1. Introduction

Surface-based air temperature measurements are an essential component of many meteorological field programs. Temperature measurements are usually taken with dataloggers that sample a temperature sensor at fixed time intervals and either store the data on location for later analysis or transmit the data to a central repository via radio or telephone links. Improvements in electronics have recently resulted in the production of lower-cost, miniaturized temperature dataloggers. Because larger numbers of dataloggers are now affordable, this development may lead to opportunities to improve the spatial sampling of air temperatures in meteorological applications.

Micro- and mesoscale climate studies may benefit greatly from this technology. Examples of applications include the mapping of temperature for frost investigations in orchards or vineyards; the determination of complex terrain influences on air temperature; and the measurement of microclimatic temperature variations in caves, vegetation canopies, or other special environments. Some of the new dataloggers are small enough to be attached to small mammals, birds, or other fauna to determine the microclimates to which they are exposed in their living spaces. The development of temperature dataloggers has been accelerated somewhat in recent years by government regulations regarding the handling and transport of food products. Of special concern is the monitoring of the temperature–time history of refrigerated shipments of foods, flowers, and other temperature sensitive goods. This paper documents the characteristics of one such datalogger, the HOBO H8 Pro, developed by Onset Computer of Bourne, Massachusetts (Fig. 1), that has been weatherproofed for outdoor use and is thus suitable for extended meteorological investigations.

2. The temperature datalogger

The HOBO H8 Pro circuitry is housed in a plastic case that is sealed from moisture by two rubber O-rings;one is at the base of the plastic case, and the other is on a threaded plug that is removed from the face of the case by the user to upload instructions to the logger and to download data. The version of the logger that we tested had an internal temperature sensor located inside the plastic case and an external thermistor in a 0.5 cm by 3.0 cm epoxy-potted cylinder at the end of a 2-m electrical lead. The manufacturer’s specifications for the logger are provided in Table 1.

The datalogger is started (launched) by removing the threaded plug and connecting an interface cable between a data port within the logger case and the serial port of a computer. A computer program provided by the manufacturer is then run, and the user selects datalogger launch instructions including the logging interval, the time at which data collection should begin, the channels to be logged, and the resolution desired (8 or 12 bits). The datalogger uses the computer’s clock to automatically set its initial time. The logger carries a unique serial number in its software. To facilitate identification of the logging location, the user may enter an additional logger name. The logger is then disconnected from the computer and installed at the field location. Data can be downloaded at any time from the logger using a computer and the same software and cable. A small handheld data-downloading and storage device (HOBO Shuttle) that facilitates the field collection of data from multiple dataloggers is also available from the manufacturer. The shuttle is connected to the datalogger via an interface cable through the data port, and a single button is pushed to download the data, reset the logger clock, check the logger’s battery, and relaunch the logger. The 468 kbytes of data storage in the shuttle is sufficient for storing data from seven full dataloggers.

3. Testing of the temperature datalogger

We performed four tests of the datalogger and shuttle. First, the time constant of the sensor/datalogger was determined in still air. Second, temperature accuracy was determined by comparison with a platinum thermometer in a stirred water bath. Third, an overnight field test was performed to compare the temperatures recorded by multiple loggers in an outdoor setting. Finally, the time accuracy of the shuttle was determined in laboratory tests. The results of the tests are provided below.

4. Field experience

Field experience was gained from a network of 15 temperature dataloggers operated on the sides of two mountains in the Columbia Basin near Richland, Washington. Eleven of the loggers were placed on a line that ran up the north face of Rattlesnake Mountain (43.394°N latitude, 119.595°W longitude, 1111 m MSL) and continued up the east ridge of the mountain to the summit. This line of loggers, placed over an elevation range from 133 to 1111 m MSL, was operated from 16 December 1998 to 30 March 1999. The experiments were designed to determine if “vertical” profiles of temperature obtained from networks of surface-based dataloggers on mountainsides could be used to indicate the presence of temperature inversions and persistent cold air pools in the Columbia Basin. The datalogger temperature profile was to be compared with 10-min average virtual temperature profiles made each half hour with a nearby Radio Acoustic Sounding System. The results of the intercomparison will be published in a separate paper; here we concentrate on temperature datalogger performance.

