TOGA-COARE Upper Air: Non-NCAR Vaisala High-Resolution Radiosonde Data 1.0 General Information The Non-National Center for Atmospheric Research (NCAR) Vaisala High Resolution Radiosonde Data is one of the upper air data sets collected by the University Corporation for Atmospheric Research/Joint Office for Science Support (UCAR/JOSS) as part of the Tropical Ocean Global Atmosphere (TOGA) Coupled Ocean-Atmosphere Response Experiment (COARE) project. Included in this data set are the high vertical resolution radiosonde data from the non-NCAR land-based sites in Darwin, Funafuti, Gove, Honiara, Hong Kong, Kanton, Kota Kinabalu, Kuching, Laoag, Legaspi, Mactan, Madang, Misima, Santa Cruz, Singapore, Tarawa, and Thursday Island. Also included are the high resolution radiosonde data from the following non-NCAR ship-based sites: R/V Hakuho Maru, R/V Keifu Maru, R/V Natsushima, and R/V Vickers. The temporal coverage varies from site to site but all data fall within the Enhanced Monitoring Period (EMP) of 1 July 1992 to 30 June 1993 and most soundings were within the Intensive Observing Period (IOP) of 1 November 1992 to 28 February 1993. When a site was in operation soundings were typically available twice per day during the EMP and four times per day (or more) during the IOP. These soundings have been corrected for the humidity measurement errors of the Vaisala RS80 radiosonde (Wang et al 2002). 2.0 Data Contact Junhong (June) Wang (junhong@ucar.edu) Katy Beierle (kbeierle@atd.ucar.edu) 3.0 Data and Format Information 3.1 NCAR Sounding Sites Site Name Longitude Latitude Elevation Time Period --------------------------------------------------------------------------- Darwin 130.88 -12.42 30.0 1 Nov 1992 - 28 Feb 1993 Funafuti 179.22 -8.52 2.0 1 Nov 1992 - 26 Feb 1993 Gove 136.49 -12.17 53.0 1 Nov 1992 - 28 Feb 1993 Honiara 159.97 -9.42 56.0 1 Nov 1992 - 26 Feb 1993 Hong Kong 114.17 22.32 66.0 1 Nov 1992 - 28 Feb 1993 Kanton -171.72 -2.77 6.0 31 Jan 1993 - 25 Feb 1993 Kota Kinabalu 116.05 5.93 2.0 2 Nov 1992 - 26 Feb 1993 Kuching 110.33 1.48 21.0 2 Nov 1992 - 27 Feb 1993 Laoag 120.53 18.18 5.0 1 Nov 1992 - 28 Feb 1993 Legaspi 123.73 13.13 19.0 1 Nov 1992 - 28 Feb 1993 Mactan 123.97 10.30 24.0 3 Nov 1992 - 13 Jan 1993 Madang 145.80 -5.22 5.0 1 Nov 1992 - 28 Feb 1993 Misima 152.83 -10.70 7.0 1 Nov 1992 - 28 Feb 1993 Santa Cruz 165.80 -10.70 24.0 12 Nov 1992 - 23 Feb 1993 Singapore 103.88 1.33 14.0 8 Nov 1992 - 27 Feb 1993 Tarawa 172.92 1.35 4.0 12 Nov 1992 - 27 Jun 1993 Thursday Island 142.22 -10.58 61.0 1 Nov 1992 - 25 Feb 1993 R/V Hakuho Maru Variable Variable 8.0 1 Nov 1992 - 4 Dec 1992 R/V Keifu Maru Variable Variable 3.0 3 Nov 1992 - 15 Nov 1992 R/V Natsushima Variable Variable 6.0 2 Feb 1993 - 18 Feb 1993 R/V Vickers Variable Variable 2.0 28 Jan 1993 - 27 Feb 1993 3.2 File Naming Conventions There are two types of data files included in these data sets. 1) High resolution radiosonde data files: These files are named as follows: zmddhhmm.sit where: z signifies that the data include the humidity corrections of Wang et al (2002). m signifies the month 1-9 = January - September a-c = October - December dd is the two digit day hh is the two digit hour (UTC) mm is the two digit minute sit is the release site: dar = Darwin fun = Funafuti gov = Gove hak = R/V Hakuho Maru hon = Honiara hnk = Hong Kong kan = Kanton kei = R/V Keifu Maru kot = Kota Kinabalu kuc = Kuching lao = Laoag leg = Legaspi mac = Mactan mad = Madang mis = Misima nat = R/V Natsushima san = Santa Cruz sin = Singapore tar = Tarawa thu = Thursday Island vic = R/V Vickers 2) Pibal (wind only) data files: These files are named as follows: sitmddhh.cls where sit is the release site (same as above) m is the month 1-9 = January - September a-c = October - December dd is the two digit nominal day hh is the two digit nominal hour (UTC) cls signifies that these data are in class format 3.2 Format Description 3.2.1 CLASS Files (named like sitmddhh.cls) 3.2.1.1 Header records The header records (15 total records) contain data type, project ID, site ID, site location, actual release time, sonde type, met processor, winds type, surface meteorological observation source, and comments provided by the opeator. Line Label Contents 1 Data Type: Project ID 2 Project ID: Data description 3 Release Site Type/Site ID: Site, Country, and WMO ID 4 Release Location (lon,lat,alt): Location of release (lon and lat first given in deg min then in decimal degrees. Altitude in meters. 