[Next] [Prev] [Contents] [Top]

An Introduction to Atmospheric and Oceanographic Datasets


Surface Observations

The term `surface observation' refers to observations made near the surface. Temperature and humidity instruments are generally located in instrument shelters about 2.0 meters above the ground (
Fig. 3.1). The shelters are meant to protect the instruments from exposure to direct sunshine, precipitation and condensation while at the same time providing adequate ventilation. Historically, the design of these shelters has varied from country-to-country and shelters within countries have evolved over time. In some instances, the shelters or the instruments introduced systematic biases. An example of a systematic error is a thermometer which consistently measures temperatures too high or low. Precipitation measurements are made by a variety of devices and methods. The most common method is to measure the water depth within a container. The accuracy of the measurements can be affected by both the intensity of the precipitation and, in particular, the wind speed. Also, it should be noted that precipitation observations often show considerable spatial variation, even at nearby locations, especially in showery weather and in areas where topography influences local conditions.

Most countries have several `classes' of observing stations making surface measurements. In practice there is considerable overlap in the tasks required of each class. For example, the U.S. classifications include first order, second order or airway, and co-operative stations. The differences are in the variables measured and the frequency of the observations. A first order station operates 24 hours-a-day and is maintained by a NWS trained and certified staff. The observations made at these stations span a wide number of variables and usually include: temperature, precipitation, surface pressure, humidity, wind speed and direction, cloud cover, snow depth, visibility, solar radiation and current weather. The observations are taken frequently, either on an hourly basis or at specific times (e.g., 00, 06, 12, 18 UTC). Second order stations are maintained by a trained staff who are supervised by NWS personnel. The majority are operated by the FAA to provide weather observations in support of aircraft operations. These stations record observations at airports which include many, but not necessarily all, of the variables measured at first order stations. The frequency of observation can be hourly but may be less. Co-operative ("co-op") stations are stations operated by of the NWS Co-op Observing Program. This program consists of approximately 7,750 stations manned by volunteers who regularly report observations. Typically, the measured variables at co-op stations include only temperature and precipitation and the observations are made much less frequently. In fact, many stations just record the temperature extremes Tmin and Tmax and precipitation once per day. Since some of the co-op stations are affiliated with agricultural interests they may also record soil temperature, soil moisture and evaporation. The precipitation reported by co-op stations is not necessarily the amount which occurred on a specific calendar day. Rather, it is a 24-hour precipitation amount. For example, it could be the total precipitation from 0500 on one day to 0500 the next day. At stations which record only once per day the reported precipitation may be less than actually occurred due to evaporation from the measuring container, especially in dry climates.

One major advantage of surface station observations is that they span the longest time periods and are, therefore, often used in climate studies. However, there are many limitations associated with these observations of which researchers should be aware. These include outliers, inhomogeneities, missing data, spatial-temporal sampling problems, different instrument types, instrument bias,etc. Inhomogeneities in observational data may result from changes in a station's geographic location, elevation, observing times, instruments, averaging techniques, surrounding environment, observers, etc. Station location and elevation changes are generally recorded and may (possibly) be accounted for, but, in general, information on other sources of inhomogeneities (i.e., metadata) is often not available. Even if metadata are available, a complicated process must be used to account for the various factors. Since the metadata are often site specific it can be labor intensive to incorporate the information correctly.

A time series of December sea-level pressures from Bombay, India spanning 1920 to 1993 illustrates a commonly encountered inhomogeneity (Fig. 3.2). A discontinuity is clearly evident around 1960. No documentation is available and neither the location nor the elevation have changed according to the available information. However, many countries reprocess data at decadal intervals and it is possible that the averaging procedure changed. Another possibility is that beginning in 1960 the observations were taken at different times of the day. In any event, some procedure should be used to identify and, possibly, adjust for the discontinuity.

Instrument quality varies from country-to-country. Surface temperatures can be recorded to within 0.5 degree C. Precipitation is often recorded to the nearest millimeter (mm). However, the recorded precipitation amounts (both rain and snow) are often underestimated due to winds, evaporation and equipment design. Underestimates typically range from only a few percent to 50% in some extreme cases. On a global basis it is estimated that precipitation amounts are underestimated by about 15%. Figure 3.3 illustrates the corrected and uncorrected annual precipitation for the former USSR and the USA.