Temperature sensors must be shielded from solar and terrestrial radiation to provide acceptable air temperature data. Thus, for our field experiments, the external thermistors were placed inside six-plate radiation shields, and the dataloggers were secured to the mounting brackets at the foot of the radiation shield to make a self-contained datalogger/shield assembly. The assemblies were secured to steel fence posts so that the thermistors were exposed approximately 1 m above the ground. Data were logged at 5-min intervals. The radiation shields used in our experiments were designed by Gill (1983) and manufactured by the R. M. Young Company of Traverse City, Michigan (Model 41301-5). The design uses 12-cm-diameter plates made of UV-stabilized white thermoplastic. The plates are stacked to cover the sensor completely, shading it from direct and reflected solar radiation, upward and downward longwave radiation, and precipitation. The design comes in several standard configurations including 6-, 9-, and 12-plate shields. The number of plates is chosen based on the size of the temperature sensor assembly that must fit in the shield. The root-mean-square (rms) temperature excess in this unaspirated shield, as reported by the manufacturer for solar radiation of 1080 W m⁻², is 0.4°C in 3 m s⁻¹ winds, 0.7°C in 2 m s⁻¹ winds, and 1.5°C in 1 m s⁻¹ winds. These and similar shields have been extensively tested (e.g., Fuchs and Tanner 1965; Gill 1983; Payne 1987; Crescenti et al. 1989; Andersson and Mattisson 1991; Anderson and Baumgartner 1998). Andersson and Mattisson (1991) performed a nearly 1-yr-long comparison of 1-min-average temperatures obtained with high-quality platinum resistance thermometers exposed in a 12-plate R. M. Young radiation shield and in an adjacent continuously aspirated Teledyne radiation screen. Temperatures differences were greatest during calm conditions under clear skies. The maximum 1-min-average error in the entire period of record was 3.36°C. Temperature differences never exceeded 1°C for longer than 1 h; and June, December, and annual rms errors were between 0.1° and 0.2°C. Anderson and Baumgartner’s (1998) field tests with a nine-plate radiation shield found a maximum instantaneous radiative overheating of 3.4°C and mean daytime errors of 0.27°C. Because of shield geometry, the maximum heating errors occurred for sun elevations of approximately 45°.

The dataloggers performed well during the experiment, providing an uninterrupted temperature record for each station, except for a 2-week data loss at one logger whose sensor cable was severed by an inquisitive elk or deer. Figure 4 presents data from sites located at various elevations on the side of Rattlesnake Mountain. In Fig. 4a time series of 5-min-temperature data are presented for a subset of six sites for the period from 19 to 23 December 1998. Pseudoprofiles obtained from the Rattlesnake Mountain dataloggers by averaging the 5-min data over 1-h intervals are shown for two periods in Figs. 4b,c. Figure 4b is an example of a near-neutral profile during a high wind night when mean winds on the summit of Rattlesnake Mountain varied between 19 and 31 m s⁻¹. Figure 4c is an example of a set of 3-hourly temperature profiles on a clear night with weak ambient winds. Shown are hour-averaged soundings at 2100, 0000, 0300, and 0600 LST. These profiles use all 11 of the Rattlesnake Mountain dataloggers (as does Fig. 4b) to illustrate the development of a nocturnal temperature inversion.

5. Conclusions

Inexpensive temperature dataloggers are now becoming available for applications in meteorology. We have investigated the characteristics of one such datalogger and have reported our operating experience. The loggers are suitable for meteorological measurements in situations in which the data do not require real-time delivery to a central location (as with radio or telephone links) or an onboard display of data (although data from this particular logger can be displayed in real time if a portable computer is connected to the logger). Because the loggers are inexpensive, large numbers of dataloggers might be purchased to facilitate detailed studies of the spatial structure of temperature fields. They are small, yet provide acceptable accuracy and sufficient storage for many meteorological and climatological investigations. In our field investigations, the datalogger temperature sensors were shielded from solar and terrestrial radiation in unaspirated radiation shields. Other investigators have reported on the magnitude of temperature errors in such shields, which can reach several degrees Celsius under intense insolation when wind speeds are low. The dataloggers record instantaneous temperatures at the selected sampling times but have no built-in capability for obtaining average temperatures from multiple samples (e.g., hour averages from 5-min samples). Such averages can, however, be performed on archived data after data collection if the sampling rate is high enough. The relatively long time constant (122 s) of the presently available external thermistor provides an averaging of its own. Sampling at rates shorter than this response time provides little additional information, even though allowable by the datalogger software.