5 UTC Release Time (y,m,d,h,m,s): Actual UTC release time. 6-11 Typically blank 12 Nominal Release Time (y,m,d,h,m,s):Nominal UTC release time. 13 Data Column headings (parameters) 14 Data Column headings (units) 15 ---- 3.2.1.2 Data records The data records each contain time from release, pressure, temperature, dew point, relative humidity, U and V wind components, wind speed and direction, ascent rate, balloon position data, altitude, and quality control information. Each data line contains 21 fields, separated by spaces. The data are right-justified within the fields. All fields have one decimal place of precision, with the exception of latitude and longitude, which have three decimal places of precision. The contents and sizes of the 21 fields that appear in each data record are as follows: Field Format Missing No. Width Parameter Units Value ----------------------------------------------------------------------- 1 6 F6.1 Time Seconds 9999.0 2 6 F6.1 Pressure Millibars 9999.0 3 5 F5.1 Dry-bulb Temperature Degrees C 999.0 4 5 F5.1 Dew Point Temperature Degrees C 999.0 5 5 F5.1 Relative Humidity Percent 999.0 6 6 F6.1 U Wind Component Meters/Second 9999.0 7 6 F6.1 V Wind Component Meters/Second 9999.0 8 5 F5.1 Wind Speed Meters/Second 999.0 9 5 F5.1 Wind Direction Degrees 999.0 10 5 F5.1 Ascension Rate Meters/Second 999.0 11 8 F8.3 Longitude Degrees 9999.0 12 7 F7.3 Latitude Degrees 999.0 13 5 F5.1 Range Kilometers 999.0 14 5 F5.1 Azimuth Degrees 999.0 15 7 F7.1 Altitude Meters 99999.0 16 4 F4.1 QC flag for Pressure Code (see below) 99.0 17 4 F4.1 QC flag for Temperature Code (see below) 99.0 18 4 F4.1 QC flag for Humidity Code (see below) 99.0 19 4 F4.1 QC flag for U Component Code (see below) 99.0 20 4 F4.1 QC flag for V Component Code (see below) 99.0 21 4 F4.1 QC flag for Horizontal Wind Code (see below) 99.0 ---------------------------------------------------------------------- Fields 16 through 21 contain the Quality Parameters provided by UCAR/JOSS. The JOSS QC flags are as follows: Code Description ---------------------------------------------------------------------- 99.0 Unchecked (QC information is `missing.') (`UNCHECKED') 1.0 Checked, datum seems physically reasonable. (`GOOD') 2.0 Checked, datum seems questionable on physical basis. (`MAYBE') 3.0 Checked, datum seems to be in error. (`BAD') 4.0 Checked, datum is interpolated. (`ESTIMATED') 9.0 Checked, datum was missing in original file. (`MISSING') ---------------------------------------------------------------------- 3.2.2 Non-CLASS Files (named like zmddhhmm.sit) 3.2.2.1 Header Records The header records (15 total records) contain data type, project ID, site ID, site location, actual release time, sonde type, met processor, winds type, surface meteorological observation source, and comments provided by the opeator. Line Label Contents 1 Data Type: Includes information on the RH corrections for that sonde (can include RH bias, RH htg, age of sonde) 2 Project ID: Site and data type. 3 Launch Site Type/Site ID: Site, Country, and WMO ID. 4 Launch Location (lon,lat,alt): Location of release (lon and lat first given in deg min then in decimal degrees. Altitude in meters. 5 GMT Launch Time (y,m,d,h,m,s): Actual GMT release time. Other header records can include: Caution: Cautionary notes. Sonde Type/Serial Number/Age(yrs): Radiosonde type, and age. GC-corrections P/T/U (hPa/C/%): Ground Correction information SAH correction: Sensor Arm Heating correction information. Nominal Launch Time (y,m,d,h,m,s): Nominal release time. 3.2.2.2 Data records The data records are similar to those in the CLASS files, except there are more spaces between each column and some values do not follow the same format (e.g. lack of decimal points) or have different missing values (especially note that the U and V wind components have varying missing values in these data, sometimes 9999.0, 999.0, 9999, or 999). 