AIR Worldwide Releases Updated Reconstructed Weather Data

AIR Worldwide Corporation (AIR) of Boston has released updated reconstructed weather data for stations throughout the U.S. The update by AIRWeather™, AIR’s weather risk management service, provides weather derivatives traders a consistent and temporally homogenized historical temperature time series for weather stations nationwide.

The update accounts for station shifts and station environmental changes that impact historical weather observations.

“In order for the weather derivatives market to function smoothly, it is imperative to address the issues of historical temperature variations on a regular basis,” said Mark Gibbas, AIRWeather senior research scientist in meteorology. “Evaluating a weather derivative contract using raw data can lead to devastating financial consequences as historical data can be inconsistent.

“For example, sensors at weather stations are often shifted, which can lead to variations in observations,” explained Gibbas. “A weather derivative contract may be written and based on the temperature at a station with 25 years of historical data. If, however, five years ago, that station moved its sensors to a cooler location the result would be five years of cooler observations and 20 years of warmer observations. By using the full 25 years of inconsistent data, the estimate will be biased toward the earlier warmer observations. The historical data will not be representative of estimated future observations, resulting in a flawed derivatives contract. AIRWeather solves this problem by reconstructing the data to represent accurate historical weather observations regardless of station changes.”

AIRWeather used its updated, proprietary data reconstruction process for its latest update. The new process involves sophisticated algorithms with improved design to detect variations in temperature resulting from station shifts and environmental changes since 1950.

AIRWeather’s reconstructed data involves de-shifting and de-trending the cleaned daily minimum and maximum temperature time series to remove data inconsistencies due to factors unrelated to weather. The new reconstructed data adds 15 more weather stations, bringing the total stations monitored by AIRWeather to 235 in the U.S. alone.