The spatial distribution of stations making surface observations over the mid-latitudes of the northern hemisphere is uneven but adequate for many research projects interested in studying large-scale phenomena. However, over much of the tropics and the southern hemisphere the number of stations are far fewer. Figure 3.4 illustrates a sample spatial distribution of stations currently reporting daily maximum and minimum temperatures and precipitation over the GTS. (The data are from NCAR ds512.0) It must be emphasized that the number and the spatial distribution of stations has varied considerably with time. In the early years (1880s onward) observations are generally from European and North American stations. The tropics and the southern hemisphere have had considerably fewer stations which persists to the present time. Thus, there are large spatial and temporal gaps in the surface station coverage.

Many countries have established national climatological reference stations to maintain a high-quality basis for studying climate change (ds564.0). These reference stations are chosen based upon length of record, location and the quality of the observations. In some instances, the climate records have been adjusted after careful analysis. These adjustments may include comparisons with nearby stations, correcting for differing observation times over the period of record, and other adjustment procedures. In the U.S., this is called the Historical Climatology Network (HCN; ds565.0). It is a product of a joint effort between NCDC and CDIAC. It consists of year-month data for 1219 stations which nominally span 1880 to the present. Daily data for a selected subset of these high-quality stations is contained within ds510.0

Climatologies and Gridded Data of Surface Variables

Climatologies for several surface variables are available. These climatologies, which are available on grids with different resolutions, represent the `average' of a quantity over a particular time period. They are available in digital form as station files or gridded fields and in printed form as contoured maps or tables. Measures of climate variability and covariability (e.g., variances, covariances, correlations, standard deviations, anomalies, percentiles, spectra, empirical orthogonal functions, extremes, etc.) are available less often. These statistical quantities are used for many purposes including agricultural planning and validation of climate models. Some examples of climatologies include the monthly mean temperature, precipitation and sea level pressure over the globe for a 30-year period, or the probability of precipitation occurring on any particular day of the year at a particular location.

As part of the daily operation procedures at meteorological centers (e.g., NMC and ECMWF), gridded analyses of a number of different variables are produced. These grids are generated at a number of levels including near the surface. A detailed description of gridded analyses is presented in Chapter 6. At NMC and NSIDC, gridded analyses of snow cover over the globe and the Northern Hemisphere are generated. These are available from NCAR.

The Global Precipitation Climatology Center (GPCC) is an an example of a special WMO effort. The GPCC is a unit for collection, analysis, and archiving monthly precipitation data on a global scale an on routine basis. It produces gridded estimates of monthly precipitation. Its products are made available from NCDC (see Appendix A).

Upper Air Observations

"Upper air" observations began in the 1940s. However, large numbers of these observations only became available starting with the International Geophysical Year (IGY; 1957-58). Today, about 1000 stations operated by about 90 countries make upper air observations up to four times per day at internationally agreed-upon times. These upper air observations are made by radiosondes or rawinsondes. A radiosonde (invented in 1927) is an expendable balloon-borne instrument which measures pressure, temperature and humidity and relays the information to an observing station where the data are recorded after corrections are made for instrument response time and, sometimes, other errors.

A rawinsonde is a radiosonde whose three dimensional position is measured as a function of time. Because the balloon drifts with the wind, the position and time information can be used to estimate the winds aloft. Before World War II, the radiosonde balloon was tracked optically using a theodolite. During the war the Army Signal Corps developed an electronic tracking system know as the Signal Corps Radiotheodolite (SCR-658) so that winds aloft measurements could be made in cloudy conditions. The SCR-658, the Army-Navy Ground Meteorological Device (AN/GMD) and the Weather Bureau Radiotheodolite (WBRT) track the radiosonde by homing on the signal that the radiosonde sends to the ground station. The radiosonde can be tracked by radar if a reflector is attached to the instrument or by using the Omega or Loran navigational systems. Modern sounding sytems use both radiotheodolites and the Omega/Loran postion finding to measure winds. In the future they may be tracked via GPS. These upper air observations are referred to as "raob" data, regardless of which instrument was actually used.

The height of each report is derived by integrating the "hydrostatic equation":


p = pressure (hPa)
z = height (m)
p = air density (kg m-3)
g = gravity (m s-2)
This equation expresses a mechanical balance between the downward force of gravity acting upon the mass of the atmosphere and a pressure gradient force acting upward. The hydrostatic equation may be integrated using the humidities, temperatures and pressures measured by the raob to provide the height of the observations to a high degree of accuracy. Only in cases of severe thunderstorms and some other small-scale systems will the accuracy of this relationship be seriously compromised.

