Back
A TECHNIQUE TO DETECT MICROCLIMATIC INHOMOGENEITIES IN HISTORICAL TEMPERATURE RECORDS K.E. Runnalls* and T.R. Oke University of British Columbia, Vancouver, Canada Abstract A technique to identify inhomogeneities in historical temperature records caused by microclimatic changes to the surroundings of a climate station is presented. Because site characteristics can have a pronounced influence on the nocturnal cooling process, this technique uses daily maximum and minimum temperatures to construct time series of monthly ‘cooling ratios’ for the test station and a neighbouring reference station. Hurst rescaling is applied to the time series of cooling ratios to aid in the identification of change points, or inflections, which signal microclimatic change in one of the station records. These change points are then compared to documented station history events, provided that sufficient metadata exists. Results for a variety of air temperature records, ranging from rural to urban stations, are presented to illustrate the applicability of the technique. Key words: temperature records, homogeneity, microclimate, Hurst rescaling 1.
INTRODUCTION
A number of homogeneity assessment techniques exist for the purpose of identifying and correcting non-climatic sources of variability in historical temperature records (e.g. see Peterson et al, 1998). Nevertheless, biases caused by microclimatic change around individual climate stations such as the painting of the thermometer screen, minor instrument relocations, vegetation growth or removal, and construction of houses, roads, parking lots, and runways, frequently escape detection. The difficulty in detection arises because, among other things, these kinds of changes typically take the form of gradual trends over time that are virtually indistinguishable from true climatic trends, and because sufficient metadata (station history information) is usually not available to identify if, or when, such inhomogeneities occur in any particular record. Consequently, microclimatic biases are potentially still present in many of the historical temperature records used for climate change analysis. This paper describes a new technique to detect microclimatic biases in historical temperature records by focusing on the role of site-specific microclimatic control on the nocturnal cooling process. Daily maximum and minimum temperatures are used to estimate the magnitude of nocturnal cooling. Then, the test station is compared to a nearby reference station by constructing time series of monthly ‘cooling ratios’. It is argued that the cooling ratio is a particularly sensitive measure of microclimatic differences between neighbouring climate stations for several reasons: firstly, because microclimatic character is best expressed at night in stable conditions; secondly, because larger-scale climatic influences common to both stations are removed by the use of a ratio and, thirdly, because the ratio can be shown to be invariant in the mean with weather variables such as wind and cloud. Hurst rescaling (Outcalt et al., 1997; Hurst, 1951) is applied to the cooling ratio time series to identify change points in the cooling ratio time series, which signal microclimatic change in one or both of the station records. 2. DESCRIPTION OF THE TECHNIQUE 2.1 Definition of the cooling ratio (R) While most homogeneity assessment techniques are applied to maximum and minimum temperature records separately, microclimatic influences on temperatures are known to influence the nocturnal cooling process, and are likely to be best expressed on calm, clear nights when surface cooling is large. Daily cooling magnitude (? T) is estimated from daily maximum and minimum temperatures (Tmax and Tmin respectively) as follows: ? T = Tmax (n-1) - Tmin (n) (1) where n is the day. In order to determine how the microclimatic character affects the temperature records, the cooling ratio, R is calculated for the test station (A), and a neighbouring reference station (B). R A:B = ? T A / ? T B *
(2)
Corresponding author address: K.E. Runnalls, Department of Geography, University of British Columbia, 1984 West Mall, Vancouver, Canada, V6T 1Z2; e-mail
[email protected]
The reference station should be a high quality record, with relatively few sources of inhomogeneities, and have sufficient metadata information to establish this. The reference station must also be subject to the same topographic, local, meso- and synoptic climatic influences. Thus, the distance requirements for selecting neighbouring stations will depend on the complexity of the surroundings. 2.2 Significance of the cooling ratio Use of the cooling ratio has several advantages over the examination of the temperature records directly. Firstly, the ratio eliminates global and regional climate trends common to both series, thereby making it easier to identify microclimatic influences in the record. Furthermore, the ratio is a dimensionless number of magnitude close to unity. Further, the most significant attribute of the cooling ratio is that although between-site temperature differences are strongly dependent on wind and cloud, cooling ratios are approximately constant in the mean for all weather conditions. That is, while temperature differences are virtually eliminated on windy and cloudy nights, the characteristic cooling ratio is preserved under all weather conditions. This can be seen in Figure 1, which shows daily values of R for 1955 as a function of wind and cloud. Wind speed (u in m s-1), cloud amount (m, in tenths) and cloud type (k, a function of cloud height having a value between 0 and 1) are combined into a single weather factor, F W following Oke (1998). F W = (1- km²) u –1/2
