Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Pays-Bas → 2008 → Global drought in the 20th and 21st centuries : analysis of retrospective simulations and future projections of soil moisture

Wageningen Universiteit (2008)

Global drought in the 20th and 21st centuries : analysis of retrospective simulations and future projections of soil moisture

Sheffield, J.

Titre : Global drought in the 20th and 21st centuries : analysis of retrospective simulations and future projections of soil moisture

Auteur : Sheffield, J.

Université de soutenance : Wageningen Universiteit

Grade : PhD thesis 2008

We describe the analysis of global and regional drought over the second half of the 20th century from a retrospective model simulation of the terrestrial water cycle, and projected 21st century changes using multi-scenario data from multiple climate models. A global meteorological forcing dataset is developed for 1948-2000 to drive the retrospective simulation by combining observations with reanalysis. Biases in the reanalysis precipitation, temperature and radiation are corrected for systematic bias and spurious trends, which exert erroneous effects on the land water budgets. A monthly soil moisture based drought index is developed from the simulation and is used to investigate the occurrence, variability and trends in drought for 1950-2000. The frequencies of short-term droughts (6 months and less) are highest in humid regions. Medium term droughts (6-12 months) are more prevalent in mid- to high-latitudes, driven by persistent frozen soil moisture anomalies. Over the Sahel and parts of high northern latitudes, the frequency of long-term droughts is at a maximum. Severe drought events are systematically identified in terms of spatial coverage, including the 1988 USA, 1982/83 Australian, 1983/4 Sahel and 1965/66 Indian droughts. There is an overall increasing trend in global soil moisture, driven by precipitation, reflected especially in North America. Regional variation is nevertheless apparent and significant drying over West Africa, stands out. Trends in drought characteristics are mostly decreasing but statistically significant changes are limited in areal extent and generally less than 10% of continental areas. Concurrent decreases in global drought spatial extent are 0.04% yr−1. Within the long-term trends we find interannual and decadal variations in soil moisture and drought characteristics driven mainly by ENSO variability, although the AMO plays an important role in many regions. Drought is driven primarily by variability in precipitation, but temperature has an effect that appears to be exaggerated in the late 20th century, especially in high northern latitudes. At global scales the soil moisture index and the PDSI are reasonably well correlated but this breaks down in cooler regions and seasons, and notably for recent years when the PDSI shows a larger drying trend, possibly due to its temperature-based evaporation estimate. To investigate future projected changes in drought, soil moisture data is analyzed for three future IPCC AR4 climate scenarios (B1, A1B, A2) from eight GCMs. A decrease in 21st century global soil moisture is accompanied by a doubling of the spatial extent and frequency of short-term droughts. Long-term droughts become three v times more common. Regionally, the Mediterranean, West African, Central Asian and Central American regions show large increases, as does mid-latitude North America but with larger inter-scenario variation. Changes under the B1 scenario are the least and the A1B and A2 results are similar. Although the changes are generally monotonic increasing, they are not statistically different from natural variability for multiple decades, in contrast to air temperature, and this depends on the drought variable, magnitude of change, natural variability and statistical confidence. In contrast, changes in the means of hydrologic variables, including soil moisture, are essentially undetectable within the 21st century, implying that changes in extremes may be more detectable than changes in mean quantities.

Mots clés : drought / soil water / computer simulation / simulation models / historical records / projections / future


Texte intégral

Page publiée le 27 mars 2008, mise à jour le 3 juin 2022