The wind speeds and directions at the pressure levels are estimated from the horizontal displacement of the balloon using: (i) the heights calculated from the hydrostatic equation, (ii) measured azimuth and elevation angles of the balloon or other location indicators, and (iii) trigonometric relationships.

The words "height" and "geopotential height" (Z) are sometimes carelessly interchanged. However, they are subtly different. Height (or altitude) refers to the absolute distance above sea-level. Geopotential height is what is actually reported in raob observations. The geopotential height is closely related to altitude but accounts for variations of gravity within latitude and height. It is defined by the following relationship

where g0 is a constant approximating the value of gravity at sea level. In the past, this value has been taken as 9.8 ms-2, but in 1993 (in the US) this changed to 9.80665 ms-2 which is the standard value of gravity at 45 degree latitude used by WMO to calibrate barometers. (In fact, this differs from the real value at 45 degree latitude which is 9.80616 ms-2). The geopotential height is proportional to the potential energy of a unit mass relative to sea level. Clearly, g = g0if there is no numerical difference between z and Z. Since g =~ g 0 in the lower atmosphere the two quantities are numerically interchangeable for most meteorological purposes. However, in the upper stratosphere and higher, the differences can be significant.

The data from raobs are recorded at numerous constant pressure levels as they rise through the atmosphere. (The mean ascent rate of a raob is about 5 m/sec. Thus, it took about 1.7 hours to rise from the surface to the 10 hPa level of Table 3.2). The levels most commonly reported are mandatory levels and significant levels. A mandatory level is a level reported by raobs from all countries by international agreement. Originally, the mandatory levels were 1000, 850, 700, 500, 400, 300, 200, 150, 100 and 50 hPa. Currently, the mandatory levels include the original mandatory levels as well as 925, 250, 70, 30, 20 and 10 hPa. International transmission codes have specific blocks reserved for data at these levels. Significant levels are levels where there are abrupt changes in the temperature or humidity profiles. (Sometimes, significant levels based upon abrupt changes in wind speed or direction are reported.) Significant levels provide additional information and can be quite useful to weather forecasting offices. In addition to mandatory and significant levels some raob reports regularly include data at 975, 950, 900, 800, 750, 650, 600 and 550 hPa. A typical sounding is presented in Table 3.2.

The unit "hPa" which stands for hecto-Pascal is favored over "millibar" (mb) which is the historically used unit for atmospheric pressure. Either is acceptable in journals. Millibar or mb will be frequently encountered in text books, articles and descriptions of dataset archives. Just remember that 1000 hPa means the same as 1000 mb.

The spatial distribution of raob stations (Fig. 3.5) is considerably less dense than the distribution of surface stations. North America and Europe have the densest reporting networks. Large spatial gaps south of 20 degrees N are evident. The time period spanned by these data records is generally shorter than that of surface stations. The earliest raob data start in the 1940s but large numbers of reporting stations only began in the late 1950s, about the time of the IGY (1957-58).

Some sources of error for raob data are similar to those for surface observations (e.g., calibration error). The homogeneity of raob observations above the surface is less affected by small changes in the location or by changing environment than near surface observations. As with surface observations, metadata are scarce. The measurement of humidity has been particularly suspect. Humidity measurements at low temperatures (e.g., -40 degrees C) are of questionable reliability. Over the U.S. humidities less than 20% were often not recorded and high humidities (~95%) which occur in clouds could not be measured. Table 3.1 lists a chronology of changes in the radiosondes used by the U.S. since 1943 and Figure 3.6 illustrates the effect upon relative humidity measurements of various instrument and sounding changes. Around the world 35--40 different types of radiosondes are being used.

The pressure instrument is accurate to about plus/minus 1-2 hPa. Temperatures are accurate to plus/minus 0.5 degree C up to about 20 km. Relative humidity accuracy is a few percent, except at very low temperatures and at very low or very high humidities. Wind accuracies are about 3 m/s The geopotential height is quite accurate. However, systematic biases in the temperature or humidity measurements can introduce large errors in the estimate of geopotential height. For example, a systematic error of 0.5 degree C in the temperature instrument could lead to an error of 25 meters or so in the 200 hPa geopotential height.