(3)
As Figure 1 shows, R is constant in the mean, and therefore independent of the wind and cloud conditions. In Figure 1(a) R is for the pairing of Vancouver International Airport and the rural reference station of Steveston and has a mean value of approximately 0.79, showing the reduction in cooling at the much more developed Airport site. In Figure 1(b), R has a mean value of 1, because in this case involving Steveston both stations have very similar, open, rural surroundings. Clearly, the cooling ratio preserves the relative microclimatic difference between stations, even when the cooling itself is limited by unfavourable weather conditions. Thus, R allows the identification of microclimatic differences between sites, without the need to account for cloud and wind conditions. Since these ancillary weather data are not available at every climate station, the use of the cooling ratio greatly simplifies the analysis and computations that would otherwise be required for the investigation of microclimatic differences. 2.5
(b)
2.5
2 (1955)
2
YVR:Steveston
1.5 1
1.5
1
R
R
Ladner:Steveston(1955)
(a)
0.5
0.5
0
0
0
0.2
0.4
0.6
0.8
1
0
0.2
0.4
0.6
0.8
1
Fw
Fw
Figure 1. Relation between cooling ratio, R and weather, represented by F W, using daily values for the year 1955. (a) Vancouver International Airport (YVR) and Steveston; (b) Ladner and Steveston. The cooling ratio is a characteristic measure of relative microclimate, and is expected to remain constant over time, except when the microclimatic controls at one of the sites change. Change points in a time series of cooling ratios therefore indicate a microclimatic inhomogeneity in at least one of the temperature records. 2.3 Application of Hurst Rescaling to time series of cooling ratios Time series of cooling ratios were analyzed using Hurst rescaling (Hurst, 1951). This technique has been used to identify the presence of physical regime transitions in a wide variety of geophysical time series (Outcalt et al., 1997), and is therefore appropriate for the identification of microclimatic regime transitions in time series of the
cooling ratio. Hurst rescaling involves both the calculation of the Hurst exponent (H), and the transformation of the original time series into a series of cumulative deviations from the mean. Values of H > 0.5 indicate that regime transitions, or significant change points are present in the time series. When the transformed series (usually rescaled by the range of the transformed series, on a scale of 0 to 1) is plotted and visually inspected, the timing of these regime transitions can be determined. For example, a period of above average conditions in the original series is transformed to an ascending trace. Below average conditions transform to a descending trace, and change points or inflections in the transformed series signal regime transitions. In the present context, change points in the transformed series signal changes in the cooling regime, and if sufficient metadata is available for the climate stations tested, then the cause of these change points can be identified. For complete details on the method of Hurst rescaling, see Outcalt et al. (1997). 3. RESULTS 3.1 Minor siting biases Figure 2 shows the cooling ratio time series and the Hurst rescaled series for Edmonton Municipal airport, and Edmonton International Airport. The Municipal Airport has experienced a number of instrument exposure problems, and is also know to be affected by Edmonton’s urban heat island (Hage, 1976), while the International Airport has much more open, rural surroundings. The Hurst exponent for this series was estimated to be 0.76, clearly indicating the presence of significant cooling regime transitions. Symbols on the rescaled trace denote events noted in the station history file, such as small instrument relocations, changes to parking in the vicinity of the instruments, painting of Stevenson screens, and automation of the instruments. Clearly, the correspondence between the inflection points on the Hurst trace and the documented station history events shows the ability of this technique to identify microclimatic inhomogeneities that are typically considered to have a negligible influence on the homogeneity of temperature records, but nonetheless are clearly exerting a bias that is measurable using Hurst rescaled cooling ratios derived from the temperature observations.
