RETRIEVAL OF ATMOSPHERIC MOISTURE PROFILES FROM DMSP DATA WITH A NEURAL NETWORK

THE PROPOSED STUDY WILL DEMONSTRATE THE FEASIBILITY OF USING A NEURAL NETWORK TO RETRIEVE ATMOSPHERIC MOISTURE, ONE OF THE ENVIRONMENTAL PARAMETERS DISCUSSED IN THE 1986 JOINT CHIEFS OF STAFF MEMORANDUM MJCS 154-86. PSI HAS ALREADY SHOWN THAT NEURAL NETWORKS ARE CAPABLE OF RETRIEVING TEMPERATURE PROFILES WITH HIGH ACCURACY FROM SATELLITE DATA, AND WORK BY OTHERS HAS SHOWN THAT CLOUD PARAMETERS CAN SIMILARLY BE OBTAINED USING NEURAL NETWORKS. BECAUSE NEURAL NETWORKS ARE INHERENTLY PATTERN PROCESSORS, THEY ARE EXCELLENT CANDIDATES TO PROCESS THE OUTPUTS OF SATELLITE-BORNE SOUNDERS. A NETWORK IMPLEMENTED ON A DESKTOP COMPUTER WILL BE TRAINED USING SIMULATED DMSP SOUNDING DATA GENERATED FROM RADIOSONDE MOISTURE AND TEMPERATURE MEASUREMENTS MADE OVER LAND AND OCEAN, IN TROPICAL THROUGH SUBARCTIC REGIONS, AND IN DIFFERENT SEASONS. IT WILL THEREAFTER BE ABLE TO QUICKLY SYNTHESIZE MOISTURE PROFILES FROM NEW SOUNDING DATA, EVEN IF THE DATA WERE COLLECTED UNDER CONDITIONS NOT CLOSELY MATCHING THOSE REPRESENTED IN THE TRAINING SET. PHASE I OF THE PROJECT WILL ESTABLISH THE FEASIBILITY AND ACCURACY OF THE METHOD. THE ULTIMATE GOAL OF LATER PHASES OF THE RESEARCH IS TO COMBINE INPUTS FROM SEVERAL TYPES OF SENSORS TO RETRIEVE, WITHOUT CUING, THREE-DIMENSIONAL MOISTURE AND TEMPERATURE PROFILES AND OTHER ATMOSPHERIC PARAMETERS SUCH AS CLOUD TYPE AND FRACTION, SURFACE EMISSIVITY, AND SURFACE TEMPERATURES IN REAL-TIME OVER LARGE GEOGRAPHIC AREAS USING ONLY A DESKTOP COMPUTER AND DATA OBTAINED BY DIRECT SATELLITE DOWN-LINK. SINCE THE METHOD IS INHERENTLY NONLINEAR AND REQUIRES ONLY A DESKTOP COMPUTER, IT SHOULD BE FASTER, LESS EXPENSIVE, AND MORE ACCURATE THAN CURRENT METHODS. IT CAN POTENTIALLY COMBINE INPUT FROM SEVERAL SENSORS TO RETRIEVE CONTOURS OF ATMOSPHEREIC PARAMETERS OVER LARGE GEOGRAPHIC AREAS IN REAL-TIME. SUCH SYSTEMS WILL ALLOW PRIVATE WEATHER, AGRICULTURE, AVIATION, AND NEWS ORGANIZATIONS TO USE SATELLITE DATA IN REAL-TIME FOR LOCAL WEATHER FORECASTING.