This data set contains hourly resolution surface meteorological data in National Center for Atmospheric Research/Earth Observing Laboratory (NCAR/EOL) Quality Control (QC) format from stations within the following networks:
Data for the Terrain-induced Rotor Experiment (T-REX) domain (34N to 40N latitude and 115W to 126W longitude) and time period (01 March 2006 through 30 April 2006) are contained within this data set. This T-REX 2006 Hourly Surface Meteorological Composite data set contains data from 933 stations.
Section 2.0 contains a detailed description of the instrumentation, siting, and algorithms used by the source network to collect the data. Section 2.1 contains a detailed description of the format of the composite data set. See Section 2.2 for information on data processing, and Section 3.0 below for the quality control processing performed by NCAR/EOL on this data set. Section 5.0 contains references.
Please review Section 4.0 for the T-REX Data Policy and additional data usage restrictions.
This data set contains surface meteorological data from the California Irrigation Management Information System (CIMIS) weather stations. CIMIS is a program in the Office of Water Use Efficiency (OWUE) in the California Department of Water Resources. There are 98 CIMIS stations included in this T-REX 2006 Hourly Surface Meteorological Composite.
All stations have their equipment and sensors mounted on a mast that is mounted on a tripod base. The sensors and their heights are as follows:
Air temperature/relative humidity (HMP35)
Sensor: Fenwall Thermistor/HUMICAP H-sensor Model: HMP35C Maker: Vaisala/modified by Campbell Scientific, Inc. Height: 1.5 m Range: 0 to 100% RH, -35 to +50 degrees C Accuracy: ±2% RH (0-90% RH), ±5% RH (90-100%), ±0.1 °C over -24 to 48 °C range Note: Both sensors are enclosed in a 12-plate naturally aspirated radiation shield made by R. M. Young.
Wind direction (wind vane)
Sensor: 2.0 meters Model: 024A Maker: Met-One Height: 2.0 meters Range: 0-360 degrees Output: 0-10 * 10³ Ohms Threshold: 0.45 m per sec (1 mph) Accuracy: ±5% Delay distance: less than 1.3 m
Wind speed (anemometer)
Sensor: Three-cup anemometer utilizing a magnet activated reed switch whose frequency is proportional to wind speed Model: 014A Maker: Met-One Height: 2.0 meters Range: 0-45 m per sec (0-100 mph) Threshold: 0.45 m per sec (1 mph) Gust Survival: 0-53 m per sec (0-120 mph) Accuracy: 1.5% or 0.11 m per sec (0.25 mph)
Precipitation (tipping-bucket rain gauge)
Sensor: Tipping-bucket rain gauge with magnetic reed switch. Model: TE525MM Maker: Texas Electronics Height: 1.0 meters Orifice: 24.5 cm (9.644 in) Resolution: 0.1 mm Accuracy: ±1% at 5 cm per hr or less.
Dew Point Temperature
Dew point temperature is the temperature to which the atmosphere must be cooled, at constant pressure and water vapor content, in order to reach saturation. It is calculated from vapor pressure (relative humidity) and air temperature data (CIMIS Sensors, cited 2007).
A small datalogger (a small microprocessor) mounted on the mast of each station is connected to the electronic sensors by wires. Once every minute, the datalogger takes a reading of each sensor and records it. It continues for one hour, then averages the 60 readings for an hourly average, or accumulates the 60 readings for an hourly total, and stores all these in its memory. It continues collecting data in this manner throughout the day (midnight to midnight). After 24 hours have elapsed, the datalogger then calculates the daily averages and totals using the 24 hourly averages and totals and stores them. The data- loggers also determines the maximum and minimum temperatures and relative humidities. These values are the highest and lowest of the one-minute readings during the day.
After 12:00 midnight (always Pacific Standard Time) a microcomputer located with the main CIMIS computer begins the interrogations of the stations. This communication is done using telephone lines. Each station has its own telephone service and modem. The microcomputer makes a phone connection with a datalogger and the data is dumped to the micro- computer. This continues through the early morning hours. The majority of the stations will be interrogated by 4:00 a.m. Some stations will not answer the phone call from the microcomputer during these early morning hours due to problems with the telephone lines themselves. Usually these stations will answer the telephone by 9:00 or 10:00 a.m.