0.8
0.6
0.4
0.2
R
Edm. Mun.:Edm. Int
. and rescaled trace
1
0 1961
1967
1973
1979
1985
1991
1997
Year
Figure 2. R Edmonton Municipal : Edmonton International and the long-term mean of the series (upper). Hurst rescaled trace (lower), with symbols denoting documented station history events at Edmonton Municipal (+) and Edmonton International (o). The Hurst exponent for the series is 0.76. 3.2 Homogeneity of urban temperature records While temperature records observed in areas of urban growth are known to contain a warming bias, site-specific changes in the characteristics of urban stations are also potential sources of inhomogeneities. Figure 3 shows the Hurst rescaled trace for Toronto (Bloor Street), Canada, with symbols denoting the timing of several minor instrument relocations. The Hurst exponent of 0.73 clearly indicates the presence of significant change points, and as with the previous example, these change points correspond closely to the documented changes to the station’s surroundings. These results imply that the conventional linear corrections applied to urban temperature records (Hansen et al., 2001; Karl et al., 1988), do not remove all sources of bias, particularly those of microclimatic origin.
Rescaled R
Toronto:Woodbridge
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1948
1960
1972
1984
1996
Year Figure 3. Hurst rescaled trace of Toronto: Woodbridge cooling ratios. Instrument relocations at Toronto denoted by (o); Woodbridge (+). 4.
CONCLUSIONS
The above results indicate that the cooling ratio is a physically relevant and sensitive measure of relative microclimatic differences between neighbouring climate stations. Hurst rescaling of cooling ratios has been shown to successfully detect microclimatic inhomogeneities that conventional techniques typically do not identify. Furthermore, these results point to the larger issue of potential biases still present in temperature records used for climate change analysis. Reconsideration and re-evaluation of the homogeneity of historical climate data bases is warranted. Because of the physical relevance of the cooling ratio approach, and the apparent skill of detecting subtle biases, this technique may also be able to be adapted to become a correction scheme for temperature records although this is not necessarily recommended. The technique may also be useful in analysis of other climate variables besides air temperature.. References Hage, K.D., 1972: Nocturnal Temperatures in Edmonton, Alberta. Journal of Applied Meteorology, 11, 123-129. Hansen, J. R. Ruedy, M. Sata, M. Imhoff, W. Lawrence, D. Easterling, T. Peterson and T. Karl. 2001. A closer look at United States and global surface temperature change. Journal of Geophysical Research, 106, D20. 23, 947-23,963. Hurst, H.E. 1951: Long-term storage capacity of reservoirs. Transactions of the American Society of Civil Engineers, 116, 770-808. Karl, T.R., H.F. Diaz, and G. Kukla 1988: Urbanization: Its detection and effect in the United States climate record. Journal of Climate, 1, 1099-1123. Oke, T.R., 1998: An algorithmic scheme to estimate hourly heat island magnitude, in Preprints, American Meteorological Society 2nd Urban Environment Symposium, Albuquerque, NM, November 2-6, 1998, 8083. Outcalt, S., K.M. Hinkel, E.Meyer, A.J. Brazel, 1997: Application of Hurst rescaling to geophysical serial data. Geographical Analysis, 29, 72-87. Peterson, T.C., D.R. Easterling, T.R. Karl, P. Groisman, N. Plummer, N. Nicholls, S. Torok, I. Auer, R. Boehm, D. Gullett, L. Vincent, R. Heino, H. Tuomenvirta, O. Mestre, T. Szentimrey, J. Salinger, E. J. Forland, I. Hanssen-Bauer, H. Alexandersson, P. Jones, D. Parker, 1998: Homogeneity adjustments of in situ atmospheric climate data: a review. International Journal of Climatology, 18, 1493-1517.
