The Parties to the United Nations Framework Convention on Climate Change hope to introduce a Reduced Emissions from Deforestation and Degradation (REDD) scheme to help developing countries cut deforestation rates below recent reference levels; Table 1 shows that determining these levels may be difficult. Rudel et al. (2) are severely constrained in their ability to convincingly identify or confirm the appearance of spared land because of the poor quality of data at their disposal: (i) They base their analysis on international agricultural statistics, also published by FAO, which lack the 100% coverage needed to confirm land transfers within farming uses and to other uses. Until 1996 FAO listed in its Production Yearbook (now the Statistical Yearbook) series national areas of arable crops, permanent crops, permanent pastures, forests and woodlands, and other land, which combined to cover 100% of national land area. These estimates were never very reliable, but from 1996 those for permanent pastures and forests and woodlands were discontinued, and subsumed with ‘‘other land’’ in a ‘‘nonarable and nonpermanent crops’’ category, because governments were not reporting grazing land (or shifting cultivation) ar-