The collected data is transferred to the CIMIS computer. The CIMIS computer then performs three tasks: 1) Reference Evapotranspiration (ETo) is calculated, 2) the data is run through the quality control program and flags are placed appropriately, and 3) the data is stored in a relational database for retrieval by CIMIS users (CIMIS Network, cited 2007).
The following are codes that were applied to data during the T-REX time period.
The entire list of quality control flags used by CIMIS can be found on the Current Hourly Flags site (CIMIS Flags, cited 2007).
The R and Y flags for the precipitation were translated to the QCF questionable flag. The R wind direction flag was translated to the QCF Unlikely Flag. The I wind speed flag was translated to the QCF questionable flag.
This data set contains surface meteorological data from the Great Basin Unified Air Pollution Control Division (GBUAPCD). Information on the instruments, data collection and processing are unavailable. There are 6 stations included in this T-REX 2006 Hourly Surface Meteorological Composite.
More information on GBUAPCD can be found at http://www.gbuapcd.org/index.htm.
PestCast is a project of the University of California Statewide Integrated Pest Management (IPM) Program and the California Environmental Protection Agency Department of Pesticide Regulation. Data are collected by University of California cooperatve observers. There are 8 stations included in this T-REX 2006 Hourly Surface Meteorological Composite.
Air temperatures are measured at a height of 5 to 6 feet. Wet-bulb and dry-bulb temperature readings are made from a psychrometer, to be used in calculating dew point and relative humidity. Wet-bulb and dry-bulb temperature readings are made at the time of observation. They are entered to a tenth of a degree when the value is less than 100F. Above 100F, only whole degrees are entered.
More information on PestCast can be found at http://www.ipm.ucdavis.edu/WEATHER/abtwxvars.html.
The Unidata Local Data Manager (LDM) (Unidata, 2007 ) distributes World Meteorological Organization (WMO) Surface data. These data are ingested by UCAR/EOL in ASCII WMO meteorological message structure format (NOAA/NWS, 2007). The primary feedset name is "WMO" which includes Public Product Service (PPS), Domestic Data Service (DDS), High resolution Data Service (HDS), and International Data Service (IDS) feedtypes. Only products that match the patterns ^S[AP].* .... ([0-3][0-9])([0-2][0-9]) and ^SX..81 .... ([0-3][0-9])([0-2][0-9]) are collected. In these patterns, S stands for surface, A for Aviation Routine Reports (FM 15 - METAR), P for Special aviation weather reports (FM 16 - SPECI), and X for miscellaneous text records. Only hourly METAR data are included in this dataset.
For information on the METAR format see the ASOS User's Guide, 1998 [PDF] and ASOS User's Guide Appendices, 1998 [PDF]. Observing, reporting, and coding standards for surface-based meteorological observations from all federal agencies are defined in the Federal Meteorological Handbook 1. (OFCM, 2005) For 20 minute METAR stations, the observation that falls between 15 minutes before the hour and the hour, inclusive, is included in this dataset. If there is no observation in this time period, then the observation closest to the hour and falling between 1 minute and 15 minutes after the hour is included in this dataset. There are 82 LDMSFCMETR stations in this TREX Hourly Surface Composite.
This dataset contains ASOS, AWOS (USDOT, 1988 ), and MANUAL stations. Station ID's are 3 characters long. Some networks use the 4-character ID to refer to these stations. To obtain the 4-character id, prepend a "K". For example, station ABR could also be referred to as station KABR.
There were 4 stations during the TREX time period that had elevation changes of a few hundredths of a meter. All of these elevation changes occurred at either 20:00 UTC or 21:00 UTC on Apr 26, 2006. Because the netCDF format we deliver the data in does not handle changes within a single file, the elevations of the last 3 or 4 times for these stations were manually changed to match the beginning of the day, so the change appears to happen 3-4 hours later than it actually does. Since these changes are so small, it was deemed this would not be a problem.
The stations and elevation changes are as follows:
|Stn||elev 1||elev 2|
This dataset is a collection of data from many networks. The data are fed to EOL over the LDM by the FSL Meteorological Assimilation Data Ingest System (MADIS). Some of these networks are themselves collections of data from unrelated sources. Each of these networks/subnetworks can contain different frequency data and different parameters. 734 stations from the MADIS LDM feed are included in this TREX Hourly Surface Composite.
The hourly parameter value given in this composite is the value closest to the hour that falls at or after 45 minutes and before (but not at) 15 minutes after the hour. MADIS QC flags are not carried forward to EOL QC format. However, the MADIS QC summary value X "Failed QC stage 1", which is a gross limit check, has been used to mask out extremely unlikely data. If the data fails this check, it is set to missing.
