Assessing Forecast Accuracy during Stormy Weather Conditions

Accurate and timely forecasts are essential to effectively respond to extreme weather such as storm surges, tropical cyclones, heavy rains, gusty winds, and alike. Global climate change has been driving stronger and frequent weather extremes. So it is vital to improve weather prediction giving utility industry more time to prepare for extreme weather events.

Let’s take a look at how Mine-Weather methodically deals with specific forecast accuracy during dangerous storms, minimizing infrastructural damages and industrial downtime. Mine-Weather extracts the calendar period of storm weather events of interest for in-depth analysis. Storm periods are defined in Mine-Weather in two different ways, one is time based and other weather based threshold.

In Time based definition, Mine-Weather system considers the actual calendar time period. Time period may include the start and end time, say prior to a storm hitting the utility service area, or after the storm is over. Users can setup storm definitions and run different time period scenarios to get a proper idea of the various factors involved in constituting a storm. Storms can be analyzed individually or by bundling several storms into one composite view.

For Weather-based threshold definition, conditions are defined and threshold value is set based on the weather variables. Users can define four weather variables in Mine-Weather they are steady wind speed, wind gust, temperature and rainfall. User can define a lower bound and upper bound value for each individual event. Any weather variable that exceed this threshold value is considered an extreme weather event. While defining extreme weather conditions in Mine-Weather, user has the option to apply condition to the actual weather data, forecast data or both.

Mine-Weather employs a two-step procedure to designate an extreme weather event. First, the user selects a series of options like storm type, event details, lower and upper bounds, calendar time periods, and alike. Then the user can apply ‘both’ or ‘either’ weather conditions based on which the system runs through the data to extract time periods that meet the selection criteria. System returns a list of the time periods that meet the select criteria, which can be further edited by the user. User can delete certain periods, add a set number of hours, link short periods, and finally commit the edits, which is then processed to create summary tables for dashboard analysis. This process can be repeated for any number of extreme weather events.

Now the system starts collecting the actual and forecast data for each of the defined time periods associated with the extreme weather event. In case of observational data the system only selects weather variables for a particular period of time. For Forecast data it is a complex process, where the system pulls forecast data that overlaps with the storm period. It is noted that prolonged periods of forecasts prior to beginning of a storm event is the most relevant forecast data for storm analysis. Mine-Weather isolates all such forecast data collected for a particular time period and analyses to see the forecast accuracy during a storm condition.

Numerous possible approaches are available to determine the exact period of extreme weather. Download our WhitePaper to read the first series of Mine-Weather in determining accuracy of forecast data during a storm event.

Download Whitepaper

About the Author

Dr. Ronald Mueller

Ron is CEO and Founder of Macrosoft, Inc.. Ron heads up all company strategic activities, and directs day-to-day work of the Leadership Team at Macrosoft. Ron is also Macrosoft’s Chief Scientist, defining and structuring Macrosoft’s path forward on new technologies and products, such as Cloud; Big Data; and AI. Ron has a Ph.D. in Theoretical Physics from New York University, and worked in physics for over a decade at Yale University, The Fusion Energy Institute in Princeton, NJ, and at Argonne National Laboratory. Ron also worked at Bell Laboratories in Murray Hill, NJ., where he managed a group on Big Data, including very early work on neural networks. Ron has a career-long passion in ultra-large-scale data processing and analysis including: predictive analytics; data mining, machine learning and neural networks.

Leave a Reply

Your email address will not be published. Required fields are marked *