Novel multivariate quantile mapping methods for ensemble post-processing of medium-range forecasts

Statistical post-processing is an indispensable tool for providing accurate weather forecasts and early warnings extremes. Most statistical univariate, with dependencies introduced via use of empirical copula. Standard methods take a dependence template from either the raw ensemble output (ensemble copula coupling, ECC) or observations (Schaake Shuffle, SSh). There are drawbacks to both methods. In ECC it assumed that simulates well, which not always case (e.g. 2-meter temperature in The Netherlands). Schaake Shuffle able capture flow dependent changes choice key. Here we compare reshuffled standard model statistics (EMOS) approach two multivariate bias adjustment approaches have been used before context: 1) correction N-dimensional probability density function transform (MBCn) 2) ranks defined optimal assignment (OA). These advantage they explicitly structure present observations. We apply ECC, MBCn OA 2-m dew point at seven stations Netherlands. Forecasts verified univariate methods, using heat index derived variables, wet-bulb globe (WBGT). Our results demonstrate spatial inter-variable more realistic compared Shuffle. variogram score shows while most skilful first days, moderate lead times longest than MBCn. Overall, highlight importance considering between variables locations forecasts.