Representing local-scale temperature patterns in complex terrain: performance of high-resolution datasets

Gridded near-surface air temperature datasets are essential for environmental and climate applications, providing spatially continuous information beyond point measurements. In mountain regions, however, accurately representing temperature is particularly challenging. Strong spatial variability, frequent departures from simple elevation-based gradients, and cold-air pooling driven by nocturnal cooling and drainage flows lead to complex temperature patterns that are generally underrepresented when interpolating temperature observations from sparse weather stations. These limitations can reduce the accuracy in capturing extreme conditions, such as hot spells in the valley bottoms and urban areas or cold spells and strong thermal inversions. High-resolution dynamical models offer a complementary, physically based perspective by explicitly resolving terrain and atmospheric processes, improving representation of temperature gradients, diurnal cycles, and local circulations. Yet, near-surface temperatures in complex terrain remain sensitive to model resolution and surface-atmosphere coupling. The distinct strengths and limitations of these approaches raise the question of how different methods perform in representing local temperature patterns in complex terrain. In this study, we compare a 1-km dataset of daily near-surface air temperature produced through an interpolation scheme with high-resolution fields from dynamical modelling to assess the abilities to represent temperature variability in a complex mountainous terrain like the one of the Adige River catchment in Eastern Italian Alps. The interpolation method estimates the vertical temperature structure through a daily fitted, non-linear temperature-elevation profile based on more than 600 station observations at multiple altitudes and accounts for topographic complexity (Frei, 2014). Model-based products include the km-scale reanalysis VHR-REA_IT (Raffa et al., 2022) obtained by a dynamical downscaling of ERA5 for Italy at approximately 2-km resolution and the Copernicus European Regional ReAnalysis (CERRA). The comparison is conducted over the period 1990-2020 and focuses on the representation of temperature extremes and their spatial variability, e.g., cold-air pooling and heatwaves, and on the description of daily vertical profiles. Interpolated fields capture local extremes and cold-air pools where observations are available but are limited in resolving broader spatial variability and vertical thermal structure. In contrast, high-resolution reanalyses provide a more physically consistent depiction of thermal gradients, although systematic differences in describing extremes emerge. Our results will illustrate how the complementarity of approaches can guide the appropriate use and integration of temperature products in mountainous regions to support temperature-related hazard monitoring and risk assessment.