Weather prediction strategies
This guide lays out the strategies and research employed to find arbitrage in climate markets. Most are in active research currently.
Probabilistic hyper-local forecasting with machine learning
A custom machine learning model trained on spatially co-located weather sensors. The publicly available weather sensors near ASOS stations don't exhibit rounding errors. Large amounts of data from around the ASOS station will produce a much higher accuracy forecast, then current forecast models that measure with a grid of .25 spatial degrees. By focusing more closely on the immediate vicinity region of the weather station better probability in relation to current market odds would be found.
Probalistic automated trading and arbitrage spotting
Weather forecasting and machine learning cannot predict the future. Forecast accuracy is not just a measure of precision, but of probabilities. This would potentially yield higher returns over time. While the weather is unpredictable, it is at the same time, extremely predictable. Also newer ai/ml forecast models are acheiving unparalled accuracy with physics level estimation of the atmosphere on vast quantities of publicly available data. Leveraging these methods for better forecasting specific to a weather station is the highest priority research objective.
Use historical OMO (one minute observations) for estimating realtime data feed values
Due to limitations in real time data available from ASOS-HFM feeds, values are inaccurate. Using historical one minute observations produced from these stations, a more accurate interpretation of real time data could be made. OMO data is only available made available to the public after 1-2 days, not in real time.
FOIA requests and resource development for OMO feeds
ASOS stations transmit data a number of different ways. There is a OMO feed available for one weather station, https://www.weather.gov/psr/HiResASOS . This is specific to this forecast office providing access to HiRes Real Time Data from the airports weather station. This is not widely available. But there are several different public feeds that data is transmitted along for NWS and the FAA. More research will need to be conducted to see if any of them can yield Hi-Resolution data for real time access to more accurate data from ASOS stations.