High resolution radiosonde data |
1 Format conversion procedures 1.1 Conversion to JOSS CLASS format (all datasets except 08508, 72208, 72402, 74494, KCEJ) ------------------------------------------------------------------------------------------ 1.1.1 The raw parameters, in their native resolution, were kept without change 1.1.2 The ascension rate was calculated for each data point (excluding the surface, where it was given a missing value) based on the altitude and time at two levels, the current level and the previous level. If the data point for which the ascension rate was being calculated had missing time or altitude, the ascension rate was not calculated at that level and was flagged as missing. If the data point prior to that for which the ascension rate was being calculated had a missing time or altitude, then the data point two data points prior to the present one was used, and so on until one with non-missing time and altitude was found. If there are no non-missing previous data points the ascension rate is set to missing. 1.1.3 The U and V wind components were calculated based on the wind speed and direction at the wind data levels 1.1.4 Dew points were calculated via the equations from Bolton (1980) Due to CLASS format constraints, Dew Points less than -99.9 Deg C were given a value of -99.9 Deg C 1.1.5 The latitude and longitude position of the radiosonde was calculated based on (Snyder, 1987). See Snyder page 159 equations 20-14, 20-15, 20-18, and 21-15. 1.2 Lajes(08508) dataset conversion to JOSS CLASS format -------------------------------------------------------- 1.2.1 Meteo-France/CNRM received the high resolution data set from INMG. Due to the lower resolution of these data in the upper levels, Meteo-France/CNRM included data from the TEMP (GTS) messages within the high resolution dataset. 1.2.2 This data set was hand entered at the site (Lajes), and contained many typographical errors. Meteo-France/CNRM and UCAR/JOSS have corrected such errors as have been found. Any remaining typographical errors should be minimal, but caution should still be used. 1.2.3 As the time from launch was not available in the original files, it was estimated with a ascension rate of 5m/s (excluding at the surface, where it was given a missing value). 1.2.4 Meteo-France/CNRM converted the data to a high resolution JOSS CLASS format file, without quality control flags. These JOSS CLASS format soundings were then passed along to UCAR/JOSS for further processing and quality control procedures as described below. 1.2.5 The raw variable vertical resolution temperature, relative humidity, altitude, pressure, dew point, wind direction, and wind speed were kept without change. 1.2.6 The ascension rate was not calculated for this data set due to the lack of time information. 1.2.7 The U and V wind components were calculated based on the wind speed and direction at the level. 1.2.8 Due to CLASS format constraints, Dew Points less than -99.9 Deg C were given a value of -99.9 Deg C and were flagged as being questionable. 1.2.9 The latitude and longitude position of the radiosonde was not calculated due to the lack of time information. 1.3 Charleston(72208) and Chatham(74494) datasets conversion -------------------------------------------------------- 1.3.1 The raw 6 s vertical resolution time, temperature, relative humidity, altitude, and pressure were kept without change. 1.3.2 The ascension rate was calculated for each data point (excluding the surface, where it was given a missing value) based on the altitude and time at two levels, the current 6 s level and the previous 6 s level. If the data point for which the ascension rate was being calculated had missing time or altitude, the ascension rate was not calculated at that level and was flagged as missing. If the data point prior to that for which the ascension rate was being calculated had a missing time or altitude, then the data point two data points prior to the present one was used, and so on until one with non-missing time and altitude was found. If there are no non-missing previous data points the ascension rate is set to missing. 1.3.3 Dew points were calculated via the equations from Bolton (1980). Due to CLASS format constraints, Dew Points less than -99.9 Deg C were given a value of -99.9 Deg C and were flagged as being questionable. 1.3.4 In past derivation of winds using the raw 6-sec resolution elevation and azimuth angle data containing elevation angle oscillations occasionally led to large oscillations in wind velocity, specifically at low elevation angles (Williams 1993). The general approach to correct this problem was to remove the outlier radiosonde position data before computing the wind components. This process required fitting a ninth order polynomial to the azimuth and elevation angle data from 360 seconds to the end of the sounding, then comparing the calculated residuals and observed values, and finally removing the outliers when present. Applying some additional smoothing helped rectify the more extensive problem occurring when low elevation angles were within 10 degrees of the limiting