A wealth of information on the MADIS system and the networks provided can be found with the FSL/MADIS website at http://www-sdd.fsl.noaa.gov/MADIS (FSL, 2007). For more information on these networks, see the FSL/MADIS Surface Network Information webpage A list of the reporting frequencies of each network can be found at http://www-sdd.fsl.noaa.gov/MADIS/mesonet_temporal_frequency.html. Note that these are the native reporting frequencies of the networks. This composite only contains the value reported closest to the hour as described above.
ESRL/GSD and its data providers disclaim liability of any kind whatsoever, including, without limitation, liability for quality, performance, merchantability and fitness for a particular purpose arising out of the use or inability to use the data.
Please review Section 4.0 for information on the T-REX Data Policy and additional data usage restrictions.
This dataset is formed by extracting Hourly Surface Meteorological Data from the T-REX 2006 Five Minute Surface Meteorological Composite. Refer to the T-REX 2006 Five Minute Surface Meteorological Composite description document for more information.
The following networks were extracted from the Five Minute Meteorological Composite. Each network is followed by the number of stations it contains in parenthesis: ASU Flux Tower Sonics (1); China Lake Handar (12); Desert Research Institute (16); NCAR/EOL Integrated Surface Flux Facility (3); NCAR/EOL Integrated Sounding System (2); Mobile Integrated Sounding System (5); Tribal Environmental Exchange Network (4); University of Houston Flux Tower (1); University of Leeds Automatic Weather Stations (15); and University of Utah HOBO stations (49).
An hourly file is created from a 5-minute file by selecting the (xx-1):55 observation as the xx hourly observation, with the exception of the precipitation parameter, which is totaled from (xx-1):05 through xx:00 (where xx is the 2-digit hour and xx-1 is the previous hour), and placed in the xx hourly observation. For example, the 2PM hourly observation contains the 1:55 5-minute observation for all parameters except precipitation. The 2PM hourly precipitation value is a sum of all 5-minute precipitation values from 1:05 to 2:00 inclusive. If any 5-minute precipitation values in this range are missing, then the hourly value is set to missing.
Please review Section 4.0 for information on the T-REX Data Policy and additional data usage restrictions.
The T-REX Hourly Surface Composite observation data contains 10 metadata parameters and 38 data parameters and flags. The metadata parameters describe the station location and time at which the data were collected. The time of observation is reported both in Universal Time Coordinated (UTC) Nominal and UTC actual time. For this Hourly surface composite, reported nominal time and actual time are the same. Days begin at UTC 0100 and end at UTC 0000 the following day. The table below details the data parameters in each record. Several data parameters have an associated Quality Control (QC) Flag Code which are assigned by the Earth Observing Laboratory Data Management Group. For a list of possible QC Flag values see the Quality Control Section 3.0.
Parameters Units ---------- ----- Date of Observation UTC Nominal Time of Observation UTC Nominal Date of Observation UTC Actual Time of Observation UTC Actual Network Identifier Abbreviation of platform name Station Identifier Network Dependent Latitude Decimal degrees, South is negative Longitude Decimal degrees, West is negative Station Occurrence Unitless Station Elevation Meters Station Pressure, QC flag Hectopascals (mb) Reported Sea Level Pressure, QC flag Hectopascals (mb) Computed Sea Level Pressure, QC flag Hectopascals (mb) Dry Bulb Temperature, QC flag Celsius Dew Point, QC flag Celsius Wind Speed, QC flag m/s Wind Direction, QC flag Degrees Total Precipitation, QC flag mm Squall/Gust Indicator Code Value Squall/Gust Value, QC flag m/s Present Weather, QC flag Code Value Visibility, QC flag Meters Ceiling Height (first layer) Hundreds of feet Ceiling Flag (first layer), QC flag Code Value Cloud Amount (first layer), QC flag Code Value Ceiling Height (second layer) Hundreds of feet Ceiling Flag (second layer), QC flag Code Value Cloud Amount (second layer), QC flag Code Value Ceiling Height (third layer) Hundreds of feet Ceiling Flag (third layer), QC flag Code Value Cloud Amount (third layer), QC flag Code Value The list of code values for the Present Weather is too large to reproduce in this document. Refer to WMO, 1988 for a complete list of Present Weather codes. The code values for the Squall/Gust Indicator are: Code Definition ---- ---------- blank No Squall or Gust S Squall G Gust The code values for the ceiling flag Indicator are: Code Definition ---- ---------- 0 None 1 Thin 2 