ISSN: 2277-3754 ISO 9001:2008 Certified International Journal of Engineering and Innovative Technology (IJEIT) Volume 3, Issue 3, September 2013 364 Abstract — The accurate and timely identification of cyclogenesis has been one of the major issues in cyclone prediction. Due to the lack of in situ observations, model simulations, which are generally different from real conditions, are used to address this challenging problem. Since model derived parameters perform poorly in predicting cyclogenesis, we developed a neural network approach to identify cyclogenesis using the data obtained from Global Positioning System Radio Occultation (GPSRO) technique. For this purpose, we used 1-dimensional variational analysis temperature and pressure profiles and atmospheric stability and moisture indices, derived from this temperature and pressure profiles using GPSRO Constellation Observing System for Meteorology Ionosphere and Climate (COSMIC) mission over North Indian Ocean as predictors for cyclogenesis. Out of the 8686 COSMIC tangent points, during May 2006 - December 2010, 118 profiles are selected after applying thresholds to different parameters at different levels. Using these 118 profiles, a neural network model is developed to predict the cyclogenesis and non-cyclogenesis locations. Out of the 118 cases, the model could accurately predict 112 non-cyclogenesis and 5 cyclogenesis points. Though the model failed to predict one cyclogenesis case, there are no false alarms. Index Terms—Artificial Neural Network, Atmospheric stability, Cyclogenesis prediction, One-dimensional variational analysis, Radio occultation observations. I. INTRODUCTION Cyclones have a life cycle of genesis to dissipation. Besides sudden intensity changes of the tropical cyclones, identification of cyclogenesis locations has been a difficult problem. So far, tropical cyclogenesis predictions are obtained from numerical model simulations using low resolution atmospheric [1], [2], [3] and atmosphere ocean coupled models [4], [5]. Since model simulations could be significantly different from the actual observations, forecasting skills using output from the models for cyclogenesis predictions are usually poor [6]. Earlier studies suggested different precursor factors for cyclogensis formation: Palmen et al. [7] show that sea surface temperature (SST) of above 26°C is necessary for tropical cyclone formation. In a study of western Pacific typhoons, Riehl et al. [8] concluded that storm formation can be inhibited by the vertical shear of the horizontal wind. Gray [9] defines a tropical cyclone genesis condition as the product of thermodynamic and dynamical terms where, SST and middle level moisture affects are included as the thermodynamic terms and vertical shear and low level vorticity affects as the dynamical terms. Climatologically values of this product are well correlated with the global tropical cyclone formation regions [10]. Finally, Gray [9], [11] considered six potential precursor environmental conditions favorable for cyclogenesis. Hennon et al. [12] suggested using in-situ observations for cyclogenesis prediction. Till the recent years this suggestion could not be implemented due to the limitations of the in situ radiosonde observations, particularly, over the oceans. However, Global Positioning System (GPS) Radio Occultation (RO), henceforth GPSRO, technique has been providing accurate and high resolution vertical profiles of temperature, humidity and pressure over a larger spatial and temporal scales [13], [14]. Presently operating GPSRO Constellation Observing System for Meteorology Ionosphere and Climate (COSMIC) mission (operating since May 2006) can track 90% of all rising and setting occultation soundings to within one kilometer of the earth‟s surface. The accuracies of COSMIC observations are similar to those of radiosonde with global mean temperature bias of -0.09K and a standard deviation (SD) of 1.72K [15] with the added advantage of availability over the oceans, where cylogenesis generally occurs. This mission provided a good amount of 1-dimensional variational (1-DVAR) temperature and pressure profiles from which a few precursor parameters required for cyclogenesis prediction could be estimated. Studies show that cyclone track and intensity forecasts have improved by incorporating GPSRO data into the model [16], [17]. However, these data have not been used so far for the prediction of cyclone formation. Thus for the first time we propose a methodology to predict cyclogenesis over the North Indian Ocean using COSMIC observations. II. DATA We used 1-DVAR temperature and pressure profiles from GPSRO COSMIC mission (http://cosmic-io.cosmic.ucar.edu) in the North Indian Ocean (5°N-25°N, 60°E-100°E) from May 2006 to December 2010 (Figure 1). The cyclogonesis locations are taken from best track data of Joint Typhoon Warning Center (JTWC) (http://www.usno.navy.mil/NOOC/nmfc-ph/RSS/jtwc/best tracks/ioindex.html). An Approach to Predict Cyclogenesis using Radio Occultation Observations Neerja Sharma, M. M. Ali National Remote Sensing Centre, Indian Space Research Organization (ISRO), Hyderabad 500037 (India)