COMPARISON OF SUMMER- AND WINTER-TIME SUBURBAN ENERGY FLUXES IN CHRISTCHURCH, NEW ZEALAND

Transcription

1 INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Climatol. 22: (22) Published online 24 May 22 in Wiley InterScience ( DOI: 1.12/joc.767 COMPARISON OF SUMMER- AND WINTER-TIME SUBURBAN ENERGY FLUXES IN CHRISTCHURCH, NEW ZEALAND RACHEL A. SPRONKEN-SMITH* Department of Geography, University of Canterbury, PB 48 Christchurch, New Zealand Received 8 November 2 Revised 7 December 21 Accepted 16 December 21 ABSTRACT Knowledge of the surface energy balance is fundamental to understanding the boundary layer meteorology and climatology of urban areas. This study reports some of the first direct measurements of energy fluxes over the city of Christchurch, New Zealand, during both summer and winter. Observations of the surface energy balance were made over two mainly residential suburbs: St Albans and Beckenham. Net all-wave radiation Q was measured with a net radiometer, the eddy covariance approach was used to measure the turbulent heat fluxes (sensible heat Q H, and latent heat Q E ), and the heat storage flux Q S was estimated as the energy balance residual. During the predominant northeasterlies and unstable conditions in summer, the fetch at St Albans includes a commercial warehouse as well as residential areas. In summer, on a daily basis, Q H is the dominant heat sink followed by Q S and Q E. However, during daytime Q S can be considerable and may approach the magnitude of Q H. Evaporation is low because the turbulent flux source areas are mainly centred over the commercial warehouse and yard, which have little greenspace. In winter the flux source areas are mainly residential for both sites, and the small daily surplus of Q is partitioned mostly into Q S, with some Q E and asmallq H that may be directed either towards or away from the surface depending largely on the synoptic conditions. Under strong inversion conditions, which occur frequently in Christchurch during winter, the turbulent heat fluxes are very small and Q H may be directed towards the surface for many hours overnight and early in the morning. During foehn events the energy partitioning is significantly altered, particularly in winter. Net radiation may be substantially decreased, evaporation is usually markedly increased and in winter Q H may be directed towards the surface for much of the event. The results highlight the importance of seasonal and synoptic controls in energy partitioning at this location, although difficulties with fetch complicate the analysis. Copyright 22 Royal Meteorological Society. KEY WORDS: surface energy balance; heat fluxes; urban climate; foehn; Christchurch 1. INTRODUCTION Knowledge of the surface energy balance is fundamental to understanding the boundary layer meteorology and climatology of urban areas. The energy balance of an urban area can be expressed as: Q + Q F = Q H + Q E + Q S + Q A [W m 2 ] (1) where the flux densities are: Q net all-wave radiation, Q F anthropogenic heat, Q H turbulent sensible heat, Q E turbulent latent heat (or evaporation), Q S sensible heat storage, and Q A heat advection (representing the net gain or loss due to transport associated with the spatial heterogeneity of sources and sinks). This equation is typically applied to a volume that reaches from a measurement height well above the roughness elements (buildings and trees), down to ground level. Although several studies have examined surface energy exchanges of urban areas in the Northern Hemisphere (for a review see Grimmond and Oke (1995)), most focus on summertime energetics with few reports of wintertime observations. This study reports some of the first direct * Correspondence to: Rachel A. Spronken-Smith, Department of Geography, University of Canterbury, PB 48 Christchurch, New Zealand; r.spronken-smith@geog.canterbury.ac.nz Copyright 22 Royal Meteorological Society

2 98 R. A. SPRONKEN-SMITH measurements of energy fluxes over the city of Christchurch, New Zealand, during both summer and winter. The dynamic meteorological environment of Christchurch (which includes common foehn events) provides a useful location to study the influence of synoptic, as well as seasonal, controls on energy partitioning. Christchurch (43 33 S, E) has a population of about 32 and is located on the Canterbury Plains to the east of the Southern Alps (Figure 1). Knowledge of the surface energy balance in this city has two important practical implications. First, during summer the region has a relatively low rainfall, so it is useful to know how weather influences suburban water use (particularly irrigation) and hence the latent heat flux. Second, Christchurch has a serious wintertime air pollution problem (Spronken-Smith et al., 22). In winter, the presence of slow-moving anticyclones, together with local topographic effects, result in the common occurrence of nocturnal surface inversions. On these cold nights there is heavy usage of coal and wood for domestic heating (Canterbury Regional Council, 1997), resulting in serious pollution events. An understanding of the dispersion of air pollution must incorporate the role of the surface energy balance, since an increase in sensible heat at the surface in response to solar heating results in vertical mixing and the growth of the mixed layer. The depth of the mixed layer determines the volume through which pollutants can disperse. In summer, stronger solar heating results in greater sensible heat fluxes and, consequently, the mixed layer is much deeper, providing better conditions for dispersion. However, in winter, under settled anticyclonic conditions, the atmosphere can become quite stable, particularly for the period from sunset through to the following morning. Furthermore, foehn events, which are common in Christchurch, can restrict the growth of the mixed layer, as a layer of warm air can overlie cooler air at the surface. So, an understanding of the surface forcings and their links to synoptic processes will assist the numerical modelling of the dispersion process. The surface energy balance of Christchurch has been the focus of previous study. Tapper et al. (1981) modelled the energy balance of the Christchurch area in an attempt to simulate winter urban heat islands. The model suggested that in the industrial and commercial areas of the city no energy is used for evaporation; 64% is converted to sensible heat and 46% goes into storage. As part of a study on the dispersion of air pollution, van den Assem (1997) made direct measurements of the energy balance over the residential suburb of Beckenham in late August (i.e. late winter), This paper will report the results of both van den Assem s (1997) wintertime study and more recent measurements of summer- and winter-time energy fluxes over a similar residential neighbourhood (St Albans) in Christchurch Urban setting 2. OBSERVATION PROGRAMME Christchurch is known as the Garden City and has an abundance of urban parks and tree-lined streets. The neighbourhoods of St Albans and Beckenham (Figure 1) primarily consist of one- and two-storeyed residential houses interspersed with a few shopping strips. The surface characteristics of the two sites are given in Table I. Both neighbourhoods have a high percentage of pervious materials: 56% in St Albans and 64% in Beckenham. These values are much higher than typical North American suburban areas. For example, Grimmond and Oke (1995) estimate pervious materials in four American suburbs to range from 4 to 44%. Observations were taken from 21 m high instrumented towers at both sites. At St Albans the tower was located to the west of a large commercial warehouse and yard (the only commercial warehouse in an otherwise residential area) over short grass (see Figure 1(b) and the aerial image in Figure 5). This location is not ideal, since the predominant winds are northeasterlies, and thus during these conditions the fetch is a mixture of residential and commercial land use. Under other wind directions, however, the fetch is entirely residential neighbourhoods. The Beckenham tower was located in a residential area at a council yard and was surrounded by deciduous trees to the south, southeast, northeast and north within a distance of 5 m from the tower. Directly around the tower was short grass, and about 3 m west of the tower were stored materials (generally less than 2 m height). Single- or two-storey buildings and abundant greenspace surrounded the council yard (Figure 1(c)).

3 SUBURBAN ENERGY FLUXES 981 (a) Main road N 2 km A Southern Alps 44 S Christchurch SA 173 E B Port Hills (b) (c) Figure 1. (a) Location of Christchurch, New Zealand, with detail showing the location of the St Albans (SA) and Beckenham (B) towers, together with the airport (A, site of long-term data records). (b) View from St Albans tower looking north; the warehouse and yard are clearly seen to the right and the nature of the residential surrounds is shown to the left. The wider neighbourhood is shown in an aerial image in Figure 5. (c) View northeast over the Beckenham neighbourhood from the lower Port Hills. Note that this image is for spring, where as measurements were taken in winter

