Documentation for the Deep Convection Tracking Database from the International Satellite Cloud Climatology Project 1. Overview The International Satellite Cloud Climatology Project (ISCCP) has produced a global dataset (currently covering the period 1983-2001) describing the properties of clouds (Schiffer and Rossow 1983, Rossow and Schiffer 1999). This dataset can be used more effectively for the study of specific cloud processes or other related meteorological phenomena if the information about particular types of clouds can be separated from these general, global statistics. To support studies related to "deep" convection, the primary process by which the tropical atmosphere is heated by precipitation and radiation, the ISCCP dataset has been analyzed to identify and describe the properties of mesoscale deep convective cloud systems. The first step in the analysis (see Section 4 for more details) is to identify all spatially adjacent clusters of satellite image pixels (in the ISCCP DX dataset, Rossow et al. 1996) that contain clouds in the upper troposphere, determined by cloud top temperatures < 245 K. Such clusters are referred to as Convective Systems (CS), whether or not they actually contain convective clouds. A further test identifies Convective Clusters (CC) as adjacent cloudy pixels with cloud top temperatures < 220 K. The location, date and time (at nominal 3-hr intervals GMT) of each cluster is collected together with a statistical summary of the CS and CC cloud properties, based on the ISCCP determinations of cloud top temperature and optical thickness. In addition, several properties describing the shape and size of the CS are calculated. The second step of the analysis (see Section 4 for more details), which produces the Convection Tracking (CT) Database, is to "track" each CS over time to form time-associated "families". Since the satellite images used for the ISCCP analysis are separated by 3-hr time intervals and spatially sampled at intervals of about 30 km, the evolution of smaller CS is not followed, so the CT Database only contains families formed from CS that are at least 90 km in radius (about 30 pixels in the ISCCP DX dataset). Tracking is done by examining the next image in the time sequence near the location of each CS in the previous image to determine whether a CS with appropriate characteristics is present that can be identified as the continuation of the previous CS. In addition to the size cut-off, two other parameters determine whether or not a family has terminated because no acceptable CS is available in the next image: the time between images and the percent spatial overlap between the two CS at adjacent times. Normally the time between images is 3 hr, but occasionally images are missing. However, the larger CS usually last for longer time periods, so a theoretical time limit is calculated as a function of CS size, based on the climatology of Machado et al. (1998). In the case of missing images, a family continues if the image separation time (hours) is less than the theoretical limit for this size CS. Overlap is determined by construction of latitude-longitude windows around the last CS member and a candidate at the next time step. This rectangle is formed by the maximum and minimum possible latitudes and longitudes for both CS. The number of common pixels in these two rectangles are counted and overlap is calculated as the ratio of the number of common pixels to the number of pixels in the larger of the two CS. This procedure is repeated for all available candidates and the one with the largest overlap is selected as the next member of the family. Two other tests are performed and the results included to provide an assessment of the quality of the tracking analysis. The first test calculates a theoretical overlap of two CS based on their eccentricity, maximum convergence/divergence and maximum propagation speed. If the actual percent overlap is less than the theoretical value, the overlap flag gives the theoretical value. If the actual overlap is greater than the theoretical value, the overlap flag is set to 99. The second test determines the rate of CS size convergence/divergence by the ratio of two CS sizes; a threshold of 0.75 is set. A flag value of 1 indicates a decrease in CS size at a rate larger than the threshold rate; a flag of 2 indicates an increase of CS size at a rate larger than the threshold rate. A flag value of 99 indicates that the rate of change of CS size is smaller than the threshold rate. If either the overlap or convergence/divergence flag values are not 99, the quality of the tracking analysis is less good. The CT Database contains the morphological and cloud property information for each member of all the families of CS identified by the analysis, even if CC never occurs. The CT Database currently covers the period January 1998 to August 2001 but will soon be extended to cover the whole ISCCP time period (1983-2001) and through at least 2006. 