angles (LA). When the elevation angle was between (LA + 7.5) and (LA + 10), the new elevation angle was computed using a 2 min linear fit. When the elevation angle was between (LA + 5) and (LA + 7.5), the new elevation angle was computed using a 3 min linear fit. When the elevation angle was less than (LA + 5), the new elevation angle was calculated employing a 4 min linear fit. No frequency smoothing occurred when the number of low elevation angle observations was greater than 20% of the total number of observations. A Finite Fourier Series analysis performed using the elevation angle's residuals allowed removal of 90-190 second periods and smoothing periods below 30 seconds. Obtaining the u and v wind components entailed fitting a 2 min second order polynomial to the position except for the beginning and end minute (or 1.5 minutes if over 50 mb) which used a 3 min fit. A linear fit was used when there were less than 15% of the total number of points, not including the beginning or end of the flight, on one side of the point under going the wind value calculation. 1.3.5 The latitude and longitude position of the radiosonde was calculated based on (Snyder, 1987). See Snyder page 159 equations 20-14, 20-15, 20-18, and 21-15. 1.4 Wallops(72402) dataset conversion to JOSS CLASS format ------------------------------------------------------ 1.4.1 These soundings arrived at JOSS in three seperate files, one containing ~ 1.2 second vertical resolution PTH data, a second file containing significant level PTH data, and a third file containing one minute vertical resolution wind data. 1.4.2 The significant level files contained the independently measured surface data point. The second data point in the significant level file contains the time after release. The time difference between these two points was saved. Since the 1.2 second data file did not have any way of finding the release point we used the pressure, temperature and RH from the second point (in the significant level file) to search the 1.2 second PTH file for the sonde's release point. This can be done since the data within the significant level files was derived from data within the 1.2 second PTH files. The 1.2 second PTH file was searched for the point matching the second point's (again from the sig level file) pressure, temperature, and relative humidity. When this matching point was found , the 1.2 second file was then searched for the datum closest to the matching point's time minus the time difference saved previously. We then assumed that this matching point was the release point and this became the zero point in the final file. All the 1.2 second vertical resolution PTH data points whose times occurred after this release point were written to the final file. The release point usually occurred well beyond the zero time within the 1.2 second PTH file. In order to make the release point's time zero, its original time was subtracted from the remaining datum's times. This is reflected in the final file's first point having a negative time. This point is the independent surface observation mentioned above and was saved for this reason. 1.4.3 Altitudes for the 1.2 second vertical resolution PTH data were derived using the hypsometric equation using the virtual temperature. Note, many times the launch point found in the 1.2 second vertical resolution PTH file had altitudes unequal but close to 13 meters. However this value was kept and all altitudes were calculated using this surface height. If the dew point was missing, the dry bulb temp was used in place of the virtual temperature. 1.4.4 Wind and 1.2 sec resolution PTH data points having the same times were merged into one point except for the zero time. The zero time wind was assumed to be an independent surface observation and therefore was assigned to the significant level file's surface point, the very first point in the final file. The wind for the zero time within the final file remained missing. Otherwise, the wind data was placed by altitude and its time was set to missing. If the wind and 1.2 sec vertical resolution PTH altitudes were sufficiently close, the points were combined, using the calculated PTH altitude. If a wind altitude was missing, it was placed in the final file by time, with its time then set to missing. 1.4.5 The raw pressure, temperature, relative humidity, wind direction, and wind speed were kept without change. 1.4.6 The ascension rate was calculated for each data point (excluding the release point and the independent surface observation where it was given a missing value) based on the altitude and time at two levels, the current level and the previous level. If the data point for which the ascension rate was being calculated had missing time or altitude, the ascension rate was not calculated at that level and was flagged as missing. If the data point prior to that for which the ascension rate was being calculated had a missing time or altitude, then the data point two data points prior to the present one was used, and so on until one with non-missing time and altitude was found. If there are no non-missing previous data points the ascension rate is set to missing. 