3.3.4 Dataset Remarks The Non-CLASS files (named like zmddhhmm.sit) have been passed through correction procedures for humidity measurement errors (the class files were not passed through this correction since they only have wind parameters). The details of this correction are available in Wang et al (2002). These corrections deal with six humidity measurement errors present during TOGA COARE including chemical contamination, temperature-dependence, basic-calibration-model, ground check, sensor aging, and sensor-arm-heating. The following short summary comes from Wang et al (2000). 1. Introduction During TOGA_COARE intensive observing period (from Nov. 1992 to Feb. 1993), a total of 11,172 radiosondes were launched at 42 stations. VIZ radiosonde systems were used at 13 stations, accounting for 3488 soundings. The remaining 7684 soundings used Vaisala RS80 radiosondes, including 4744 RS80H soundings and 2940 RS80A soundings. Sounding systems provided by NCAR's Atmospheric Technology Division (ATD), including Integrated Sounding Systems (ISS) and the Cross-chain Loran Atmospheric Sounding Systems (CLASS), launched 3760 Vaisala RS80 radiosondes (mostly RS80H) at eight stations (referred as NCAR data). The NCAR systems recorded radiosonde data at the surface prior to launch, an important factor in subsequent error detection and correction development. The rest of the Vaisala RS80 soundings (total 3924) were made with existing non-NCAR systems (referred as non-NCAR data). A dry bias has been found in the TOGA_COARE Vaisala radiosonde data (Zipser and Johnson 1998). Using data containing the dry bias introduces substantial errors in derived radiative and thermodynamical parameters (Guichard et al. 2000) and in calculated atmospheric heat and moisture budgets (Johnson and Ciesielski 2000). In addition, the spatial distribution of different types of radiosondes used at 42 sites during TOGA COARE (Vaisala RS80A and RS80H, and VIZ) introduced false spatial variations in large-scale moisture. When the full un-corrected COARE data set is assimilated into a numerical weather prediction model, it can lead to under-prediction of clouds and precipitation (Lorenc et al. 1996). ATD and Vaisala identified the cause of the dry bias as contamination of the polymer used as the dielectric material in the capacitive RS80 humidity sensor (Humicap). ATD collaborated with Vaisala to develop physical models based on laboratory tests to correct the dry bias and improve other aspects of Vaisala humidity measurements. 2. Correction Algorithm and Evaluation The Vaisala RS80A and RS80H use "Humicap" thin-film capacitance sensors whose temperature-compensated capacitance is proportional to the ambient water vapor concentration. The two versions differ primarily in the chemical composition and properties of the sensor dielectric material, and in the accuracy of the algorithm for sensor temperature-dependence used in the data processing. Humidity errors for both RS80A and RS80H include "chemical contamination error", "temperature-dependence error", "sensor-arm-heating error", and others. A series of laboratory tests was conducted in the calibration lab at Vaisala in Finland and used to develop the correction algorithms. The correction algorithms are described in detail in our papers (in preparation) and are only briefly described below. The chemical contamination error is due to the occupation of binding sites in the sensor polymer by non-water molecules emitted from the sonde packaging material, and produces a dry bias. The dry bias due to chemical contamination depends on sensor type (A vs H), the age of the sonde, and RH, and affects measuremennts throughout the troposphere. For various reasons, the RS80A has much smaller contamination error than RS80H. One-year-old RS80H and RS80A sondes have ~4% and ~2% dry-bias errors at 50% RH, respectively. Contamination errors generally increase with age and with RH. Age information was not available for NCAR soundings in COARE, so an independent reference RH measurement at the surface was used instead. That surface reference RH was compared to the radiosonde RH before launch and used in a modified correction algorithm. For non-NCAR soundings, the dry bias was corrected based on known or estimated sonde ages. Based on the