Clear below 12,000 feet 3 Estimated 4 Measured 5 Indefinite 6 Balloon 7 Aircraft 8 Measured/Variable 9 Clear below 6,000 feet (AUTOB) 10 Estimated / Variable 11 Indefinite / Variable 12 12-14 reserved 15 Missing The code values for the Cloud Amount Indicator are: Code Definition ---- ---------- 0 0 ( or clear) 1 1 okta or less, but not zero or 1/10 or less, but not zero 2 2 oktas or 2/10-3/10 3 3 oktas or 4/10 4 4 oktas or 5/10 5 5 oktas or 6/10 6 6 oktas or 7/10-8/10 7 7 oktas or more, but no 8 oktas or 9/10 or more, but not 10/10 8 8 oktas or 10/10 (or overcast) 9 Sky obscured by fog and/or other meteorological phenomena 10 Sky partially obscured by fog and/or other meteorological phenomena 11 Scattered 12 Broken 13 13-14 Reserved 15 Cloud cover is indiscernible for reasons other than fog or other meteorological phenomena, or observation is not made.
This data set contains hourly observations for the T-REX domain and time period. The component data sets from which this data set was compiled are available on-line in native format via the T-REX Master Table of data sets (NCAR/EOL, 2006)
Calculated Sea Level pressure is computed from station pressure, temperature, dew point, and station elevation using the formula of Wallace and Hobbs (1977).
When not present in the raw data, the dew point temperature was computed by NCAR/EOL from station pressure, temperature, and relative humidity using the formula from Bolton (1980). This calculation was done for the following networks: China Lake Handar, Desert Research Institute (DRI) Automatic Weather Station Data, NCAR/EOL/ISFF, NCAR/EOL/ISS, NCAR/EOL/MISS, TREX, University of Houston Flux Tower, and University of Leeds AWS.
This T-REX Hourly Surface Composite does not contain any Sea Level Pressures.
The T-REX 2006 Hourly Surface Meteorological Composite was formed from several datasets:
These T-REX 2006 Hourly Surface Meteorological Composite datasets were collected over the T-REX 2006 domain (i.e., 34N to 40N latitude and 115W to 126W longitude) and time period (1 March 2006 through 30 April 2006) and were combined to form a surface composite. The composite was quality controlled to form the final T-REX 2006 Hourly Surface Meteorological Composite. NCAR/EOL Mobile ISS (MISS) is included in the T-REX 2006 Hourly Surface Meteorological Composite, but no mobile data were included in the final Horizontal Quality Control processing.
During the NCAR/EOL Horizontal Quality Control (NCAR/EOL HQC) processing, station observations of pressure, temperature, dew point, wind speed and wind direction were compared to "expected values" computed using an objective analysis method adapted from that developed by Cressman (1959) and Barnes (1964). The NCAR/EOL HQC method allowed for short term (>/= 30 day) variations by using 30 day standard deviations computed for each parameter when determining the acceptable limits for "good", "questionable", or "unlikely" flags. "Expected values" were computed from inverse distance weighted station observations within a 100 km Radius Of Influence (ROI) centered about the station being quality controlled (the station being quality controlled was excluded); i.e.;
theta_e = < theta(i)/w(i) > / < w(i) >
Where theta_e is the "expected value" of the parameter at the site in question, theta(i) is the observed value of the parameter at site i, w(i) is the weighting factor for site i (here the inverse of the distance between site i and the station being quality controlled), and <...> is the sum over all stations "i" in the current ROI that have valid observations of the parameter at the time in question. Only sites with elevation differences of less than or equal to 200.0 meters were included in the calculation of the "expected value". Data were always compared at like solar times.
To determine an observation's HQC flag setting, the difference between the actual observation and its "expected value" was compared to that parameter's normalized standard deviation. Normalizing factors (also called the sensitivity coefficients) were chosen to control the "good", "questionable", and "unlikely" flag limits for each parameter. See Table 3-1 for T-REX 2006 normalizing factors. Table 3-2 contains the HQC flag limit ranges derived from the normalizing factors given in Table 3-1 and estimated standard deviations for each parameter so that 95% of the QC limits applied to the T-REX 2006 data fell within these ranges. For example, 95% of the observed sea level pressure values that were flagged as "good" were within 1.3 mb of the expected value. The significant overlap of the ranges seen in Table 3-2 was partially due to seasonal and station differences in standard deviations. The actual HQC limits applied at any particular time depended upon the dynamic nature of the particular station's parameter values over time.
