Like most computer forecasts, CIP/FIP may benefit from the ingestion of newer, higher quality data available today. One particular source has gained our interest. The National Weather Service introduced an ice accumulation forecast product in recent years. This product lies within their National Digital Forecast Database (NDFD) and is available to the general public. Since this forecast exclusively predicts the ground accumulation of freezing precipitation, it is unclear if it may be useful in a forecast for in-flight conditions. The goal of this research project is to investigate how effective the accumulation forecast has been since its debut and see how it may be used in model predictions like CIP and FIP.
The first task of this project is to analyze the structure of the archived forecasts. The icing accumulation forecasts are saved in grib2 files. The forecasts themselves are rather easy to read with some simple code, but complications arise when attempting to view multiple forecasts leading up to a single event.
The process began with choosing appropriate event dates to start downloading and organizing data. Dates were picked from early 2014 based on observed surface icing conditions across the U.S. The NDFD icing data is stored as day 1-3 forecasts and day 4-7 forecasts. A forecast is output each hour throughout the day, which is then saved as a file for archiving. Each hour's file consists of periods six hours long going out to the third or seventh day. The icing accumulation is given as a six-hour accumulation ending at the end of the current period:
“… An ice accumulation grid will be specified whenever at least a trace of ice accumulation is forecast for any hour during a valid period. Valid periods for the NDFD begin and end at 0600, 1200, 1800, and 0000 UTC.” (NWS 3)
To make things easier, only the 1-3 day forecasts were used. Once an event time is picked from a selection of files going back three days, a forecast v. time graph would capture the accumulation prediction variability leading up to the event.
The graphic will show how the forecast performed as the event approached. It is important to compare the final results, but it may be just as important to see how much variability was in the forecast as time progressed. Though the forecast involves model predictions, the final output is done by a forecaster at each NWS office. The icing potential forecast is also dependent on the output of other products:
“If a WFO is not producing a Snow Amount grid for zero amounts, a WFO does not have to produce an Ice Accumulation grid for zero amounts. However, all WFOs must create ice accumulation grids when at least a trace of ice is in their forecast.” (NWS 3)
The second task of this project will be to compare the forecast icing accumulation amounts to the actual observed amounts in the region. Data will be acquired from multiple weather stations across an affected area. If a majority of the forecast accumulation is within a reasonable amount to the observed, then that event will be considered a positive hit. The problem here is determining a proper way to compare a 6-hour accumulation forecast to one minute observed data. Once the forecast can be compared correctly for an event, the process comes down to collecting enough events for meaningful analysis.
“National Digital Forecast Database (NDFD) Ice Accumulation Grids NWS Product Description Document (PDD) October 12, 2011.” http://www.nws.noaa.gov/infoservicechanges/ndfd_iceaccum.pdf, 2011, web, Sept 2015.