4 982 R. A. SPRONKEN-SMITH Table I. Surface characteristics of St Albans and Beckenham neighbourhoods within 2 km 2 of the towers. The plan area values are given as percentages of the total area. For location of sites see Figure 1 Characteristic St Albans Beckenham Land use Residential Residential Immediate tower surrounds Grassed area adjacent to large warehouse and concrete and asphalt yard Grassed area adjacent to asphalt storage yard Buildings (%) Roads and pavement (%) Residential greenspace (%) 5 56 Parks and other greenspace (non-residential) (%) 6 8 Total impervious (%) Total pervious (%) Roughness length z (m).42 a.45 b,.6 c Zero-plane displacement z d (m) 2.44 a 4.67 b,4.5 c a Estimated using empirical formulae of Raupach (1994). b Estimated using empirical formula of Kondo and Yamazawa (1986). c Estimated using the mass conservation method of De Bruin and Moore (1985) Climatic setting The climate of Christchurch is strongly influenced by the presence of the Southern Alps to the west and the Pacific Ocean to the east. Situated in the mid-latitude westerly wind belt, the region s weather is characterized by the reasonably regular passage of anticyclones at approximately weekly intervals punctuated by either classical mid-latitude depressions or more frequently trough and cold front systems. Anticyclonic circulation is most frequent in summer, whereas in winter a cyclonic circulation dominates (Sturman et al., 1984). Two mesoscale features are important in Christchurch s climate: the foehn effect and sea breezes. With the Southern Alps approximately 1 km to the west, Christchurch is often subject to the foehn effect prior to the passage of cold fronts over the region. This local nor wester often causes rapid increases in wind speed and temperature and dramatic decreases in relative humidity. During summer, onshore flows (northeasterlies) can result from three different origins: synoptic anticyclonic northeasterly flow, lee-trough-generated onshore northeasterlies and sea breezes, which are typically manifest as northeasterly or east-northeasterly winds (Sturman and Tapper, 1996). Meandaily temperaturesrangefrom 17 C in summer (January)to 6 C in winter (July), with daily maximum temperatures of 22 C in summer and 11 C in winter (Tomlinson and Sansom, 1994). Frosts are common in winter, with about 19 days of ground frosts occurring in July and 16 days in August. The region has a relatively low annual rainfall of 6 7 mm that is fairly evenly distributed, with a tendency for a winter maximum. Monthly rainfall in summer is only about 4 mm, so irrigation of residential gardens and urban greenspace is common Measurements Instruments to measure the surface energy balance were mounted at 2 m on the St Albans tower and 18 m on the Beckenham tower. Estimates of surface roughness z, and zero-plane displacement height z d at St Albans were made using the morphometric methods of Raupach (1994) (see details in Appendix A). At Beckenham, van den Assem (1997) used both an empirical method (Kondo and Yamazawa, 1986) and a mass conservation method (De Bruin and Moore, 1985) to estimate z and z d. The surface roughness estimates for both neighbourhoods are in Table I. Tennekes (1982) approximation for the lower limit of the constant

5 SUBURBAN ENERGY FLUXES 983 flux layer (z l >> z d + z ) yields a lower limit of about 11 m at St Albans and 15 m at Beckenham, so it is probable that turbulent flux measurements from both towers are in the constant flux layer and are thus representative of the local scale (i.e. 1 2 to 1 4 m). However, the level of the physical blending height, above which turbulent mixing is horizontally homogeneous, varies with mixing activity (Schmid, 1997), and thus it is important to calculate source areas for the turbulent fluxes. These source areas were estimated for the St Albans fluxes, using the flux source area model (FSAM) of Schmid (1994) (data were unavailable for these calculations at Beckenham). Net all-wave radiation Q was measured with a Radiation Energy Balance System Q 7.1 net radiometer and the eddy covariance approach was used to measure the turbulent heat fluxes. A Campbell Scientific (CSI) one-dimensional sonic anemometer and fine-wire thermocouple system (CA27) measured fluctuations in vertical wind speed and air temperature, and a CSI krypton hygrometer (KH2) measured fluctuations in absolute humidity. The vertical velocity, air temperature and humidity fluctuations were sampled at 1 Hz and covariances determined over 15 min periods at St Albans and 1 min periods at Beckenham. Standard flux corrections were made for air density and oxygen absorption (Webb et al., 198; Tanner and Greene, 1989). The heat storage flux Q S was estimated as the energy balance residual. This has the inherent problem that any errors in other fluxes accumulate in this term. It does not explicitly include any sources of anthropogenic heat Q F or sources or sinks due to advection Q A from cooler or warmer surfaces upwind. Neither Q F nor Q A were