Mean Year-Month Statistics for Raob Data

NCAR has computed mean year-month statistics (i.e., the average of a variable or quantity of the span of one month in a particular year) for a number of raob stations at mandatory levels. This includes not only the mean temperature, wind speed, geopotential height and humidity, but also quantities such as sensible heat transport, momentum transport and eddy quantities. These are contained in NCAR dataset ds391.0.

Other Sources of Upper Air Data

Other common sources of upper air data are aircraft reports (AIREP), pilot-balloon (PIBAL) observations and data derived from satellites (see Chapter 5). An AIREP is code used to transmit aircraft observations to an operational meteorological center. These reports can be particularly useful to forecast centers when they come from locations over the oceans where there are few station observations. A PIBAL is a balloon which is visually tracked while ascending in order to measure upper level winds. It is used to support aircraft operations. No temperature or humidity measurements are made.

Table 3.1*
Chronology of changes in U.S. radiosondes
Date Change
1943 Lithium chloride humidity element replaced hair hygrometer.
1943 Dark ceramic resistance sensor replaced glass-tube electrolytic temperature element.
1948 Relative humidity computed using saturation relative to water instead of ice
1948 Change observing times from 2300, 1100 UTC to 0300, 1500 UTC.
1949 Smaller temperature sensor to reduce response time.
1950 Correction for solar radiation introduced (until 1960).
1957 Change observing times from 0300, 1500 UTC to 0000, 1200 UTC.
1960 Introduced white-coated temperature elements.
1965 Carbon humidity element, began reporting low humidities.
1972 Redesigned humidity ducts to reduce solar effects.
1973 Stopped reporting relative humidity below 20%.
1980 New carbon hygristors, new relative humidity transfer equation.
1988 Precalibrated hygristor replaced type requiring preflight calibration.
1988 New VIZ sonde with new humidity duct.
New Space Data Division (SDD) manufactured radiosonde at some stations. Differences between VIZ and SDD noted.
1993 Relative humidity to be reported over broader range, values <20% and up to 98% (instead of 95%) in cloud.
1993 Gravitational constant used to define geopotential height from geopotential changed from 9.8 to 9.80665 m s .
* From Trenberth, K.E., 1994: Atmospheric Circulation Climate Changes, 31, 427-453.

Table 3.2
Typical Raob Report (NMC decode)
(deg C)
(deg C)
1005.0 99999.0 10.6
1000.0 131.0 10.41.4 999.0 999.0
962.0 99999.0 8.60.3999.0999.0
925.0 99999.0 7.62.5270.0 40.0
858.0 99999.0 2.6 0.0999.0999.0
850.0 1467.0 2.40.0265.0 46.0
841.0 99999.0 2.60.0 999.0999.0
817.0 99999.0 3.211.0999.0999.0
800.0 99999.0 2.25.0999.0999.0
788.0 99999.0 3.29.0 999.0999.0
750.0 99999.0 0.614.0 999.0999.0
700.0 3032.0 -1.512.0265.0 44.0
634.0 99999.0 -5.523.0999.0 999.0
600.0 99999.0 -6.332.0999.0999.0
572.0 99999.0 -9.311.0 999.0999.0
546.0 99999.0 -11.114.0999.0999.0
500.0 5640.0 -16.121.0250.060.0
400.0 7290.0 -27.19.0255.059.0
336.0 99999.0 -37.37.0999.0999.0
300.0 9290.0 -43.59.0255.058.0
291.0 99999.0 -45.110.0 999.0999.0
250.0 10490.0 -52.98.0 260.064.0
245.0 99999.0 -52.910.0999.0999.0
200.0 11900.0 -63.113.0255.064.0
174.0 99999.0 -68.3 12.0999.0999.0
150.0 13660.0 -58.320.0255.032.0
100.0 16210.0 -59.931.0270.017.0
75.4 99999.0-59.5 31.0999.0999.0
70.0 18440.0 -59.531.0275.025.0
55.0 99999.0 -59.531.0999.0999.0
50.0 20540.0 -59.331.0270.016.0
30.0 23750.0 -59.131.0245.015.0
26.0 99999.0 -60.331.0 999.0999.0
20.0 26280.0 -60.731.0 265.010.0
10.0 30610.0 -58.731.0290.033.0
Note: 99999.0 and 999.0 represent missing data.
1 Some form of humidity is reported. These include: relative humidity (%), specific humidity (q) or dew-point depression (TD).

