Non-parametric kernel-based estimation and simulation of precipitation amount

The probability distribution of precipitation amount strongly depends on geography, climate zone, and time scale considered. Closed-form parametric distributions are not sufficiently flexible to provide accurate universal models for over different scales. In this paper we derive non-parametric estimates the cumulative function (CDF) wet periods. CDF obtained by integrating kernel density estimator leading semi-explicit expressions functions. We investigate an adaptive plug-in bandwidth (KCDE), using both synthetic data sets reanalysis from Mediterranean island Crete (Greece). show that KCDE provides better than standard empirical (staircase) estimate kernel-based use normal reference bandwidth. also demonstrate enables simulation means inverse transform sampling method.