CMWS Model History

The CMWS (Copiague Maplecourt Weather Station) Model started development in August of 2008. The initial release was on January 17, 2009. It was intended to automate the analysis of numerical computer data outputs from the North American Model (NAM) and Global Forecasting System (GFS). These are the two main operational weather computer models supported by the National Oceanic Atmospheric Administration (NOAA).

The most challenging aspect of this project was the extraction of the numerical outputs from the models. Rather than re-inventing source code, we searched the World Wide Web (WWW) and came across a website with the following address, http://www.wxcaster.com/models_text.htm.

The website had the specifications needed to go forward. With this stumbling block out of the way, we needed a way to store the extracted data. The decision was made to use PostgreSQL and MySQL as the Database Management System (DBMS) for persistent data management. A flexible and scalable database schema was formulated to encompass both the NAM and GFS model data extraction outputs. Latest Database schema

The next step was to manipulate the data to provide the best analysis. Most weather forecasters use model data to guide their forecasts. However some use computer models exclusively by either using an average of the models or a ratio of the models. They also look at the biases of a particular computer model and the range of the models. Short range models include the NAM and medium to long range include the GFS. The best solution (highest forecast confidence) is when two or more computer model outputs are the same or similar enough.

The CMWS Model automates this analysis. The method used is statistical forecasting. Statistical forecasting uses the numerical outputs from the GFS and NAM computer models to predict the future weather conditions by identifying trends, patterns and propriety logic within the data to develop a forecast. This method uses mathematical formulas to identify the patterns and trends while testing the result for mathematical reasonableness and confidence. Latest Model Run

The CMWS Model outputs verification analysis graphs, which are plots of selected weather parameters such as temperature, dew point, mean surface pressure, wind speed and wind direction. This provides users with a visual analysis of the CMWS Model performance. Latest Model Verification

Release 1.0 Uses the average of the NAM and GFS models, 50/50 ratio.

Release 1.5 Uses a weighted average of the two models, based on forecast hours (lower forecast hours have more weight than higher forecast hours since higher errors are propagated in longer range forecast). Model verification analysis graphs are also provided for 2 meter temperature, dewpoint, mean surface pressure, wind speed and wind direction. These graphs are plots of the observed data and CMWS model data forecast.

Release 2.0 Uses the same method in Release 1.5, but incorporates differences between the observed data to the forecasted data. The average difference is used to normalize the output.

Release 2.1 Uses the same method in Release 2.0, but a logarithmic weighted average of the two models is used.

Release 2.2 Major User Interface Overhaul. One important new feature is to allow users select different locations.

Release 2.2.1 Bug fixes found in release 2.2.

Release 2.3 Added three new output parameters, precipitation type, sky condition (cloud cover) and wind chill temperature. Removed three output parameters, 850mb temperature, max temperature and min temperature.

Release 2.4 Use time-based regression model for model differences.

Release 2.5 Use Relative Humidity to determine precipitation output parameter and Only include the last 7 days of model and observed data.

Release 2.6 Fixed a major security hole.

Release 2.7 New utility to convert zipcodes to the nearest ICAO code. (zipcode2icao)