Handout (4.3 MB)
To understand this relationship, I employ the High-Resolution Rapid Refresh (HRRR) model’s Vertically Integrated Smoke (VIS) data and compare it to direct-normal solar radiation measurements from the Atmospheric Radiation Measurement user facility’s Southern Great Plains Multifilter Rotating Shadowband Radiometer (MFRSR) network. The 13 MFRSR devices in Kansas and Oklahoma track the sun across the sky and measure direct-beam radiation data, which predominately powers photovoltaic production. Using data from January 1, 2021, to July 31, 2023, I utilize the HRRR-VIS hourly analysis to identify days where smoke was present over each station. I can then calculate the total reduction from expected maximum radiation using MFRSR measurements and a standard solar model.
Two interesting features have presented themselves. First, there is a critical threshold where smoke values greater than 150 mg per square meter have a significant impact on the reduction of solar radiation. Smoke values below this threshold can lead to a wide range of solar reduction values, anywhere from 0% to 30%. However, once the smoke load goes above 150 mg per square meter, the reduction values reach 40% to 50%. Second, there seems to be a spatial and temporal trend in the data, wherein the geographic locations of the fire as well as time of year alter the properties of the smoke and thus create a range of reductions in solar radiation. I aim to continue investigating how the origin of wildfires, and when they take place, impacts the smoke and solar radiation reduction. I am also eager to fit a regression with the data based on these factors, with potential applications in the use of artificial intelligence for solar forecasting. With the significant smoke loads experienced from the 2023 Canadian wildfires in large Northeastern metropolitan areas, and historic wildfires taking place in California over the past decade, power companies and photovoltaic energy producers need a better way to forecast how this smoke will impact their operations. This research aims to give further insight into the many complex variables that must be considered to accurately forecast photovoltaic power generation.

