The aim of this paper is to present the latest version of the model SWAP (Soil Water-Atmosphere-Plants), describing the interactions between the land surface and the atmosphere and being oriented on the coupling with atmospheric models. The distinctive features of SWAP are as follows: (i) it is a physically based model especially with respect to a description of the processes of the winter-spring period which are usually described very schematically; (ii) in solving the systems of equations we tried to use analytical methods contrary to the usual practice of the application of numerical ones, that allowed us to avoid many problems associated with solving the numerical equations (such as instability, great consumption of the computer resources and the calculational time and so on); (iii) a relatively small number (18) of model parameters, most of which can be obtained from literature; (iv) when calculating the partition of non-intercepted rainfall into infiltration and surface runoff, the spatial variability of hydraulic conductivity at saturation is taken into account. Relatively simple mathematical formalism and application of the analytical approach make the model to be compact and sufficiently rational.
Previous version of SWAP was validated during participation in PILPS (Project for Intercomparison of Land surface Parameterization Schemes). Since then the model has been advanced in its representation of the evaporation from a bare soil, formation of snow cover and drainage, as well as the calculation of some soil and snow parameters. New version was validated using the data of agrometeorological station Petrinka (the Kursk region, Russia). The station is situated in the forest-steppe zone with limited plants water supply. Using the Petrinka data set we had an opportunity to test SWAP for different types of the land surface: bare soil, grass and crops and on the long term - from 5 to 13 years.
The stand-alone simulations were carried out for fallow (for 5 years), grassland (for 13 years) and agricultural field (11 years) with a 1-day time step. The values of simulated water storage in 1-m soil layer were compared with the corresponding values measured at the Petrinka. The root mean square error of the calculations is equal to 32, 25 and 37 mm for fallow, grassland and crop area, respectively. The coefficient of correlation between the calculated and measured data is equal to 0.75 for fallow, 0.90 for grass and 0.81 for crops. The obtained results show that SWAP reproduce annual and interannual variation of soil water storage under water stress conditions fairly well.
It is difficult to determine the extent to which the discrepancies between the calculated and observed values depend on quality and completeness of input variables, validation data and parameters or on scheme-specific errors. For this reason it is necessary to continue the validation of the scheme under a wide spectrum of natural conditions. This will require great efforts directed towards collecting the appropriate data of high quality.