Document Type



The paper uses two years (1997–1999) of data from the North American Land Data Assimilation System at National Centers for Environmental Prediction to analyze the variability of physical variables contributing to the hydrological cycle over the conterminous United States. The five hydrological variables considered in this study are precipitation, top layer soil moisture (0–10 cm), total soil moisture (0–200 cm), runoff, and potential evaporation. There are two specific analyses carried out in this paper. In the first case the principal components of the hydrological cycle are examined with respect to the loadings of the individual variables. This helps to ascertain the contribution of physical variables to the hydrological process in decreasing order of process importance. The results from this part of the study had revealed that both in annual and seasonal timescales the first two principal components account for 70–80% of the variance and that precipitation dominated the first principal component, the most dominant mode of spatial variability. It was followed by the potential evaporation as the secondmost dominant process controlling the spatial variability of the hydrologic cycle over the continental United States. In the second case each hydrological variable was examined individually to determine the temporal evolution of its spatial variability. The results showed the presence of heterogeneity in the spatial variability of hydrologic variables and the way these patterns of variance change with time. It has also been found that the temporal evolution of the spatial patterns did not resemble white noise; the time series of the scores of the principal components showed proper cyclicity at seasonal to annual timescales. The northwestern and the southeastern parts of the United States had been found to have contributed significantly toward the overall variability of potential evaporation and soil moisture over the United States. This helps in determining the spatial patterns expected from hydrological variability. More importantly, in the case of modeling as well as designing observing systems, these studies will lead to the creation of efficient and accurate land surface measurement and parameterization schemes.