Anthropogenic climate change has driven over 5 million km² of drylands towards desertification

Land degradation is a significant global issue, with varying estimates of affected areas. Dryland vegetation changes are primarily driven by anthropogenic climate change (ACC), encompassing altered water availability, temperature increases, and CO2 fertilization effects, and land use (LU) practices like grazing, cropping, and deforestation. Unsustainable LU is generally considered the main negative driver of dryland degradation, although climate change's negative impacts are also substantial. Paradoxically, satellite data since 1980 show increased global vegetation greening, attributed to CO2 fertilization, increased rainfall, temperature changes, and LU. However, model discrepancies exist regarding the relative contributions of these factors, with some attributing almost all greening to CO2 fertilization while others emphasize climate or LU. These models often don't account for rapid ecosystem changes (breakpoints) caused by events such as extreme fires or shifts in agricultural practices. This study aims to quantify the extent of global desertification, defined as land degradation in arid, semi-arid, and dry sub-humid areas, considering ACC, LU, climate variability (CV), and CO2 fertilization while accounting for rapid ecosystem shifts. The study utilizes satellite-based GIMMSv3.1g Normalized Difference Vegetation Index (NDVI) data and a modified Time Series Segmented Residual Trends (TSS-RESTREND) method to attribute vegetation changes to the identified drivers.

Existing research on land degradation and desertification presents a range of estimates for the extent of degraded areas, from less than 10 to 60 million km². Studies have shown the negative impacts of unsustainable land use practices on dryland ecosystems, alongside evidence suggesting that anthropogenic forcing has already led to increased arid areas. However, a global increase in vegetation greenness since 1980 is documented, with multiple studies offering contrasting explanations for this trend, highlighting the difficulty in disentangling the relative contribution of climate change, CO2 fertilization, and land-use change. The challenge of accounting for rapid ecosystem changes caused by discrete events such as fires or abrupt policy changes is also a prominent knowledge gap identified in previous work. The UNCCD and IPCC definitions of desertification, and the use of NDVI as a proxy for vegetation growth, are also discussed in relation to existing literature.

This study quantified global desertification using the UNCCD definition of land degradation and the GIMMSv3.1g NDVI dataset (1982–2015). A non-parametric trend analysis (Theil-Sen slope estimator and Spearman's significance test) was applied to peak growing season NDVI (NDVImax) to assess overall vegetation change. Attribution of these changes to CO2, climate variability (CV), climate change (CC), and land use (LU) was performed using a modified TSS-RESTREND method. This approach quantifies the effects of interannual CV, long-term climate changes, CO2 fertilization, and ecosystem breakpoints caused by LU. A 12-member ensemble of statistical model runs was employed to address uncertainties arising from different gridded datasets (four precipitation and three temperature datasets). To account for the CO2 fertilization effect, a theoretical relationship linking increased photosynthesis to rising CO2 concentrations was used to scale the NDVI data and remove the CO2 effect. The scaled NDVI data were then used with the TSS-RESTREND method to attribute the remaining changes to climate and land use, separating climate change from climate variability. The method incorporates a change detection component to identify structural changes in ecosystems and uses a weighted mean of ensemble results considering both C3 and C4 photosynthetic pathways. Statistical significance was determined using the Spearman's rank correlation coefficient test and the Benjamini-Hochberg procedure to control the false discovery rate. The IPCC protocol for determining ensemble significance and agreement was also applied. Multiple datasets were used for precipitation, temperature, and land cover data (see Table 1 for details).

Globally, 6% of drylands experienced desertification (significant negative NDVImax change) between 1982 and 2015, while 41% showed significant greening, and 53% experienced no significant change. CO2 fertilization was the largest absolute driver of global dryland vegetation change, followed by LU, CV, and CC. However, the global mean per-pixel contribution of CO2 was significantly larger than those of CV, CC, or LU. When CO2 was removed from the analysis, 60.4% of global dryland vegetation change was attributed to LU, and 39.6% to CC and variability, highlighting the importance of explicitly considering the CO2 fertilization effect. Anthropogenic climate change (ACC, combining CO2 fertilization and CC) had a net positive (greening) effect globally, but also a desertifying impact across 12.55% (5.43 million km²) of drylands, disproportionately affecting poorer nations. In areas experiencing desertification, negative LU was the primary driver in 79.9% and a contributing factor in 99% of areas. Even with smaller average effects compared to LU, climate change remained an important factor, compounding the negative impacts of unsustainable land use. Widespread greening was observed, primarily driven by CO2 fertilization, but LU and CV were also crucial in certain regions such as the Sahel, India, China, and Australia. The study employed an ensemble approach to minimize uncertainties associated with observational datasets and structural change detection to account for ecosystem breakpoints.

This study addresses the research question of quantifying the extent and drivers of global desertification by providing a comprehensive assessment of the relative contribution of various factors. The findings highlight the complex interplay between anthropogenic climate change and land use in shaping dryland vegetation dynamics. The significance of CO2 fertilization as a global driver of greening is confirmed, but the study demonstrates the importance of considering land-use changes, climate variability, and climate change in a spatially explicit manner to understand regional trends. The disproportionate impact of desertification on developing economies underscores the need for effective policy interventions. The results are consistent with some regional studies but offer a larger scale and more nuanced view by explicitly considering CO2 fertilization and the dynamic nature of land-use change. The study’s findings are relevant to policymakers, researchers, and conservationists working on dryland management and climate change mitigation.

This study provides a comprehensive assessment of global desertification, highlighting the significant contribution of anthropogenic climate change and unsustainable land use practices. Despite widespread greening, a considerable area of drylands has undergone desertification, impacting millions of people, predominantly in developing countries. The importance of accounting for both the CO2 fertilization effect and the complex spatial patterns of land use changes in future studies and mitigation efforts is emphasized. Future research could focus on improving the resolution of datasets in specific regions, exploring the interaction between different drivers of change at finer scales and investigating the effects of different policy interventions on preventing desertification.

The study relies on NDVI as a proxy for vegetation productivity, which may not capture all aspects of land degradation (e.g., shrub encroachment). The accuracy of the attribution analysis is dependent on the quality and availability of climate and land use datasets, particularly in data-scarce regions. The model used assumes a linear relationship between NPP and NDVI, which might not be entirely accurate in all dryland ecosystems. While the ensemble approach reduces uncertainties, some remain due to the inherent variability in climate and land use data. The study's focus is on long-term trends, and it may not capture short-term fluctuations in vegetation or impacts from unusual events.