Using Integrated EOF Analysis for Evaluation of WRF Simulations in Urban Environments

Abstract Evaluation methods for regional climate models (RCMs) commonly rely on point comparisons with observed meteorological fields, which provide limited understanding of the spatial and temporal representations of important factors affecting urban areas in models. These factors are not only complex but also difficult to differentiate, which complicates their analysis. This study thus develops an innovative approach using the empirical orthogonal function (EOF) analysis to compare urban heat island and precipitation patterns in RCM simulations with those from observations, taking advantage of the capacity of the method for data disaggregation. The method was tested on summer daily maximum and minimum temperature ( T max and T min ) and precipitation P in the Chicago Metro Area (CMA). Using observed data, the EOF analysis on temperature consistently produced coherent patterns that reflect known impacts of urban environments on climate and weather. EOF evaluation of corresponding 4-km WRF simulations against observations confirmed a strong warm bias (∼3°C) for simulated T min in the urban area, as observed in point comparisons against stations; further analysis, however, suggested that the shape and time behavior of the urban pattern were well represented. The EOF analysis on T max , which showed no problems in the point comparison, revealed important differences in shape (urban area of influence on temperatures) and time [principal component (PC) correlation of −0.5] for the urban pattern between datasets, suggesting the need for model improvements. Results showed no systematic urban effects on summer P for the CMA for observations or simulations, but the analysis of winter patterns suggested a possible urban enhancement on P over the city. Significance Statement Evaluating, diagnosing, and refining regional-scale climate models over urban areas is an essential part of improving the projection of urban weather and climate impacts. The primary objective of our project is to develop better techniques for evaluating and diagnosing regional climate model simulations over urban areas. Most traditional approaches rely on one-to-one comparisons between simulated and observed values using meteorological stations or remote sensing data, providing limited information about the ability of the model to capture underlying physical processes that ultimately affect urban weather and climate. Using the empirical orthogonal function analysis to disaggregate daily maximum and minimum temperature, and precipitation data into spatial patterns, we provide an innovative way to evaluate urban effects simulated by regional climate models and compare them against urban patterns from gridded station observations.