Data were never changed, only flagged.
HQC was only applied to sea level pressure, calculated sea level pressure, temperature, dew point, wind speed and wind direction. If the calculated sea level pressure quality control information was available, its flag was applied to the station and sea level pressures. If the calculated sea level pressure could not be quality controlled, the sea level pressure quality control flag was applied to the station pressure. If the sea level pressure could not be quality controlled, the station pressure quality control flag was not overridden.
Table 3-1 Normalizing factors used for T-REX 2006 Hourly Surface Meteorological Composite
Parameter Good Questionable Unlikely --------- ---- ------------ -------- Sea Level Pressure (SLP) 0.2 0.2 0.5 Calculated SLP 0.4 0.4 1.0 Dry Bulb Temperature 0.5 0.5 1.0 Dew Point Temperature 0.5 0.5 1.0 Wind Speed 2.25 2.25 4.0 Wind Direction 1.22 1.22 2.2Table 3-2 Ranges of HQC flag limit values for T-REX 2006 Hourly Surface Meteorological Composite
Parameter Good Questionable Unlikely --------- ---- ------------ -------- Sea Level Pressure (mb) < 1.3 [0.6-3.2] > 1.4 Calculated SLP (mb) < 2.7 [1.1-6.8] > 2.8 Dry Bulb Temperature (deg.C) < 2.5 [0.9-5.0] > 1.7 Dew Point Temperature (deg.C) < 2.9 [0.9-5.7] > 1.7 Wind Speed (m/s) < 7.7 [0.5-13.7] > 0.9 Wind Direction(degrees) < 150.7 [38.6-180.0] >69.6
The squall/gust wind speed data were not quality controlled.
General consistency checks were also applied to the dry bulb temperature, wind direction, squall/gust, and the relationship between precipitation and cloud amount/cloud cover. If the dew point temperature was greater than the dry bulb temperature both values were flagged "questionable". Also, wind direction for observed "calm" winds was given the same QC code as the wind speed. If the wind speed was greater than the squall/gust, then the squall/gust QC code was set to "questionable". If precipitation was reported, but the cloud amount was "none" or "clear", then both the cloud amount and precipitation values were flagged "questionable".
Several impossible values were also checked. Negative wind speeds were flagged "unlikely". Negative squall/gust wind speeds were flagged "unlikely". Wind directions of less than 0 degrees or greater than 360 degrees were flagged "unlikely". If these consistency checks would have upgraded the quality control flags previously set by HQC or gross limit checks, then they were not applied. However, if these consistency checks would have degraded the previously set QC flags, they were applied.
The NCAR/EOL HQC scheme relied on spatial and temporal continuity to flag the data. It has been shown that this method works very well for temperature, dew point, pressure, and wind speed, but is not a very good scheme for the wind direction.
Gross limit checks were also used to determine the quality of the precipitation values. The gross limits are shown in Table 3-3. Negative precipitation was flagged "unlikely".
Table 3-3 - Precipitation Gross Limit Values
Parameter Good Questionable Unlikely --------- ---- ------------ -------- Hourly Precip <20.0 mm >=20.0 mm >= 50.0 mmCertain "questionable" and "unlikely" data values were also manually inspected. After inspection, the quality control flag may have been manually modified to better reflect the physical reasonableness of the data. Data were never modified, only flagged. See Table 3-4 for a list of the possible quality control flags and their meanings.
Table 3-4 - Quality Control Flags
QC Code Description ------- ----------- U Unchecked G Good M Normally recorded but missing. D Questionable B Unlikely N Not available or Not observed X Glitch E Estimated C Reported value exceeds output format field size or was negative precipitation. T Trace precipitation amount recorded I Derived parameter can not be computed due to insufficient data.
Any redistribution of this data must include this data acknowledgment statement.
Please use the following citation as an acknowledgment, if using this data in any scientific report/ paper/presentation:
Any redistribution of this data must include this data acknowledgment statement.
Any redistribution of this data must include this data acknowledgment statement.
