Effect of volcanic eruptions on the hydrological cycle

Abstract

Large explosive volcanic eruptions inject SO2 into the stratosphere where it is oxidised to sulphate aerosols which reflect sunlight. This causes a reduction in global temperature and precipitation lasting a few years. Here the robust features of this precipitation response are investigated, using superposed epoch analysis that combines results from multiple eruptions. The precipitation response is first analysed using the climate model HadCM3 compared to two gauge based land precipitation datasets. The analysis is then extended to a large suite of state-of-the art climate models participating in the Coupled Model Intercomparison Project Phase 5 (CMIP5). This is the first multi-model study focusing on the precipitation response to volcanoes. The large ensemble allows analysis of a short satellite based dataset which includes ocean coverage. Finally the response of major world rivers to eruptions is examined using historical records. Whilst previous studies focus on the response of just a few rivers or global discharge to single eruptions, here the response of 50 major world rivers is averaged across multiple eruptions. Results are applicable in predicting the precipitation response to future eruptions and to geoengineering schemes that seek to counteract global warming through reducing incoming solar radiation.

The main model-simulated features of the precipitation response include a significant global drying over both land and ocean, which is dominated by the wet tropical regions, whilst the dry tropical ocean regions get significantly wetter following eruptions. Monsoon rainfall decreases, whilst in response to individual eruptions the Intertropical Convergence Zone shifts away from the hemisphere with the greater concentration of volcanic aerosols. The ocean precipitation response is longer lived than that over land and correlates with near surface air temperature, whilst the land response correlates with aerosol optical depth and a reduction in land-ocean temperature gradient.

Many of these modelled features are also seen in observational data, including the decrease in global mean and wet tropical regions precipitation over land and the increase of precipitation over dry tropical ocean regions, all of which are significant in the boreal cold season. The land precipitation response features were robust to choice of dataset. Removing the influence of the El Nino Southern Oscillation (ENSO) reduces the magnitude of the volcanic response, as several recent eruptions coincided with El Nino events. However, results generally remain significant after subtraction of ENSO, at least in the cold season.

Over ocean, observed results only match model expectations in the cold season, whilst data are noisy in the warm season. Results are too noisy in both seasons to confirm whether a long ocean precipitation response occurs. Spatial patterns of precipitation response agree well between observational datasets, including a decrease in precipitation over most monsoon regions. A positive North Atlantic Oscillation-like precipitation response can be seen in all datasets in boreal winter, but this is not captured by the models. A detection analysis is performed that builds on previous detection studies by focusing specifically on the influence of volcanoes. The influence of volcanism on precipitation is detectable using all three observational datasets in boreal winter, including for the first time in a dataset with ocean coverage, and marginally detectable in summer. However, the models underestimate the size of the winter response, with the discrepancy originating in the wet tropics.

Finally, the number of major rivers that undergo a significant change in discharge following eruptions is slightly higher than expected by chance, including decreased flow in the Amazon, Congo, Nile, Orange, Ob and Yenisey. This proportion increases when only large or less humanly influenced basins are considered. Results are clearer when neighbouring basins are combined that undergo the same sign of CMIP5 simulated precipitation response. In this way a significant reduction in flow is detected for northern South American, central African and less robustly for high-latitude Asian rivers, along with a significant increase for southern South American and SW North American rivers, as expected from the model simulated precipitation response.