2. Software Included are a FORTRAN90 program and a makefile for reading and searching the CT Database. to compile: make Ctread to run: Ctread year month day lat lon time window_flag degree sample output: Included are sample output files named by the location of the sample family. where: year: 4 digit year month: 2 digit month day: 2 digit day lat: latitude to search -- decimal degrees (-90 to +90) lon: longitude to search -- decimal degrees (0 to 360 east) (Greenwich = 0) time: nearest hour (0 - 24) window_flag: 0 = search range of latitude-longitude determined by minimum-maximum values given by CS center position and radius 1 = search latitude-longitude range given by degree in all directions from CS center degree: range of search window in latitude-longitude The software searches all geostationary satellite positions where a match is possible, determined by a range (+/-) 60 degrees latitude-longitude from the nominal sub-satellite point (SSP). A match is declared if the specified latitude-longitude are within the CS radius (or the specified search window) for the specified date and time. The time range is implicitly (+/-) 1.5 hours. If a match is possible, the CT files for the specified and previous month are searched for a match. For every match found, the program returns a header followed by the complete information for the entire convective tracking family containing the matched CS. Each family is uniquely identified for each month by a 3 letter geographical satellite position code and a family number. The header summarizes the search result with the following information for the matching CS: satellite position lifetime of family in hours age of matched CS in hours (precision is 3 hr) year month day hour CS radius in km CS center latitude (-90 to +90 degrees) CS center longitude (0 to 360 degrees east) 3. Database File Format Each file is a formatted ASCII file, containing one year of data, grouped by month, and by each of five satellite positions (when data are available for that position). A Table of Contents file lists all available files and the nominal sub-satellite latitude-longitudes for each position (SSP). Each position is defined by the following range of SSP longitudes (degrees east): GOE: 248 310 (USA, GOES-EAST) GOW: 225 262 (USA, GOES-WEST) GMS: 130 160 (JAPAN) MET: 0 (EUROPE, METEOSAT) INS: 63 94 (ASIA) Each line of the CT Database contains the following quantities (-99 means undefined) for each CS in the family containing the matched CS: Satellite position (GOE GOW GMS MET INS) Family number (unique by year-month and satellite position) Largest CS radius in this family (km) Minimum CS Cloud Temperature in this family (K) Number of DX pixels with Cloud Temperature < 200 K (largest CS) **Lifetime of this family as number of CS it contains **Position of this CS within family as number of image steps Year Month Day Hour CS radius (km) CS center latitude (-90 to +90 degrees) CS center longitude (0 to 360 degrees east) CS inclination of semi-major axis with respect to north CS eccentricity CS squared correlation of linear regression CS convective fraction (ratio of CC area to CS area) Number of CC in this CS Radius of largest CC (km) Mean radius of all CC Largest CC center latitude (-90 to +90 degrees) Largest CC center longitude (0 to 360 degrees east) Gradient of Cloud Temperature (K/km) CS mean Cloud Temperature (K) CS minimum Cloud Temperature (K) Largest CC Cloud Temperature (K) Standard Deviation of CS Cloud Temperature (K) Largest CC Horizontal Wind Direction with respect to north Largest CC Horizontal Wind Speed (m/s) CS Horizontal Wind Direction with respect to north CS Horizontal Wind Speed (m/s) Percent non-overlap between this and last CS Percent overlap between this and last CS Overlap flag (99, percent theoretical overlap if larger than actual) Convergence/divergence flag (1, 2 or 99) ISCCP Internal Parameter Land/Water flag (majority of CS pixels) CS Mean Cloud Optical Thickness (ice) CS Maximum Cloud Optical Thickness (ice) Minimum Possible Latitude this CS (degrees) Maximum Possible Latitude this CS (degrees) Minimum Possible Longitude this CS (degrees) Maximum Possible Longitude this CS (degrees) 44 - 47 ISCCP Internal Variables **The read software changes these two quantities into family lifetime in hours and age of this CS in hours as determined from the dates/times of each CS in the family. 