Please also refer to the T-REX Data Policy for general usage, citation and acknowledgment policy.
ASOS User's Guide, 1998 , ASOS Project Office, NOAA, National Weather Service, Washington D.C., June 1998. [Available online from http://www.nws.noaa.gov/asos/aum-toc.pdf ]
ASOS User's Guide Appendices, 1998 , ASOS Project Office, NOAA, National Weather Service, Washington D.C., June 1998. [Available online from http://www.nws.noaa.gov/asos/appen.pdf ]
Barnes, S. L., 1964: A technique for maximizing details in numerical weather map analysis. J. Appl. Meteor., 3, 396-409.
Bolton, D., 1980: The computation of equivalent potential temperature., Mon. Wea. Rev., 108, pp 1046-1053.
CIMIS, cited 2007, Network Maintenance [Available online from
CIMIS, cited 2007, Current Hourly Flags [Available online from http://wwwcimis.water.ca.gov/cimis/dataQcCurrentHourly.jsp]
CIMIS, cited 2007, Sensor Specifications [Available online from http://wwwcimis.water.ca.gov/cimis/infoStnMaint.jsp]
Cressman, G. P., 1959: An operational objective analysis system. Mon. Wea. Rev., 87, 367-374.
De Wekker, S. F. J., and C. D. Whiteman, 2006: On the time scale of nocturnal boundary layer cooling in valleys and basins and over plains. J. Appl. Meteor., 45 (6), 813-820.
FSL, cited 2007: Meteorological Assimilation Data Ingest System (MADIS) Surface Network Information [Available online from http://www-sdd.fsl.noaa.gov/MADIS]
Mayr, G. J., L. Armi, S. Arnold, R. M. Banta, L. S. Darby, D. D. Durran, C. Flamant, S. Gabersek, A. Gohm, R. Mayr, S. Mobbs, L. B. Nance, I. Vergeiner, J. Vergeiner, and C. D. Whiteman, 2004: GAP flow measurements during the Mesoscale Alpine Programme. Meteorology and Atmospheric Physics, 86, no. 1-2, 99-119.
NCAR/EOL ISFF, cited 2006: NCAR Integrated Surface Flux Facility at T-REX [Available online from http://www.eol.ucar.edu/rtf/projects/trex/isff/]
NCAR/EOL ISS, cited 2006: NCAR Integrated Sounding System at T-REX [Available online from http://www.eol.ucar.edu/rtf/projects/t-rex/iss/]
NCAR/EOL, cited 2006: T-REX Master Table of Datasets [ Available online from http://data.eol.ucar.edu/master_list/?project=T-REX]
NOAA/NWS, cited 2007: WMO Message Structure 2000 Paraphrased Version [Available online from http://www.nws.noaa.gov/tg/head.html]
OFCM, cited 2005: Federal Meteorological Handbook No. 1 Surface Weather Observations and Reports [Available online from http://www.ofcm.gov/fmh-1/fmh1.htm]
Unidata, cited 2007: Unidata Local Data Manager (LDM) [Available online from http://www.unidata.ucar.edu/packages/ldm/]
United States Department of Transportation (USDOT), 1988. AWOS Operations Manual, Federal Aviation Administration.
Wallace, J.M., P.V. Hobbs, 1977: Atmospheric Science, Academic Press, 467 pp. World Meteorological Organization (WMO), 1988: Manual on Codes Volume I, Part B - Binary Codes. WMO, Geneva, Switzerland.
Whiteman, C. D., J. M. Hubbe, and W. J. Shaw, 2000: Evaluation of an inexpensive temperature data logger for meteorological applications. J. Atmos. Oceanic Technol., 17, 77-81.
Whiteman, C. D., S. Zhong, W. J. Shaw, J. M. Hubbe, X. Bian, and J. Mittelstadt, 2001: Cold pools in the Columbia Basin. Weather and Forecasting, 16, 432-447.
Whiteman, C. D., T. Haiden, B. Pospichal, S. Eisenbach, and R. Steinacker, 2004: Minimum temperatures, diurnal temperature ranges and temperature inversions in limestone sinkholes of different size and shape. J. Appl. Meteor., 43 (8), 1224-1236.
Whiteman, C. D., S. Eisenbach, B. Pospichal, and R. Steinacker, 2004: Comparison of vertical soundings and sidewall air temperature measurements in a small Alpine basin. J. Appl. Meteor., 43 (11), 1635-1647.