1.4.7 Dew points were calculated via the equations from Bolton (1980). Due to CLASS format constraints, Dew Points less than -99.9 Deg C were given a value of -99.9 Deg C and were flagged as being questionable. 1.4.8 The latitude and longitude position of the radiosonde was calculated based on (Snyder, 1987). See Snyder page 159 equations 20-14, 20-15, 20-18, and 21-15. 1.4.9 The missing pressures for the one-minute winds not combined with an exisiting 1.2 second vertical resolution PTH datum were linearly interpolated using the pressures and altitudes from the times immediately before and after the missing pressure's time. The following equation details the procedure: wind pressure = p1 + (p2 - p1)*[(Walt - h1)/(h2 - h1)], under the conditions:h2 > Walt > h1 and p2 < p1, where hn is the height of the nth pressure surface, pn is the nth pressure, and Walt is the altitude of the wind data. If an altitude was missing, then the equation became the following: wind pressure = [p1 + p2]/2. These values were only calculated to make these wind values visible for visual quality control (see below). These values should not be used. 1.5 R/V KNORR dataset conversion to JOSS CLASS format ------------------------------------------------- 1.5.1 FASTEX Experiment ----------------- 1.5.1.1 The raw 10 s vertical resolution time, temperature, relative humidity, altitude, pressure, dew point, wind direction, and wind speed were kept without change. 1.5.1.2 The ascension rate was calculated for each data point (excluding the surface, where it was given a missing value) based on the altitude and time at two levels, the current 10 s level and the previous 10 s level. If the data point for which the ascension rate was being calculated had missing time or altitude, the ascension rate was not calculated at that level and was flagged as missing. If the data point prior to that for which the ascension rate was being calculated had a missing time or altitude, then the data point two data points prior to the present one was used, and so on until one with non-missing time and altitude was found. If there are no non-missing previous data points the ascension rate is set to missing. 1.5.1.4 Due to CLASS format constraints, Dew Points less than -99.9 Deg C were given a value of -99.9 Deg C and were flagged as being questionable. 1.5.1.5 The latitude and longitude position of the radiosonde was calculated based on (Snyder, 1987). See Snyder page 159 equations 20-14, 20-15, 20-18, and 21-15. 1.5.1.6 The quality codes from the source agency were ignored, except in one case. The winds within the first 120 seconds were flagged as questionable as they were in source data. This was due to these winds being interpolated from the surface and the 120 sec data point due to the 240 sec smoothing interval used by the source agency to derive wind values. 1.5.2 LabSea Experiment ----------------- 1.5.2.1 The raw 10 s vertical resolution time, temperature, relative humidity, altitude, pressure, dew point, wind direction, and wind speed were kept without change. 1.5.2.2 The ascension rate was calculated for each data point (excluding the surface, where it was given a missing value) based on the altitude and time at two levels, the current 10 s level and the previous 10 s level. If the data point for which the ascension rate was being calculated had missing time or altitude, the ascension rate was not calculated at that level and was flagged as missing. If the data point prior to that for which the ascension rate was being calculated had a missing time or altitude, then the data point two data points prior to the present one was used, and so on until one with non-missing time and altitude was found. If there are no non-missing previous data points or if the ascension rate value was < 1 m/s or >12m/s the ascension rate is set to the estimated value of 5 m/s. 1.5.2.4 Due to CLASS format constraints, Dew Points less than -99.9 Deg C were given a value of -99.9 Deg C and were flagged as being questionable. 1.5.2.5 The latitude and longitude position of the radiosonde was calculated based on the estimated trajectory of the radiosonde. The trajectory of the radiosonde was calculated based on altitude, ascension rate and wind values at the current level and the prior calculated position. If the wind is missing at the current level, the first previous wind value is used. If there is no non-missing previous data points, the estimated trajectory of the radiosonde is the vertical above the launching point. 1.5.2.6 No quality code is set except when Dew Point are given a value of -99.9 Deg C. 2 Quality Control Procedures 2.1 JOSS Quality Control Procedures ------------------------------- These datasets underwent the JOSS QC process which consisted of internal consistency checks and visual quality control. The internal consistency checks included gross limit checks on all parameters and vertical consistency checks on temperature, pressure, and ascension rate. JOSS then visually examined each sounding. 