4. Analysis Methodology and Parameter Definitions The basic ISCCP Analysis of the sampled weather satellite radiances, measured at infrared (approximately 11 microns) and visible (approximately 0.6 microns) wavelengths (Schiffer and Rossow 1985), determines whether a particular image pixel is clear or cloudy and, if cloudy, determines the cloud top temperature (K) and visible optical thickness (daytime only) by comparing the measured radiances to the output of a radiative transfer model (assuming the pixel is completely cloud covered) to produce the ISCCP DX dataset (Rossow et al. 1996, Rossow and Schiffer 1999). The cold-topped clouds included in the CT Database are all assumed to be composed entirely of ice polycrystals with an effective radius of 30 microns. The Clustering Analysis is similar to that devised by Machado (cf. Machado and Rossow 1993 and references therein) and identifies groups (clusters) of spatially adjacent cold-topped cloudy pixels in each satellite image. Two sets of clusters are identified, those composed of cloudy pixels with cloud top temperatures < 245 K, called Convective Systems (CS), and those composed of cloudy pixels with cloud top temperatures < 220 K, called Convective Clusters (CC). The CS represent all high-level cloud systems, including isolated cirrus, and convective systems including a characteristic mix of cloud types: deep convective towers, stratiform anvil clouds and cirrus. The CS is described by its location, size, shape and the properties of its clouds, including whether there are embedded CC, their location and properties. The size of CS and CC is reported as the radius (km) of a circle with the same area as covered by the image pixels composing the CS and CC: R = (N x A/PI)**1/2, where N is the number of pixels and A is the area of pixel (assumed to be 900 km**2). The shape of the CS is described by the eccentricity and inclination of the semi-major axis (with respect to north) of a best-fit ellipse with the same area and same latitude-longitude range covered by the actual CS. The squared correlation coefficient indicates the regression of the variations of latitude and longitude as another CS shape indicator. The CS cloud properties are summarized by seven quantities: the mean, standard deviation and minimum cloud top temperatures, the spatial gradient of the cloud top temperatures (determined by the maximum and minimum temperature and their separation distance, K/km), the number of image pixels with cloud top temperatures < 200 K (indicating convection penetrating into the stratosphere) and the mean and maximum cloud optical thickness. The CC properties (if any are present) are summarized by the number and mean radius of the CC embedded in the CS, the location, radius and cloud top temperature of the largest CC and the fraction of CS area covered by CC. The Tracking Analysis, essentially the same as developed by Machado et al. (1998), associates individual CS over multiple images (minimum of two, equivalent to 3 hr lifetime) into a family, assigning a unique family number by year-month and geostationary satellite position. The basic result is the lifetime of the CS, indicated by the number of CS composing it, but reported in hours. Additional family statistics reported are the largest CS radius and the minimum cloud top temperature in the family. The motion (speed in m/s and direction with respect to north) of the CS and of the largest embedded CC are also reported at each time step, together with the percent overlap and non-overlap (which do not necessarily add to one because of the way that overlap is estimated) and the overlap and convergence/divergence quality flags that indicate whether the CS evolution fits climatological behavior. The CT Database Search program returns the age (in hours) of the individual CS that most closely matches the input space-time coordinates, a summary of the particular CS and its family information, together with the complete family information. 5. References Machado, L.A.T., and W.B. Rossow, 1993: Structural characteristics and radiative properties of tropical cloud clusters. Mon. Wea. Rev, 121, 3234-3260. Machado, L.A.T., W.B. Rossow, R.L. Guedes, and A.W. Walker, 1998: Life cycle variations of mesoscale convective systems over the Americas. Mon. Wea. Rev., 126, 1630-1654. Rossow, W.B., A.W. Walker, D. Beuschel and M. Roiter, 1996: International Satellite Cloud Climatology Project (ISCCP) Description of New Cloud Datasets. WMO/TD - No. 737, World Climate Research Programme (ICSU and WMO), Geneva, February 1996, pp. 115. Rossow, W.B., and R.A. Schiffer, 1999: Advances in understanding clouds from ISCCP. Bull. Amer. Meteor. Soc., 80, 2261-2287. Schiffer, R.A., and W.B. Rossow, 1983: The International Satellite Cloud Climatology Project (ISCCP): The first project of the World Climate Research Programme. Bull. Amer. Meteor. Soc., 64, 779-784. Schiffer, R.A., and W.B. Rossow, 1985: ISCCP global radiance data set: A new resource for climate research. Bull. Amer. Meteor. Soc., 66, 1498-1505. 6. Contacts for Information William B. Rossow (wrossow@giss.nasa.gov) Cindy Pearl (cpearl@giss.nasa.gov)