On September 27th, Capt. William R. Ryerson, from Davis-Monthan AFB, talked about potential research projects between the USAF and the department. He presented three topics which will be discussed below.

The first project idea was related to how the Air Force forecasts weather events. They currently use what Capt. Ryerson called ‘regime forecasting’. In this method, forecasters would classify the general weather patterns over the Western United States into one of fifteen regimes. Based upon the regime, forecasters could make subjective changes to numerical models based upon previous experience. This has two major disadvantages: it’s subjective and overly simplistic. He suggested using a ’stratified climatology’ technique instead. In this method, the climatology is segmented based upon an objective index. An example he gave was to use El Nino. There would them be three climates of the Western United States, during the ENSO positive years, ENSO negative years, and ENSO neutral years. This seems like a good application for an artificial neural network (ANN). ANNs are able to process a set of complex variables and output something potentially useful. The downside is that no real understanding is made into why the ANN did what it did. ANNs have previously been used in atmospheric research, for instance Ramirez et.al. [2005] used a ANN to construct a nonlinear mapping between output data from a regional ETA model and surface rainfall.

The second project idea was related to increasing the accuracy of wind gust forecasts. According the Capt. Ryerson, the wind gusts in the Southwester United States tend to be underestimated. He talked about using the T1 and T2 indicies from the skew T to perhaps increase accuracy. I don’t know what those are. There was also some discussion about the difficulties in predicting microbusts due to the differences in wet, dry, and hybrid microbusts. There was a recent attempt to “describes the procedure and methodology to formulate the convective weather potential” [Yerong, 2007]. This study was limited to the Pearl River Delta region of South China, but a similar technique could be applied to the Southwest US.

The final project idea was about optimizing operational models. In particular, he is interested in improving and verifying upper level clear air turbulence forecasts. The Air Force uses a derivative of the Ellrod-Knapp index [Ellrod and Knapp, 1992] to forecast turbulence. The problem is that there is little verification of the forecasts. When pilots enter a region of turbulence, they typically will fly out of it either by climbing, decending, or changing heading. Also, they do not keep an archive of either their forecasts or the verification through pilot reports. He talked about using the 0-hour model output (I don’t recall which model) to use as “ground truth” for the verification process. This seems odd since the 0-hour output will be highly dependent upon the initial conditions, but if the initial conditions were known, there would be no need to use the model output in the verification process.

References:

Ellrod, G.P., and D.I. Knapp, 1992: An Objective Clear-Air Turbulence Forecasting Technique: Verification and Operational Use. Wea. Forecasting, 7, 150?165.

Ramirez, M.C.V., H.F. de Campos Velho and N.J. Ferreira, 2005, Artificial neural network technique for rainfall forecasting applied to the Sao Paulo region, Journal of Hydrology, 301, 146-162.

Yerong, F., W. Ying, P. Taoyong, and Y. Jinghua, 2007, An algorithm on convective weather potential in the early rainy season over the Pearl River Delta in China, Advances in Atmospheric Sciences, 24, 101-110. DOI 10.1007/s00376-007-0101-2

(No Ratings Yet)

 Loading ...

Related Posts:

Trackback URI |