determined, although the measured terms do include their influence, as they are incorporated in the measured and residual fluxes. Tapper et al. (1981) gave a crude estimate of a wintertime Q F for Christchurch of about 4 W m 2, whereas more recent data (Canterbury Regional Council, 1997) suggest an average of around 6 W m 2. These estimates are relatively low in comparison with most Northern Hemisphere cities, but may be due to the relatively low building and population density of the city (Tapper et al., 1981). Advection could be a significant flux at both sites during sea breezes and foehn events, and this should be kept in mind when interpreting Q S. Measurements of air temperature and relative humidity (using a Vaisala HMP35D probe) and wind speed and wind direction (using Vector Instruments, A11ML cup anemometers and WP2/L wind vanes) were made at the towers. Soil moisture was measured at sites adjacent to the St Albans tower using the gravimetric method. In summer, soil moisture ranged from 4% (by weight) in dry grassed areas to 17% over moist grass. In winter, soil moisture ranged from 41 to 74% (average of 57%). In addition, in summer, 5 residences in the vicinity of the St Albans tower were surveyed to determine their patterns of external water use (irrigation of gardens and lawn, roadside berm, car washing, etc.). On average, people spent about 8 min watering their gardens and lawns, three to four times a week. Observations of the energy fluxes at St Albans were made both in summer (from 29 January (Year Day, YD 29) to 18 February (YD 49) in 1996) and winter (from 16 July (YD 197) to 9 August (YD 221) in 1997). Owing to inclement weather conditions and intermittent instrumentation problems (especially problems with thermocouple breakage during gusty foehn events), only 5 days in each of the summer and winter periods have complete daytime data. A further day with foehn conditions in summer is presented, since it has nearly complete daytime data (1 out of 12 daytime hours when Q > ). These case-study days illustrate both seasonal and synoptic forcing of the surface energy balance. In January and February 1996 the temperatures were about normal, but rainfall was much lower (only 4%) than normal. In July and August 1997 the mean daily temperatures were about normal (5 7 C), although mean daily maxima were slightly higher and mean daily minima were slightly lower than normal. Consequently, there were 44 ground frosts compared with the normal 34, although there were 23 air frosts, which is about normal. Rainfall for these winter months was 9% of normal (Penney, 1997). Observations of energy fluxes at Beckenham were made in late winter from 11 to 24 August 1995 (YDs ). Possibly because the dataset was focusing on nocturnal conditions (owing to interest in particulate pollution, which peaks at night in Christchurch), there are only 2 days with complete daily data. The weather in August 1995 was wetter than normal (about 15 mm of rain compared with the 3 year average of 6 mm), but temperatures were about normal, with a mean daily temperature of 7.1 C and 18 ground frosts three more than the norm (Penney, 1997).

6 984 R. A. SPRONKEN-SMITH 3.1. Energy partitioning in summer 3. RESULTS AND DISCUSSION The surface energy balance of the selected days in summer is shown in Figure 2(a). During these days the only precipitation that occurred was.2 mm at 18: on YD 44. Though the synoptic conditions varied throughout these days, the energy partitioning is reasonably consistent, especially for days with northeasterlies (all days except YD 33). Since different synoptic conditions can result in onshore flow (northeasterlies) the similarity of energy partitioning is to be expected, since source areas should be similar. (a) Energy flux density (Wm -2 ) Un, ND Q* Q H Q E Q S Cy, NW NW winds An, N Un, ND An, ND Cy, ND Time (day) (b) Energy flux density (Wm -2 ) Q* Q H Q E Q S -1-2 An, ND Cy, NW NW var winds An, SW SW winds An, ND An, ND 197, , , , , 1997 Time (day) Cy, W An, NW NW var winds E winds 233, , 1995 Figure 2. Surface energy balance for selected days during (a) summer and (b) winter periods. The summer data are for St Albans in January and February 1996, and the winter data are for St Albans in July 1997 and Beckenham in August The synoptic classification and local wind direction are annotated on the graph. Un is unspecified synoptic circulation, Cy is cyclonic; An is anticyclonic; N is northerly, NE is northeasterly, NW is northwesterly, E is easterly, SW is southwesterly and var is variable flow