work done by ATD and Vaisala, Vaisala changed the desiccant type in the RS80 sonde packages in August 1998 and introduced a new type of protective shield on the sensor boom in May 2000 for RS80 series radiosonde. The desiccant change is expected to reduce the dry bias by 30-50%, and the protective cap is expected to prevent contamination errors completely. The temperature-dependence (TD) error results from an approximation of a linear function of temperature to the actual nonlinear temperature dependence of the sensors, and also introduces a dry bias. The TD error mainly exists at temperatures below -20(C, increases substantially with decreasing temperatures below -30(C, and is much larger for RS80A than RS80H. The TD correction for RS80A has a correction factor of 0.15, 0.75 and 2.3 at -40(C, -60(C and -80(C, respectively. A RS80A_measured RH of 80% at -40(C would increase to 92% after correcting the TD error. The sensor-arm-heating (SAH) error was attributed to radiaitonal heating of the temperature and humidity sensors and sensor arm. The SAH error at the surface was estimated by using the radiosonde-measured temperature before launch and an independent surface reference temperature for NCAR soundings, and by using a statistical approach for non-NCAR 10-second soundings. When applicable, the SAH correction was applied to the first sixty seconds of a sounding based on a 13-second thermal time constant of the sensor arm. The uncorrected NCAR radiosonde data includes the SAH correction, but the correction assumed that the total difference in RH at the surface measured by radiosondes before launch and an independent surface instrument was due to the SAH (Cole and Miller 1995). The difference seen at the surface was actually the sum of the SAH error and the contamination dry bias. In the COARE data set released here, each corrected sounding was examined based on individual Skew-T plot analysis, and obvious glitches were corrected or removed. Various summary plots were generated at each station for both day and night soundings to evaluate the performance of the correction algorithms, including scatter plots of comparisons between near-surface mixing ratios (MR) extrapolated from independent surface instruments and averaged MR in the mixed layer from radiosonde data before and after corrections, vertical profiles of mean and standard deviation of corrections, and histograms of CAPE and RH before and after corrections. In general, all corrected soundings look 'reasonable'. Despite the extraordinary level of attention given to this sounding data set, we recognize that further analysis might reveal additional complexities and problems. We ask that all users who find problems in the corrected data or who have concerns about ATD's treatment of the data contact us. We pledge to continue our efforts to make this particular data set one of the highest quality soundings data sets ever collected. 4.0 References Cole, H., and E. Miller, 1995: A correction for low-level radiosonde temperature and relative humidity measurements. Preprints, Ninth Symposium on Meteorological observation and Instrumentation, Charlotte, North Carolina, 27-31 March 1995, American Meteorological Society, pp. 32-36. Guichard, F., D. Parsons, and E. Miller, 2000: Thermodynamical and radiative impact of the correction of sounding humidity bias in the tropics. J. Climate, 13, 3611-3624. Johnson, R. H., and P. E. Ciesielski, 2000: Rainfall and radiative heating from TOGA COARE atmospheric budgets. J. Atmos. Sci., 57, 1497-1514. Lorenc, A. C., D. Barker, R. S. Bell, B. Macpherson, and A. J. Maycock, 1996: On the use of radiosonde humidity observations in mid-latitude NWP. Meteorol. Atmos., Phys., 60, 3-17. Wang, J., H. L. Cole, D. J. Carlson, E. R. Miller, K. Beierle, A. Paukkunen, and T. K. Laine, 2002: Corrections of humidity measurement errors from the Vaisala RS80 radiosonde -- application to TOGA COARE data. J. Atmos. Ocean. Tech., (in press). Zipser, E. J., and R. H. Johnson, 1998: Systematic errors in radiosonde humidities a global problem? Preprints, Tenth Symposium on Meteorological Observations and Instrumentation, 11-16 January 1998, Phoenix, Arizona, American Meteorological Society, pp. 72-73.