Leaf Area Index Specification for Use in Mesoscale Weather Prediction Systems

Abstract

Abstract The energy budget at the surface is strongly influenced by the presence of vegetation, which alters the partitioning of thermal energy between sensible and latent heat fluxes. Despite its relevance, numerical weather prediction (NWP) systems often use only two parameters to describe the vegetation cover: the fractional area of vegetation occupying a given pixel and the leaf area index (LAI). In this study, the Consortium for Small-Scale Modelling (COSMO) limited-area forecast model is used to investigate the sensitivity of regional predictions to LAI assumptions over the Italian peninsula. Three different approaches are compared: a space- and time-invariant LAI dataset, a LAI specification based on Coordination of Information on the Environment (CORINE) land classes, and a Moderate Resolution Imaging Spectroradiometer (MODIS) satellite-retrieved dataset. The three approaches resolve increasingly higher moments both in time and space of LAI probability density functions. Forecast scores employing the three datasets can therefore be used to assess the required degree of accuracy needed for this parameter. The MODIS dataset is the only one able to capture the expected vegetative cycle that is typical of the Mediterranean ecosystem and noticeably improves the 850-hPa temperature and humidity forecast scores up to +72 h forecast time. This suggests that accounting for LAI temporal and spatial variability could potentially improve the prevision of lower-level variables. Nevertheless, model biases of 2-m screen temperatures are not substantially reduced by the more detailed LAI specification when comparisons to synoptic observing stations are performed. Using long-term measurements collected by the CarboEurope project, a detailed verification of sensible and latent heat flux predictions is also presented. It shows that the desirable positive impact arising from a better LAI specification is nullified by the large uncertainties in the initialization of the soil moisture, which remains a crucial parameter for the reduction of screen-level biases.