7 SUBURBAN ENERGY FLUXES 985 The ensemble average of five mainly clear-sky days (there was occasional patchy cloud on several days) shows the slight dominance of Q H over Q S on an hourly basis during the daytime (Figure 3(a)). Q H peaks at solar noon at about 28 W m 2. Q S reaches about 2 W m 2 at 1: and remains at this level until 13:. Thereafter, Q S declines to become negative at 17:. Evaporation is lower than expected, peaking at about 12 W m 2 at solar noon. The daily and daytime (Q > ) summaries of energy fluxes for northeasterly and northwesterly flow are presented in Table II. During northeasterly flow Q H is the dominant energy sink, both on a daytime and daily basis, accounting for 46% and 5% of Q respectively. Q S also acts as a strong sink, using about 32% of Q, whereas Q E uses only 22% of Q on a daytime basis. The synoptic type for YD 33 was a cyclonic northwesterly giving typical foehn conditions with warm temperatures (maximum of 34 C) and low relative humidity (17%), resulting in vapour pressure deficits as high as 4.3 kpa. Owing to the gusty conditions, the thermocouple that was used to measure Q H broke, giving only 1 h of daytime data. Compared with days (a) 6 Energy flux density (Wm -2 ) Q E Q S Q H Q* Time (h, LAT) (b) 6 Energy flux density (Wm -2 ) Q S Q H Q* Q E Time (h, LAT) Figure 3. Ensemble energy balances for mainly clear-sky days during (a) summer (5 day mean) and (b) winter (4 day mean) at St Albans. Standard errors are also shown. Time is given as local apparent time (LAT)

8 986 R. A. SPRONKEN-SMITH with local northeasterly flow, much more net radiation was used in evaporation, resulting in a lower Bowen ratio β of 1.29 (Table II). The 5 day ensemble average ratio Q H /Q (Figure 4(a)) increases to about.45 by 9: and remains at this level until the late afternoon, when it further increases to about unity. Q E /Q is initially higher at.33, probably due to the evaporation of dew and moisture from irrigation, and then decreases to about.2, remaining steady at this level until late afternoon, when it increases in importance. Q S /Q is fairly constant through the daytime at about.3.4, decreasing and becoming negative in the late afternoon. The Bowen ratio increases steadily through the morning, peaking at about 2.3 by 9: and remaining at this level for several hours. The 5% flux source areas for 14: during both northeasterly (YD 3) and northwesterly (YD 33) airflow in summer are shown in Figure 5. Under northeasterly flow and unstable conditions, 5% of the flux source area lies directly over the warehouse and concrete and asphalt yard, and even the 9% flux source area (not shown) only just extends into the residential area. Under northwesterly flow and unstable conditions, the 5% turbulent flux source area is still not reaching far into the residential neighbourhood, although the 9% flux source area (not shown) does extend to nearly 2 m. At other times of the day, when there is less instability, the flux source areas reach further into the surrounding residential neighbourhood. The implications of these different source areas must be carefully considered when interpreting fluxes measured at this site. The summertime surface energy balance has similar features to those observed in Tucson, Arizona, by Grimmond and Oke (1995). Ignoring foehn events in Christchurch, the daytime dominance of Q H (46% of Q ) and Q S (33% of Q ) is comparable to that observed in Tucson (47% and 28% respectively). Evaporation is low in both Christchurch and Tucson (22% and 25% of Q respectively). This similar partitioning is surprising, as the Tucson suburb, although it has a similar mix of housing, has a xeriscape landscape with low-water use vegetation, whereas the mesiscape vegetation in Christchurch is regularly watered in summer Table II. Daily and daytime (Q > ) mean energy fluxes and flux ratios for summer (St Albans, 1996) and winter (St Albans, 1997, and Beckenham, 1995) periods under local northeasterly (NE), northwesterly (NW), northerly (N) and easterly (E) breezes (for synoptic types see Figure 2). All data are for mainly clear-sky days, except for the winter N and NW. The number of days for the averages are given by n and the number of hours in the daytime averaging period (i.e. when Q > ) are also given. Note that for the summer foehn event (i.e. NW) the daytime averaging period is shorter due to the gusty conditions breaking the thermocouple involved in measuring Q H, and only values for Q and Q E are reported for the daily average Flux (MJ M 2 day 1 ) Q H /Q Q E /Q Q S /Q β Q Q H Q E Q S Summer St Albans Daytime (NE, n = 5, 12 h) Daytime (NW, n = 1, 1 h) Daily (NE, n = 5) Daily (NW, n = 1) Winter St Albans Daytime (NE, n = 4, 9 h) Daytime (N, n = 1, 9 h) Daily (NE, n = 4) Beckenham Daytime (E, n = 1, 9 h) Daytime (NW, n = 1, 9 h) Daily (E, n = 1) Daily (NW, n = 1)

