Evaluation of HRRR Smoke for IDSS and Policy Making in the Northern Great Plains

Wildfire smoke is a common concern for air quality. In the past three years, the Northern Great Plains (NGP) has seen an increasing amount of wildfire smoke being advected into the region from Canada. The longevity and spatial extent of the event means many communities have been affected. The Centers for Disease Control and Prevention (CDC) found a 17% increase in Emergency Department visits related to Asthma from April to August 2023. The increase in emergency department visits was directly correlated to an increase in the Air Quality Index (AQI) categorization for that day. The High-Resolution Rapid Refresh (HRRR) model includes near-surface and vertically integrated smoke products to help forecast the impacts of any given smoke event. Both AQI and HRRR-Smoke measure the values of PM 2.5 and can be directly related. However, few studies exist that explore the verification of smoke forecasts. Forecasts for any weather event are essential to Impact-Based Decision Support Services (IDSS); thus, the quality of smoke forecasts is equally essential for IDSS.

To verify HRRR-Smoke forecasts, prolific smoke events are defined using polar orbiting satellite imagery and regional lidar data (ceilometers) as days where PM 2.5 measurements caused poor air quality (≥ 100) in the County Warning Area (CWA) of the Fargo/Grand Forks (FGF) National Weather Service Forecast Office (NWSFO). PurpleAir air quality sensors are used to verify the HRRR-Smoke performance forecasting smoke events between the summers of 2021-2023. Each event is analyzed using contingency tables to compare forecasts to observed PM 2.5 values attributed to the different AQI categories. Contingency table values are used to calculate the hit rate, Heidke Skill Score (HSS), and bias ratio, which are the forecast verification metrics used in this study. Additionally, ceilometer data will display smoke structure and whether layers are surface-based or elevated. These observations will be qualitatively compared to near-surface and total column smoke in HRRR. Previous case studies, such as the 2018 California Camp Fire, assessed the quality of HRRR-Smoke forecasts for observed PM 2.5 values and found both a time lag and an underestimation of PM 2.5. We hypothesize similar results in our study.

This study will determine the accuracy of the HRRR in predicting future smoke events and identifying any biases that may be present. This study aims to verify HRRR-Smoke forecasts, identify biases in forecast models, and determine vertical smoke composition. Model verification and identified biases will aid operational meteorologists in forecasting future smoke events and improve IDSS management. Upon completion of a quantitative analysis for HRRR-Smoke, a qualitative analysis of IDSS and existing air quality policies for the FGF CWA will be conducted. The qualitative analysis will involve a survey to assess AQI thresholds for IDSS and explore current state/local policies for smoke events.