Downscaling Hourly Air Temperature of WRF Simulations Over Complex Topography: A Case Study of Chongli District in Hebei Province, China

Abstract

The air temperature is one of the most frequently measured meteorological parameters. Air temperature forecasting is a crucial climatic factor-related task required in many different applications in areas such as agriculture, industry, energy, the environment, and tourism (Abdel-Aal, 2004;Cifuentes et al., 2020). Certain applications include short-term load forecasting for power utilities (Li et al., 2016), protection against freezing injury of various fruits (Chung et al., 2006), adaptive temperature control in greenhouses (Dombaycı & Gölcü, 2009), prediction of cooling and energy consumption in residential buildings (Ben-Nakhi & Mahmoud, 2004), and establishment of a planning horizon for infrastructure upgrades, insurance, energy policy, and business development purposes (Smith et al., 2007). Therefore, accurate forecasting of the air temperature with high spatiotemporal resolution has always been an important goal for meteorologists and weather forecasters.Mountainous areas also represent important areas of human activities, such as fruit cultivation, tourism, and skiing in the winter season. Therefore, hourly air temperature forecasting and high-resolution prediction in mountainous areas is necessary and important for human activities. In contrast to plain areas, the topography of a mountainous area is very complex, which can cause the air temperature to vary within small terrain units. For example, in a 1-km grid, the difference in altitude between the slope top and foot can reach as much as 500 m, and the temperature difference as high as 3°C, assuming that the slope of the grid is 0.5. It is a difficult task for air temperature forecasting and modeling in complex geographical zones because the temperature is influenced by several nonlinear processes, such as the interaction between large-scale air mass circulation and local airflow, airflow-topography interactions, and the interplay between radiation and topographic shading.Currently, many numerical weather prediction (NWP) models are available for air temperature prediction, for example, the European Centre for Medium-Range Weather Forecasts (ECMWF) model, the fifth-generation mesoscale model (MM5), and the Weather Research and Forecasting (WRF) model. All these NWP models allow day-ahead air temperature prediction, which is usually adopted in practice. However, uncertainties in model

A new statistical downscaling approach for short‐term forecasting of summer air temperatures through a fusion of deep learning and spatial interpolation

“…The use of spatial interpolation provided spatial information for areas where no weather stations existed and enabled the development of a U‐Net‐based statistical downscaling model at the local scale. In addition, previous studies have only evaluated the forecast accuracy of statistical downscaling models at trained observation stations (Yi et al ., 2018; Zhang et al ., 2022). This study performed a systematic model evaluation by dividing the weather stations into seen and unseen stations, which made it possible to evaluate the spatial generalization of the bias‐correction ability of the downscaling model across the study area.…”

“…In addition, dynamic downscaling induces errors due to imperfect parametrization (Caldwell et al ., 2009). In contrast to dynamic downscaling, statistical downscaling uses the statistical relationship between predictors (e.g., NWP model forecast outputs and topographic information) and in‐situ observations, resulting in fast computation and easy application (Lee & Singh, 2018; Zhang et al ., 2022). Statistical downscaling has been reported to perform comparably to dynamic downscaling (Eden & Widmann, 2014; Monjo et al ., 2016).…”

PISCOeo_pm, a reference evapotranspiration gridded database based on FAO Penman-Monteith in Peru

“…These gridded data were at a resolution of ~10 km (0.1°), and to be used in the construction of PISCOeo_pm, the spatial scale was reduced using two covariables: LST and DEM (Tables 1 and 2 ). Reducing the spatial scale of PISCOt was inspired by other works in which the geographically weighted regression (GWR) 70 , 71 technique was applied 72 – 74 . The methodology used is divided into three steps: Spatial downscaling considering the normals values (1981–2010).…”

Vegetation Classification and a Biomass Inversion Model for Wildfires in Chongli Based on Remote Sensing Data

“…The study area is encompassed by mountains, with undulating terrain in the east, middle, and west, and three large ditches traversing the entire region. The total area of the study area is approximately 2300 km 2 , with elevations ranging from 812 m to 2169 m [32]. Geomorphologically, the region can be divided into two types: tectonic denudation plateau areas and eroded mountain areas.…”