Stochastic generation of meteorological variables and effects on global models of water and carbon cycles in vegetation and soils
Global models of water and carbon cycles in continental vegetation and soils are usually forced with monthly mean climatic data-sets and thus neglect day to day variations of the weather. This treatment may be justified for empirical models based on parametrizations validated at a monthly timescale. Mechanistic models handling hydrological and biological processes at much shorter timescales might, however, be largely affected by such an approximation, since the various processes described are highly nonlinear. A random generator of daily precipitations and temperatures applicable at the global scale has thus been developed from worldwide meteorological data covering 6 years of observations. The probability of a wet day is correlated to the weather encountered the previous day. The amount of precipitation, the daily mean temperature and the diurnal range of temperature are described from the statistical point of view by the cumulative distribution functions (CDF) of three random variables. The CDFs relative to temperatures are different for rainy and dry days. This stochastically generated weather field is used as input to IBM (Improved Bucket Model) and CARAIB (CARbon Assimilation In the Biosphere), two global models of respectively soil hydrology and vegetation productivity. Large differences in both the geographical distribution and the global value of soil water, vegetation productivity and carbon stocks are obtained between the model runs using monthly uniform weather on one side and randomly generated weather on the other. The main contribution to this difference at the global scale arises from the precipitation generation occurring as a result of high degree of nonlinearity of the interception scheme used in IBM.