Inferring Changes in Terrestrial Water Storage Using ERA-40 Reanalysis Data: The Mississippi River Basin

Abstract

Terrestrial water storage is an essential part of the hydrological cycle, encompassing crucial elements of the climate system, such as soil moisture, groundwater, snow, and land ice. On a regional scale, it is however not a readily measured variable and observations of its individual components are scarce. This study investigates the feasability of estimating monthly terrestrial water-storage variations from water-balance computations, using the following three variables: water vapor flux convergence, atmospheric water vapor content, and river runoff. The two first variables are available with high resolution and good accuracy in the present reanalysis datasets, and river runoff is commonly measured in most parts of the world. The applicability of this approach is tested in a 10-yr (1987?96) case study for the Mississippi River basin. Data used include European Centre for Medium- Range Weather Forecasts 40-yr reanalysis (ERA-40) data (water vapor flux and atmospheric water vapor content) and runoff observations from the United States Geological Survey. Results are presented for the whole Mississippi River basin and its subbasins, and for a smaller domain covering Illinois, where direct measurements of the main components of the terrestrial water storage (soil moisture, groundwater level, and snow cover) are available. The water-balance estimates of monthly terrestrial water-storage variations show excellent agreement with observations taken over Illinois. The mean seasonal cycle, as well as interannual variations, are captured with notable accuracy. Despite this excellent agreement, it is not straightforward to integrate the computed variations over longer time periods, because there are small systematic biases in the monthly changes. These biases likely result from inaccuracies of the atmospheric assimilation system used to estimate the atmospheric water vapor convergence and can be corrected in part with the application of a simple detrending procedure. It is noteworthy that the critical domain size for water-balance computations, using high-resolution reanalysis data such as ERA-40, appears to be much smaller than for raw radiosonde data. The Illinois domain has a size of only ?2 ? 105 km2 and is shown to be suitable for the computation of the water-balance estimates. A comparison for other regions would be needed in order to confirm this result.