2.1.1 Automated Quality Control Procedures 2.1.1.1 Gross Limit Checks These checks were conducted on each sounding and data were automatically flagged as appropriate. Only the data point under examination was flagged. JOSS conducted the following gross limit checks on the FASTEX sounding datasets. In the table P = pressure, T = temperature, RH = relative humidity, U = U wind component, V = V wind component, B = bad, and Q = questionable. ---------------------------------------------------------------- Parameter(s) Flag Parameter Gross Limit Check Flagged Applied ---------------------------------------------------------------- Pressure < 0 mb or > 1050 mb P B Altitude < 0 m or > 40000 m P, T, RH Q Temperature < -80C or > 30C T Q Dew Point < -99.9C or > 25C RH Q > Temperature T, RH Q Relative Humidity < 0% or > 100% RH B Wind Speed < 0 m/s or > 100 m/s U, V Q > 150 m/s U, V B U Wind Component < 0 m/s or > 100 m/s U Q > 150 m/s U B V Wind Component < 0 m/s or > 100 m/s V Q > 150 m/s V B Wind Direction < 0 deg or > 360 deg U, V B Ascent Rate < -30 m/s or > 10 m/s P, T, RH Q ---------------------------------------------------------------- 2.1.1.2 Vertical Consistency Checks These checks were conducted on each sounding and data were automatically flagged as appropriate. The Auto QC software employed six-second averaging for these checks. These checks were started at the lowest level of the sounding and compared neighboring six-second average values. In the case of checks ensuring that the values increased/decreased as expected, only the data point under examination was flagged. However, for the other checks, all of the data points used in the examination were flagged. All items within the table are as previously defined. ---------------------------------------------------------------- Vertical Consistency Parameter(s) Flag Parameter Check Flagged Applied ---------------------------------------------------------------- Time decreasing/equal None None Altitude decreasing/equal P, T, RH Q Pressure increasing/equal P, T, RH Q > 1 mb/s or < -1 mb/s P, T, RH Q > 2 mb/s or < -2 mb/s P, T, RH B Temperature < -15 C/km P, T, RH Q < -30 C/km P, T, RH B from surface to 800 mb: > 25 C/km (not applied at p < 275mb) P, T, RH Q > 40 C/km (not applied at p < 275mb) P, T, RH B for pressures < 800 mb: > 5 C/km (not applied at p < 275mb) P, T, RH Q > 30 C/km (not applied at p < 275mb) P, T, RH B Ascent Rate change of > 3 m/s or < -3 m/s P Q change of > 5 m/s or < -5 m/s P B ---------------------------------------------------------------- 2.1.2 Visual Quality Control Procedures Each sounding was then visually examined for problems that were not able to be captured via the automated checks described in item .1 above. These problems typically included oddities in the dew point and wind profiles. These two parameters can be highly variable, and hence, the automated checking is more difficult. The visual checking procedure has two main purposes: First, as a check on the results provided by the automatic checks, and second, as a more stringent check on the more variable parameters. 3 Additional quality control processing informations or dataset remarks: Site -------- 03005 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- The VIZ winds had occasionally significant problems. As an example of typical problems there is the 10 February release at 0300 UTC which had a 100 mb layer centered around 800 mb with wind speeds over 100 m/s and rapid directional shifts. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 30 May,1997: problem with the wind data launch time corrected on the following dates: 19970219 14:16 19970206 14:22 19970206 17:25 19970117 05:27 19970209 05:21 19970224 02:19 19970118 05:28 19970219 20:17 03026 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 03240 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 03496 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 03502 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 03808 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 03920 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 03953 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 04018 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 04220 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 04270 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 04339 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 04360 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 01 Dec,1997: addition of radiosoundings at 06UTC and 18UTC 06011 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 07110 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- The pressure levels above 100 mb, and especially those above 50 mb were significantly affected by "bouncing". That is there is a data point with a much higher than average ascension rate followed by a data point with a much lower than average ascension rate. The automated check that flags this type of error is from section 6.1.2 the ascension rate change check. Over 3.2% of all data points at Brest were affected. The average rate over all FASTEX soundings was only 0.28%. And for Brest, of those data points above 50 mb about 40% were flagged as either questionable or bad due to this problem. The Loran-C chain receiving had intermitent significant problems during FASTEX, this led to winds of uneven quality at Brest. At lower and middle levels (up to 300 mb) about 5% of the winds were either flagged as bad or questionable or were missing. At levels above 300 mb this went up to 10-20%. An example of the significant problems that are sometimes present is found in the 5 February sounding taken at 0600 UTC. This sounding includes many drastic wind speed changes of 10-40 m/s as well as drastic directional shifts. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 29 Jul,1997: Soundings dated 19970115:05:57 and 19970215:11:19 interpolated from middle resolution data 07145 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- The pressure levels above 100 mb, and especially those above 50 mb were significantly affected by "bouncing". That is there is a data point with a much higher than average ascension rate followed by a data point with a much lower than average ascension rate. The automated check that flags this type of error is from section 6.1.2 the ascension rate change check. Over 10.1% of all data points at Trappes were affected. The average rate over all FASTEX soundings was only 0.28%. And for Trappes, of those data points above 50 mb about 43% were flagged as either questionable or bad due to this problem. There was also a higher rate of above acceptable ascension rates (> 10m/s) at Trappes (2.11%) versus the entire FASTEX sounding dataset (0.36%). The Loran-C chain receiving had intermitent significant problems during FASTEX, this led to winds of uneven quality at Trappes. At nearly all levels about 5-10% of the winds were either flagged as bad or questionable or were missing. An example of the significant The winds at Trappes were also of uneven quality. At nearly all levels about 5-10% of the winds were either flagged as bad or questionable or were missing. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 29 Jul,1997: Sounding dated 19970218:17:20 interpolated from middle resolution data 28 Aug,1997: Soundings dated 19970118:14:47 and 19970119:23:25 interpolated from middle resolution data 07510 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- The pressure levels above 100 mb, and especially those above 50 mb were significantly affected by "bouncing". That is there is a data point with a much higher than average ascension rate followed by a data point with a much lower than average ascension rate. The automated check that flags this type of error is from section 6.1.2 the ascension rate change check. Over 8.0% of all data points at Bordeaux were affected. The average rate over all FASTEX soundings was only 0.28%. And for Bordeaux, of those data points above 50 mb about 50% were flagged as either questionable or bad due to this problem. There was also a higher rate of above acceptable ascension rates (> 10m/s) at Bordeaux (2.30%) versus the entire FASTEX sounding dataset (0.36%). The Loran-C chain receiving had intermitent significant problems during FASTEX, this led to winds of uneven quality at Bordeaux. At nearly all levels about 5-15% of the winds were either flagged as bad or questionable or were missing. An example of the significant problems that are sometimes present is found in the 24 January sounding taken at 0600 UTC. This sounding includes wind speed oscillations of 20 m/s, drastic speed and direction shifts among other problems. 08001 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 08508 Dataset Remarks (from OFPS checking - February 1998) ----------------------------------------------------- Typograpical Errors ------------------- As noted above, there were many typographical errors within this data set that were corrected by Meteo-France and UCAR/JOSS. There may be other typographical errors that were too small to have been detected or had no obvious correction. Caution should be used. Humidity Values --------------- Data from two different data sets were used in the development of this data set by Meteo-France, the high resolution data set provided by INMG and the TEMP messages. These data sources used different methods in their treatments of very large dew point depressions. The high resolution data set (those data points within the sounding the relative humidity as measured. The TEMP messages (those data points within the sounding that are at pressure levels that are not evenly divisible by 5) set any dew point depression greater than 30 C to 30 C (i.e. leading to an artificially high relative humidity value relative to that which was measured). The effect of this is that in very dry conditions at lower levels the data points from the TEMP message have relative humidity values that are higher than those at the surrounding high resolution data points. This is really noticable only when the relative humidity drops to 1-2% at pressure levels from the surface to about 500-600 mb. Such data have been flagged as questionable during the UCAR/JOSS viusal quality control processing. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 18 Dec,1997: correction on data for the following dates: 19970208 11:16 19970209 23:24 19970216 23:26 30 Dec,1997: addition of radiosoundings from January 30 Jan,1998: minor corrections on dew point values and on launching dates. 08522 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 10 Jun,1997: updating of quality flags for interpolated values 08579 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 10 Jun,1997: updating of quality flags for interpolated values 71600 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 71801 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- Many of the soundings from St. Johns had significant low level humidity problems. The first radiosonde humidity is much lower than the independently measured surface humidity. The subsequent radiosonde humidity values then return to the expected levels. An example of this type of problem can be seen in the sounding from 1 January at 0000 UTC, a small portion of which is shown below: Time Press Temp Dewpt RH sec mb C C % 0.0 971.2 -7.2 -10.9 75.0 3.6 968.0 -7.2 -36.0 8.0 15.0 962.2 -7.9 -18.1 43.8 30.6 950.4 -8.7 -14.3 63.8 60.0 927.8 999.0 999.0 999.0 75.0 916.3 -11.4 -13.7 83.1 The surface RH is 75%, the first radiosonde RH is only 8%, the radiosonde RH then recovers to more expected values. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 03 Jul,1997: data received in the FASTEX Central Archive: pressure, height, temperature, dewpoint depression and relative humidity at significant levels (no high resolution data available for meteorological parameters); 1mn height and wind data. data calculated in the FASTEX Central Archive: U and V wind components, ascension rate, latitude and longitude position of the radiosonde . due to CLASS format constraints, dew points less than -99.9 Deg C were given a value of -99.9 Deg C and were flagged as being estimated. 71816 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 71906 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 72208 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 72402 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 74494 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- These soundings were of generally good quality. 78016 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- Several of the soundings from Bermuda had significant problems. As an example of the most severe problem there is the sounding from 23 January at 0000 UTC. Below is a small section of that sounding: Time Press Temp Dewpt RH Uwind Vwind Wspd Dir sec mb C C % m/s m/s m/s deg 1410.0 433.9 -23.5 -50.0 7.0 8.1 -1.0 8.2 277.0 1420.0 431.0 -23.9 -49.1 8.0 8.1 -1.1 8.2 278.0 1430.0 428.0 -24.3 -49.4 8.0 8.1 -1.1 8.2 278.0 1440.0 425.0 -24.8 -49.8 8.0 8.1 -1.1 8.2 278.0 1450.0 422.2 -25.2 -50.2 8.0 8.0 -1.1 8.1 278.0 1460.0 419.5 -17.7 -35.9 19.0 16.6 -0.6 16.6 272.0 1470.0 430.6 -18.1 -36.2 19.0 16.8 -0.9 16.8 273.0 1480.0 427.8 -18.5 -36.0 20.0 16.9 -1.2 16.9 274.0 1490.0 424.8 -19.0 -36.5 20.0 17.0 -1.5 17.1 275.0 1500.0 421.8 -19.5 -36.4 21.0 17.2 -1.8 17.3 276.0 1510.0 419.0 -19.9 -36.3 22.0 17.4 -2.1 17.5 277.0 Not the dramatic change in all parameters between 1450 and 1460 seconds. All data prior to 1450 seconds is consistent with the 1450 second data, while all data after 1460 seconds is consistent with the 1460 second data. It appears as though this sounding was either combined with another at some point in its processing at Bermuda, or else, less likely, had measurement problems. Other soundings that may also have this problem include: 8 January 0600 UTC and 1800 UTC 23 January 0000 UTC and 0600 UTC 5 February 0000 UTC 25 February 1200 UTC Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 10 Jun,1997: updating of quality flags for interpolated values EOGW Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- There are a significant amount of missing winds in these files. Up to 400 mb about 20% of the winds are missing, starting at 300 mb that increase to 25% and the amount of missing increase such that above 100 mb about 50% of the winds are missing. Many files have very little or no winds. Victor Bugaev used the Vaisala Global Positioning System (GPS) windfinding system which was in a development period during FASTEX. FASTEX was the first operational use of the Vaisala GPS windfinding system and some software problems were present that led to this missing wind problem. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 16 Jul,1998: Ship position corrected for the sounding dated 19970213 05:31 longitude=-34.92 not 34.81 FNOU Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- The soundings from Fort Fleur D'Eppe appeared to be of generally good quality. FNPH Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- The soundings from Fort Desaix appeared to be of generally good quality. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 25 Jul,1997: Sounding dated 19970113:11:06 interpolated from middle resolution data 25 Aug,1997: Corrections at low levels ; soundings dated: 19970207:23:06 19970219:11:15 19970221:23:12 19970222:23:02 19970224:23:19 FZVN Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- The winds at this site were often missing. About 25-30% of all data point had missing winds. Le Suroit used the Vaisala Global Positioning System (GPS) windfinding system which was in a development period during FASTEX. FASTEX was the first operational use of the Vaisala GPS windfinding system and some software problems were present that led to this missing wind problem. KCEJ Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- The winds within 200 mb of the surface were missing about 30% of the time. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 07 Jul,1997: LABSEA experiment data received in the FASTEX Central Archive: position of the ship at launching date (decimal degrees); pressure (hPa), geopotential (m), temperature (C), humidity (%), dewpoint (C), wind speed (kt) and direction (deg); data every 2 seconds for the first 300 seconds after launch, every 5 seconds after. data calculated in the FASTEX Central Archive: U and V wind components, ascension rate, latitude and longitude position of the radiosonde. due to CLASS format constraints, dew points less than -99.9 Deg C were given a value of -99.9 Deg C and were flagged as being estimated. 08 Aug,1997: LABSEA experiment addition of 5 soundings on the following dates: 19970202 23:03 19970213 22:25 19970215 20:09 19970220 14:15 19970227 10:05 OXTS2 Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- The soundings from Irena Arctica appeared to be of generally good quality with the following caveats. First, Irena Arctica was among those most affected by a problem of "bouncing". That is there is a data point with a much higher than average ascension rate followed by a data point with a much lower than average ascension rate. The automated check that flags this type of error is from section 6.1.2 the ascension rate change check. Over 2.5% of all data points at Irena Arctica were affected. The average rate over all FASTEX soundings was only 0.28%. There was also a higher rate of above acceptable ascension rates (> 10m/s) at Irena Arctica (1.06%) versus the entire FASTEX sounding dataset (0.36%). Second, several soundings had a low level humidity problem in which the first radiosonde humidity is much lower than the independently measured surface humidity. The subsequent radiosonde humidity values then return to the expected levels. An example of this type of problem can be seen in the sounding from 16 January at 1200 UTC, a small protion of which is shown below: Time Press Temp Dewpt RH sec mb C C % ------ ------ ----- ----- ----- 0.0 999.5 -0.7 -2.1 90.2 10.0 995.6 -0.7 -12.3 41.1 20.0 990.2 -0.8 -6.4 65.8 30.0 984.5 -1.4 -5.8 72.0 40.0 978.0 -1.9 -5.6 75.8 The surface RH is 90%, the first radiosonde RH is only 41%, the radiosonde RH then recovers to more expected values. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 5 Sep,1997: correction on latitude and longitude on the following date: 19970212 22:59 TFTA Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- The winds at this site were often missing. About 25-30% of all data point had missing winds. Aegir used the Vaisala Global Positioning System (GPS) windfinding system which was in a development period during FASTEX. FASTEX was the first operational use of the Vaisala GPS windfinding system and some software problems were present that led to this missing wind problem. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 16 Jul,1998: Ship position corrected for the sounding dated 19970212 05:53 latitude= 45.85 not 48.85 V2EZ Dataset Remarks (from OFPS checking - September 1997) ----------------------------------------------------- The soundings from Godafoss appeared to be of generally good quality with the following caveat. There were two varieties of low level humidity problems observed with the soundings from the Godafoss. First, several soundings beginning on 18 February had very low independently measured surface humidities. For example, the 18 February sounding taken at 18 UTC had an independently measured surface RH of 1.0%, while the first radiosonde RH was 60.2% and the radiosonde RH values above that point were consistant with that value. Second, several soundings had a low level humidity problem in which the first radiosonde humidity is much lower than the independently measured surface humidity. The subsequent radiosonde humidity values then return to the expected levels. An example of this type of problem can be seen in the sounding from 16 February at 0000 UTC, a small protion of which is shown below: Time Press Temp Dewpt RH sec mb C C % ------ ------ ----- ----- ----- 0.0 991.5 1.8 0.3 89.8 10.0 986.4 1.4 -12.4 35.0 20.0 980.6 0.8 -6.1 59.9 30.0 975.1 0.2 -5.8 64.0 The surface RH is 90%, the first radiosonde RH is only 35%, the radiosonde RH then recovers to more expected values. Additional quality control processing information ------------------------------------------------- Timestamp Comment --------- ------- 16 Jul,1998: Ship position corrected for the soundings dated: 19970105 23:40 longitude =-67 not -60.7 19970121 23:21 latitude = 52.33 not 55.33 19970219 23:00 longitude =-46.42 not -40.42
Updated: 2 February 1999