Further development of adaptive parameter tuning in the numerical weather prediction models of Deutscher Wetterdienst

Abstract Numerical weather prediction (NWP) models have a large number of tunable parameters that need to be optimized in order to obtain the best possible forecast quality. This may be accomplished manually, which is usually referred to as expert tuning, or in a semi‐automated way using so‐called meta‐models that support the parameter‐tuning process. At the German Weather Service (Deutscher Wetterdienst, DWD), an alternative methodology has been developed since about 2020, which was introduced as adaptive parameter tuning (APT) in a preceding publication. APT uses information extracted from data assimilation (DA) to optimize model tuning parameters or physical quantities derived from external parameter data, focusing on parameters affecting the near‐surface fields of 10‐m wind speed, 2‐m temperature, and 2‐m relative humidity. This article describes the further development of the APT performed since the first publication, mainly comprising additional predictor fields and additional model quantities modified by APT. The predictor fields are generally calculated as time‐filtered DA increments of temperature, humidity, and wind speed at the lowest model level, serving as proxies for model biases in the respective variables. The existing predictor fields for daily averages and the diurnal temperature amplitude are complemented by daytime‐only averages for temperature and humidity with weights peaking in late afternoon, reflecting specific diurnal‐cycle biases that typically reach their maximum around sunset. Based on these predictors, additional APT components for the hydraulic diffusivity of the soil, bare‐soil evaporation resistance, and laminar transfer resistance are introduced. Moreover, the soil moisture analysis (SMA), which is part of the DA system, is complemented by APT‐based post‐processing to improve the forecast scores in situations where the SMA does not perform well. Besides that, APT components for the albedo of land and sea ice have been added.