9 SUBURBAN ENERGY FLUXES 987 (a) 3 2 Flux ratio Q H /Q* Q E /Q* Q S /Q* β Time (h, LAT) (b) 3 2 Flux ratio Q H /Q* Q E /Q* Q S /Q* β Time (h, LAT) Figure 4. Ensemble flux ratios for (a) summer (5 day mean) and (b) winter (4 day mean) at St Albans (see Section 2.3). The similarity of energy partitioning is probably due to the different source areas for the fluxes at the two locations. Whereas the Tucson site has typically residential flux source areas, at St Albans these areas include a commercial warehouse and yard with little pervious groundcover, and hence low evaporation and high Q H and Q S. Thus the magnitude of Q S is higher than typically found in a mesiscape landscape but lower than that reported for a light industrial site (with only 4% plan-area vegetated) in Vancouver (Grimmond and Oke, 1999) Energy partitioning in winter The time series of the wintertime surface energy balance for both St Albans and Beckenham (Figure 2(b)) shows that energy partitioning in winter is more variable on a daily basis. At this time of year there seems to be a much stronger synoptic control on the energy fluxes. Very little precipitation fell during the selected study days: only.2 mm at 21: on YDs 219 and 221 in 1997 and intermittent light rain on the afternoon of YD 233 in At St Albans, in the middle of winter, under anticyclonic conditions, the turbulent fluxes are very

10 988 R. A. SPRONKEN-SMITH N 1 2 m Figure 5. The 5% turbulent flux source areas at 14: LAT during northeasterly (YD 3) and northwesterly (YD 33) flow at St Albans in summer (solid white lines) and during northeasterly (YD 221) and northerly (YD 198) flow in winter (dashed lines). Note that the flow during the foehn event on YD 198 had a more northerly component (rather than the more typical northwesterly flow) at the tower level. The tower location is indicated by the small circle small and storage dominates the energy balance. In contrast, at Beckenham, in late winter, under anticyclonic conditions, storage only dominates early in the morning, until the turbulent fluxes increase as the nocturnal inversion is eroded. Then Q H dominates through the daytime and is much higher at Beckenham, peaking at about 2 W m 2 compared with only 1 W m 2 at St Albans (Figure 2(b)). Daytime evaporation at Beckenham is nearly three times higher than at St Albans (Table II), with Q E peaking at about 8 W m 2 around solar noon (Figure 2(b)). This higher evaporation is probably a consequence of both higher vapour

11 SUBURBAN ENERGY FLUXES 989 pressure deficits (peaking at 1.85 kpa on YD 233 and.79 kpa on YD 235 compared with peaks in 1997 of.3.7 kpa for non-foehn conditions and 1. kpa for the foehn event) and different source areas. At St Albans the plan area of pervious groundcover is 56%, whereas at Beckenham it is 64% (Table I), and, clearly depending on stability and wind directions, the flux source areas at St Albans may not include much greenspace. The ensemble average of 4 days from St Albans (excluding YD 198, the foehn event) is shown in Figure 3(b). The storage flux dominates during daytime and the turbulent fluxes are very small, with Q H peakingat7wm 2 around solar noon and Q E remaining fairly steady through the middle part of the day at 3 W m 2. There is a strong surface inversion at night; this persists into the morning and, owing to the suppression of turbulence, results in Q H and Q E not increasing until the inversion breaks down at around 1:. Owing to nocturnal inversions, the St Albans daytime and daily mean fluxes and flux ratios (Table II) may be very different. For example, under northeasterly flow, the mean daytime Q H is.9 MJ m 2 day 1,whereas the daily mean is.12 MJ m 2 day 1. Although evaporation is similarly low on both a daytime and daily basis (with average Q E /Q of.1), the ratio Q H /Q ranges from.16 in the daytime to.3 on a daily basis. During the daytime, storage accounts for 74% of Q, whereas on a daily basis it accounts for nearly all of Q (93%). This dominance of storage as an energy sink results largely from the morning period under strongly stable conditions, when turbulent fluxes are very small and it takes time for the convective mixed layer to grow to the tower measurement height. Although there may be considerable uptake of sensible heat by buildings and the surface substrate, it is unlikely to be as large as indicated in Figure 3(b). It is likely that, during this time, there may be advection of heat occurring in the layer below the tower measurement height. This issue of advection in the morning period under strongly stable conditions deserves further research. The dominance of Q S throughout the daytime at St Albans is seen in Figure 4(b). The ratio Q S /Q is particularly high in the morning due to the inversion, which suppresses the turbulent fluxes. The ratio Q H /Q