A meta-analysis at a local scale reveals that land-use change has caused species richness to decline by approximately 8.1% on average globally, mainly as a result of large increases in croplands and pastures. See Article p.45
The main effects humans have on our planet seem to manifest in factors of two: we have doubled the rate at which nitrogen enters the biosphere by using fertilizer; we have diverted half of the fresh water and half of all plant productivity for our own purposes; and we have modified about half of the planet's land1,2. It is widely speculated that the last of these — modifying roughly 50% of all land — is the biggest human-caused threat to biodiversity, but this theory has never been comprehensively assessed. On page 45 of this issue, Newbold et al.3 describe an ambitious attempt to evaluate the global impact of land-use change on terrestrial biodiversity.
The authors assembled a data set of more than 380 previous studies comparing the biodiversity of sites with no human change (original or primary vegetation) with similar sites modified for human use. They combined this data set with further data on the global land-use changes made by humans over the past 500 years4, and also with several predictions of how humans might modify land use over the next 100 years. The processes controlling biodiversity are highly scale-specific5, and most previous studies have focused either on extinctions at global scales or on the number of species in certain regions (and the latter is actually often increasing6). By contrast, Newbold et al. analysed their data at a local scale, typically smaller than a football field, which is more relevant to the way humans interact with nature.
The headline finding is that land-use change has caused the number of species (species richness) contained in these small plots of land to decline by 8.1% over 500 years when averaged across the globe. The authors also find a 10.7% decline in the number of individual organisms, with an additional decline in richness resulting from this loss of individuals rather than from an underlying biodiversity loss.
The study excels in providing details of where and why these losses occurred. Types of human modification that were evaluated include conversions to pasture, cropland, tree plantations, urbanization and secondary vegetation (land that was disturbed but allowed to regrow). The authors found that areas of secondary vegetation show little species loss, whereas land converted to pasture, cropland and urban areas show heavy loss (see Figure 1 in the paper3). Considering that the most common human land-use conversions are to secondary vegetation, cropland and pasture, it becomes clear that most of the impact on biodiversity has resulted from conversion of pristine vegetation to cropland and pasture (Fig. 1). As a result, most of the loss has occurred in the prairie steppes of North America and Eurasia, in heavily grazed subtropical dry areas, and in countries that have experienced rapid agricultural growth to support heavy population growth, such as Indonesia, India, Brazil and China (see Figure 3 of the paper3). However, the average 8.1% decline hides a lot of variability — for example, the authors find that low-intensity urban land use and moderate human densities (roughly speaking, suburbs) are actually associated with increased species richness.
Three other recent papers7,8,9 also assembled data from large numbers of studies at similar scales and over a variety of time periods, but these studies found that, on average, there has been no change in local species richness. The difference between these findings and those of Newbold et al. is quite small (0% over 3–260 years versus 8% over 500 years) relative to the uncertainty (the 95% error range in Newbold and colleagues' study is 3.5–12.9% species loss, not including uncertainties in land-use change). However, the perceived qualitative difference will loom large in people's minds.
One possible explanation for the discrepancy is that different methods were used — two of the previous analyses7,8 used studies that tracked diversity on single pieces of land through time, which has the advantage of directly measuring the quantity of interest but the disadvantage of potentially introducing spatial sampling biases. But I suspect the main explanation is that Newbold et al. targeted a single human impact — land-use change — whereas the three other studies averaged across all forms of human impact. Some of these, such as increased fertilization and bringing about species invasions, can actually increase species richness (and might also explain the species increase in suburban regions found by Newbold and colleagues).
It would be odd if the negative effects of land-use change documented by Newbold et al. were exactly counterbalanced, such that the net effect of all types of human impacts averaged out to zero (at the local scale). Yet that might be the most parsimonious explanation for the results across these four studies. And it might not be so odd if ecological processes strongly regulate local species richness10. This would require rates of local extinction to increase when there are too many species, and arrival rates of locally novel species to increase when there are too few species, effectively pushing the system towards a 'set point' for species richness. There is good evidence for this in an island context11, but more study of such processes is needed for mainland communities10.
Newbold and colleagues' study has two limitations that are worthy of comment. First, their compilation of the literature demonstrates the existence of publication bias — studies showing positive effects were published less often than statistically likely. Newbold and colleagues have been careful to address this problem and I do not think it greatly affected their results. But it is a wake-up call that our literature on this topic is biased.
Second, their estimates of diversity loss focused on the endpoints of changes, such as urbanization versus pasture, without segregating the data on the basis of the land's starting point, such as grassland versus forest. It seems probable that there have been larger declines in species richness in land that started as forest than as grassland (because forests typically start with more species), but many of the conversions to pastures and croplands that drive the authors' results started as grasslands, not forests — this could lead to a bias in the results. As better data become available, future studies should unpack these nuances.
Despite these limitations, Newbold and colleagues' study is by far the best empirical demonstration to date that land-use change negatively affects local biodiversity on average, and effectively puts this question beyond debate. It also shows that the greatest negative impacts come from the introduction of cropland and pasture, both in areas of rapid population growth and in the breadbaskets of Eurasia and North America, rather than from urbanized land and harvested land that is allowed to regrow. This has clear implications for conservation policies and priorities.
The scenarios examined by the authors for the next 100 years suggest that average species richness will fall by a further 3.4% if there are no interventions into practices of land-use change. Yet they also show that, even within the range of what economists consider to be realistic mitigation scenarios, we have the ability to reverse trends in biodiversity loss and even to increase average local richness. This paper provides clear guidance for policy-makers — the ball is now in their court.
About this article
Journal of Geographical Sciences (2016)