Data Mining Atmospheric/Oceanic Parameters in the Design of a Long-Range Nephelometric Forecast Tool

The Department of Defense calls for long-range forecasts to aid in the planning of operations. The goal of this research was to explore the feasibility of predicting, one month in advance, the total monthly cloud cover over the country of Afghanistan. In an attempt to reach this goal, the following objectives were achieved: 1) climatological synoptic study of Afghanistan; 2) survey of Real Time Nephanalysis, outgoing longwave radiation (OLR), and surface observational data; 3) examination of teleconnection indices and sea surface temperatures; 4) standard statistical analysis for prediction; and 5) classification tree analysis (CART). In addition, due to current world events, CART analysis was also applied over the country of Iraq (see Appendix C). Data were examined using standard statistical regression techniques, including linear and multiple linear regression, and then CART analysis was used for exploring possible concealed predictive structures. Standard statistics showed a strong negative correlation between monthly average OLR and surface observational total cloud cover from the fall through spring months. However, linear regression revealed very weak relationships between the predictor and predictand variables. As well, CART results contained misclassification rates that exceeded established thresholds for operational use. Further studies using CART for atmospheric science applications should be pursued.