shows a slight increase throughout the daytime, peaking at.31 in the mid afternoon (Figure 4(b)), and is high at night due to a significant downward sensible heat flux. The ratio Q E /Q is low during the daytime, at about.1, and at night there may be significant dewfall. The Bowen ratio is very variable throughout the day (Figure 4(b)). In the morning, under inversion conditions, it is negative and large ( 4.5), due to a downwarddirected Q H. However, once the inversion breaks down it increases to about 2.5, which is remarkably similar to daytime summer values. In contrast to the energy partitioning on anticyclonic clear sky days, warm air advection during foehn events may result in a daytime downward-directed and quite significant Q H (26% of Q at St Albans and 3% of Q at Beckenham) (Table II). Evaporation also increases markedly under these conditions, consuming 28 29% of Q at both sites. Consequently, the magnitude of daytime Q S approaches Q (which is often low due to extensive cloud cover from a nor west arch) and dominates the surface energy balance. The 5% turbulent flux source areas at 14: in St Albans under northerly (YD 198) and northeasterly (YD 221) airflow in winter (Figure 5) reach much further into the residential neighbourhood than during summer, primarily due to much more stable conditions. Although the source areas are still not ideal for northeasterly flow, during northerly flow the measured fluxes are a good representation of the residential neighbourhood. Again, at other times of the day, when conditions are more stable, the source areas will extend further into the residential neighbourhood. The wintertime energy partitioning shows some major differences to that reported by Grimmond (1992) for Vancouver, British Columbia, in winter. Grimmond (1992) found that, during the daytime, 45% of the available energy was channelled into Q E, with 36% into Q H and 19% into Q S. The importance of Q E in the Vancouver study was thought to be due to frequent rain events and small radiant input. This different energy partitioning reflects the very different wintertime synoptic regimes in the two cities. 4. CONCLUSIONS The characteristics of the suburban surface energy balance in the mid-latitude Southern Hemisphere city of Christchurch are broadly similar to those observed in Northern Hemisphere cities at similar latitudes.

12 99 R. A. SPRONKEN-SMITH However, the dataset is rather limited, and problems with surface heterogeneity complicate interpretation of flux data. During the predominant northeasterlies and unstable conditions in summer, the fetch at St Albans includes a commercial warehouse and residential areas. In summer, on a daily basis, Q H is the dominant heat sink followed by Q S and Q E. However, during daytime Q S can be considerable and may approach the magnitude of Q H. Evaporation is low, since the source areas consist of mainly impervious material. In winter there are some similarities in energy partitioning at St Albans and Beckenham. At both sites the small daily surplus of Q is partitioned mainly into Q S, although more so at St Albans (probably due to measurements occurring in mid-winter when conditions are typically more stable with smaller turbulent fluxes). Daytime evaporation is about three times higher at Beckenham than at St Albans, reflecting the higher vapour pressure deficits later in winter, as well as the greater area of greenspace in the Beckenham neighbourhood. At both sites, but particularly at St Albans in mid-winter, Q H is relatively low and may be directed either towards or away from the surface, depending largely on the synoptic conditions. Under strong inversion conditions, which occur frequently in Christchurch during winter, the turbulent heat fluxes are very small and Q H may be directed towards the surface for many hours from sunset through to mid-morning the next day. During these strongly stable conditions, there is an anomalously high storage flux after sunrise. This storage flux could include a substantial low-level (i.e. below measurement height) advective component, and future measurements of wintertime energy exchanges should try to resolve this issue. During foehn events the energy partitioning is significantly altered, particularly in winter. Net radiation may be substantially decreased if the characteristic nor west arch cloud cover is present. Because of the warm and dry foehn wind, evaporation is usually markedly increased, and in winter Q H may be directed towards the surface for much of the event. The results highlight the importance of seasonal and synoptic controls in energy partitioning at this location. In winter, strong inversions have a significant impact on surface energetics, and in both winter and summer