Bayesian SST Products: Format of the new files: Binary random access files with no extra bytes between records. All records are ascii or 1-byte integers, so that byte-swapping should not be an issue. The new files have 3 sections: Header record: Includes record length and other basic metadata SST records: SST values for the entire region - scaled to 1 byte Pclear records Bayesian cloud probability estimate, log-scaled to 1 byte In addition to the new files we are also generating files in the old file format, which includes scaled SST values only, with no metadata and no cloud probability estimates. These old format files are generating by applying a mask to the SST records cut off at a specified Cloud Probability threshold. File names Hourly products - derived from three half-hourly satellite images sst1b_2006_152_17.Z Hourly SST product / Bayesian format for day152 at 17-Z sst1o_2006_152_17 Hourly SST product, old format, screened for Pcloud < 2%. sst1bE_2006_152_17 Product based on GOES-EAST only sst1bW_2006_152_17 sst1oE_2006_152_17 sst1oW_2006_152_17 3-Hourly products - derived from 3 hourly products centered at nominal epoch sst3b_2006_152_17.Z sst3bE_2006_152_17 sst3bW_2006_152_17 sst3o_2006_152_17 sst3bE_2006_152_17 sst3oW_2006_152_17 Daily products - derived from up to 24 hourly products on the specified day-of-year sst24b_2006_152_17 sst24bE_2006_152_17 sst24bW_2006_152_17 sst24o_2006_152_17 sst24bE_2006_152_17 sst24oW_2006_152_17 hourly Matchups - match_yyyy_ddd.hhB match_yyyy_ddd.hhS The SST results are, as before, represented on a rectangular lat/lon grid The Format of the header record: ASCII record consisting ver Version ID - 8 characters - may be blank ncol Record length in bytes - also, number of pixels in one line nhrec Number of header records (always 1, for now) nrow Number of lines in grid x0 longitude, in degrees, of NW corner of grid (center of pixel) dx longitude stepsize, in degrees dy the N-S cell size in degrees. y0 latitude, in degrees, of NW corner of grid (center of pixel) iy Year of nominal data epoch id day-of-year of nominal data epoch ih UTC Hour of day of nominal data epoch im minute of nominal data epoch ( currently 0 ) iavh Averaging time, in hours SST information SST records begin following the header record(s), i.e. nhrec*ncol bytes past the start of the file. The SST is in nrows records of ncols each. For pixel IX in row IY, the pixel location is given by Longitude = x0 - dx * (IX-1) ; Latitude = y0 - dy * (IY - 1) SST values are represented as unsigned one-byte integers. They are scaled as in the old GOES-SST product. The SST value corresponding to a given count values is: T(K) = 270.0 + 0.15 * Count Count values of 0-6 represent flags, not SSTs. 0 = Space; 2=Land; 1,3,5,6 = pre-bayesian gross screening flags. Baysian cloud screen information: Pclear records begin after the SST data, i.e. (nhrec+nrow)*ncol bytes past the start of the file. Values represent the Bayesian estimate of the probability that the pixel is clear. Values are on a quasi-logarithmic scale to offer more differentiation at low cloud probabilities. Count thresholds for various cloud probabilities: Bayesian Cloud contamination probability estimate, Counts threshhold 0.01 0.1 1.0 2.0 5 10 20 50 252 237 181 157 122 95 67 30