Table 3.3
Some NCAR Datasets Containing Hourly Data
ds470.0USA3001938-92hrly-3hr *L
ds521.0USA2551969-76hrly-3hr *M

Table 3.4
Some datasets containing daily/monthly Tmin, Tmax, Precipitation
Area No.
PeriodType Order Vol
ds463.0Global7,5001967-80DS L
ds473.0Antarctica 1980-91D S
ds473.5Greenland 1987-91D S
ds508.0USA1371881-1985 DTSS
Pacific Is.
 11,000 1888-1993DT L
ds511.0USA 1381880-1987DTM
ds512.0Global 7500 1979-7/93MDTS M
ds516.0Canada7801874-1991 DTL
ds522.0USA372 1888-1976DTS
 Canada1271874-1979D T 
ds523.0 Australia10,0001939-82DTM
ds524.0Russia2231880-1989 DTS
ds560.0USA/PAC IS110001851-presMTM
ds565.0USA1200880-1987MT M
ds825.0Central Eng11659-presMDTS

Table 3.5
NCAR Datasets containing Precipitation only
Area No.
PeriodType Order Vol
ds482.0 Australia 14000 1932-82 DMT M
ds482.1 Australia 191 1840-1990 DMT S
ds484.0 Pacific 200+ 1971-94 DT S
ds485.0 China 180 1951-82 DT M
ds517.0 Brazil 2300 1910-74 D T S
ds578.0 China 90 1950-80 M T M
ds825.0 Cen. England 11659-pres DM T S

Table 3.6
NCAR Datasets containing Monthly Mean Sfc. Temperature (T), Precipitation (P), Sea Level and/or Station Pressure (SLP/STP)
ds564.0Globe(GHCN)7500x x x x 1701-1990
ds570.0Global4400 x x x x 1731-pres
ds571.0Africa1000 - x - - 1880-1973
ds572.0S. America 680 - x - - 1891-1983
ds574.0 USA3900 x x - - 1941-1980
ds575.0India4000 - x - - 1901-1970
ds576.0Canada3900 x x - - 1831-1982
ds577.0 Australia10000 - x - - 1831-1982
ds578.0China90 - x - - 1951-1980
ds578.1China160 x x - - 1951-1990
ds885.1USA x x - - 1895-pres

Table 3.7
Some Datasets with Snow Information at NCAR
Source Region Base
ds315.0 NSIDCNHem1966-91wkly, monly
ds512.0 CACGlobal1979-presdaily, monly
dsCAC (Basist)added in near future

Table 3.8
Gridded Global Precipitation Climatologies, Yr-mo Datasets, etc
Source Grid
ds207.0RAND (Moeller)4 - 5global  no
ds238.0DOE4 - 5land1951-89anomalies
ds290.0 Shea (NCAR)2.5global1950-79st. dev.
ds541.2Petty (COADS)2.5ocean1958-91 
ds768.0Cogley/Briggs1land1950-79st. dev.
ds865.0Jaeger 4 - 5global no

ds701.5MSU (Spencer)2.5ocean1979-pryear-mo
ds728.0GPI (NOAA)2.5plus/minus 381986-pryear-mo
ds728.1CMAP (Xie-Arkin)2.5global1987-6/94year-mo
ds729.0SSMI (Chang)5.0plus/minus 501987-91year-mo

Table 3.9
Gridded Global Surface Temperature Climatologies
Source Grid
ds205.0Crutcher and
5 No Hem
So Hem
ds217.0Oort(GFDL)2.5 - 5global1958-73yes

Table 3.10
NCAR Datasets containing Large Amounts of "Raob" Data
Area Period Freq;
ds353.0Global1962-72D; ST,p,Td,rh,u,v,zL
ds353.4Global1974-presD; S T,p,Td,rh,u,v,zL
ds390.0Global (US ctrl)1948-presD; TT,p,Td,rh,u,v,zL
ds390.1U.S. control1948-pres D; TT,p,Td,rh,u,v,zL
ds390.5Global (US ctrl)1943-74D; TT,p,Td,u,v,zL
ds391.0Global (US ctrl)1948-presM; TT,p,Td,u,v,zM
 derived quan 
ds430.0Global 1950-presM T,p,q,u,v,zS

Surface Observations
Climatologies and Gridded Data of Surface Variables
Upper Air Observations
Mean Year-Month Statistics for Raob Data
Other Sources of Upper Air Data

An Introduction to Atmospheric and Oceanographic Datasets
[Next] [Prev] [Contents] [Top]