EFFECT OF WEATHER STATION LOGGING INTERVAL ON THE PRECISION OF DEGREE-DAY ESTIMATES

The relationship between data-logging intervals and degree-day estimat es was examined to determine the longest interval giving equivalent in formation to estimates based on 12-min intervals and, so, the most eff icient interval for estimation of degree-days. Emphasis was placed on users of inexpensive, unsophisticated loggers with short-term (<3-5 ye ars) rather than long-term temperature-monitoring needs. The study was based on 3 years (1990-92) of 12-min, spot-measured temperature recor ds from Somersby, New South Wales, obtained using a discrete mode data logger. Longer log intervals were simulated in 2 ways. Firstly, the sp ot measures of temperature obtained at 12-min intervals were used to d erive intervals of 36 min, and 1, 2, 3, 4, 6, and 12 h, by excluding 1 2-min readings not corresponding to these intervals. Essentially, the data derived are those that would have been obtained if the discrete m ode logger was originally set to these intervals. Secondly, as a way o f approximating data collected from a logger operating in continuous m ode, temperature data recorded at 36 min, and 1, 2, 3, 4, 6, 12, and 2 4 h, were estimated by averaging 12-min temperature readings over thes e intervals. For both spot and averaged data, degree-days were estimat ed by integrating the area under the daily temperature curve using eac h of the log intervals as the time-step. Estimates were also based on each of 7 lower developmental thresholds ranging from 10 to 13 degrees C. Our analysis showed that estimates of degree-days using spot data and, to a lesser extent, averaged data were insensitive to changes in log interval and lower threshold. For both spot and averaged temperatu re data, intervals up to 6 h did not give significantly different dail y degree-day estimates from those based on 12-min intervals. However, when degree-days were estimated using averaged data, there was a signi ficant dependence between log interval and season: the sensitivity of degree-day estimates to changes in log interval was highest in winter and lowest in summer. This indicates that selection of the most effici ent log interval was seasonally dependent: shorter intervals should be selected for data collection concentrated at the cooler times of the year than at the warmer times of the year. This also implies that sele ction of the most efficient log interval is dependent on geographic va riations in climate. Precise estimates (<+/-5% difference from degree- days based on 12-min intervals) of degree-days based on averaged data were obtained at Somersby using 4-h log intervals in summer, autumn, a nd spring, and 2-h log intervals in winter. For degree-days accumulate d over a full year, 2-h log intervals gave very precise estimates that differed by only 1-2 days from accumulated degree-days based on 12-mi n intervals, while 4-h intervals gave differences of about 4-8 days. F or spot-measured estimates of degree-days, in contrast, no season x lo g interval effect was found. Consequently, there was no seasonal and, by implication, geographic dependency to log interval selection within the range of the data. Annual accumulated degree-day estimates based on 6-h log intervals differed by <8 days compared with those based on 12-min intervals. The annual accumulated difference was up to about 1 day for log intervals up to 3 h. We discuss how recommended log interv als may be modified depending on the aims of a study, its precision re quirements, and as a result of seasonal and geographic variations in t emperature regime. We examine why degree-day estimates are affected by log interval duration for spot and averaged data and consider the imp lications of our results for the application of the degree-day concept and interpretation of previous research. Ways of overcoming equipment limitations (small data storage capabilities, high data retrieval cos ts, low battery life) other than through manipulating logging interval s are examined. Finally, we argue that for the estimation of degree-da ys, operating a datalogger in discrete mode operation has considerable advantages over continuous mode operation.