the foehn events significantly alter energy partitioning. Further insight into the synoptic and surface controls on energy partitioning could be gained in a longer-term study that characterizes the turbulent flux source areas through time and stratifies the flux data according to the surface characteristics and synoptic conditions. Future research needs to be conducted at a more homogeneous suburban site to elucidate clearly the suburban fluxes, and the role of low-level advection under strongly stable conditions in winter should be further investigated. ACKNOWLEDGEMENTS Special thanks to Stanley van den Assem for data collection and to Sarah Dyer and Megan Pearce for assistance in the field. Thanks also to Tim Nolan for preparation of figures, Greg Lauer and Glenn Waterland for the generation of a surface characteristic database and to Associate Professor Andy Sturman and the referees for comments on the manuscript. Funds to support the project were provided by the University of Canterbury. APPENDIX A: ESTIMATION OF ROUGHNESS CHARACTERISTICS FOR ST ALBANS A.1. Method Raupach (1994) expresses the zero-plane displacement height z d and roughness length z in terms of building height z H and the frontal area index λ F : z d = 1 + exp[ (c d2λ F ).5 ] 1 z H (c d 2λ F ).5 (A.1) where c d is a free parameter and λ F = L y z H ρ el (A.2)

13 SUBURBAN ENERGY FLUXES 991 Table A.I. Aerodynamic characteristics of the St Albans neighbourhood compared with values given by Grimmond and Oke (1999: 1281) for similar low height and density urban neighbourhoods z H (m) z d (m) z O (m) λ F f O = z d /z H f O = z O /z H St Albans Low height and density neighbourhoods where L y is the mean breadth of the roughness elements perpendicular to the wind direction and ρ el is the density (number of roughness elements per unit area). The roughness length can then be calculated from: ( z = 1 z ) ) d exp ( k Uu + ψ h (A.3) z H z H where u U = min (c s + c R λ F ).5, ( u U ) max (A.4) and k is von Karman s constant (.4), U is the horizontal windspeed, u is the friction velocity, ψ h is the roughness sublayer influence and c S and c R are drag coefficients for the substrate surface at height z H in the absence of roughness elements, and of an isolated roughness element mounted on the surface respectively. Raupach (1994) suggests values of c S =.3, c R =.3, (u /U)max =.3, ψ h =.193 and c d = 7.5. To determine z H and λ F a geographic information system (GIS) database describing the surface characteristics of the neighbourhood within 2 km 2 of the tower was compiled following the method of Grimmond and Souch (1994). Although the database has information on plan area of buildings, roads and greenspace (residential and parks), it does not explicitly have data on tree cover, so the roughness estimates are likely to be underestimated. A subset of the database (35 m 2 ) near the tower site was sampled in the field to determine building heights and widths of the side of the buildings that face the northerly winds. A.2. Results The mean building dimensions of St Albans, together with estimates of the zero-plane displacement and roughness length and other nondimensional roughness properties, are given in Table A.I. The mean building height of 6.1 m is comparable to the 5.7 m found for residential neighbourhoods in six North American cities (Grimmond and Oke, 1999). Buildings constitute only 22% of the plan area (Table I). Ideally, the plan area of roughness elements should include trees as well as buildings, but information on trees was not included in the GIS database. Grimmond and Oke (1999) found that the inclusion of trees in residential neighbourhoods of North American cities could increase the plan area of roughness elements by 5 1%. The frontal area index λ F is low at only.85, but again Grimmond and Oke (1999) suggest that this could increase two to four times by including trees. Consequently, the roughness parameters z d and z are probably underestimated at 2.44 m and.42 m respectively. Despite this, these estimates are consistent with values cited by Grimmond and Oke (1999) for a low-density isolated flow neighbourhood, and it is likely that had trees been included in the database the overall roughness classification would have remained the same. REFERENCES Canterbury Regional Council Christchurch inventory of home heating and motor vehicle emissions. Technical Report. Environmental Management Group. Prepared by NIWA, May 1997, Report No. R97/5. De Bruin HAR, Moore CJ Zero plane displacement and roughness length for tall vegetation, derived from a simple mass conservation hypothesis. Boundary-Layer Meteorology 31: Grimmond CSB The suburban energy balance: methodological considerations and results for a mid-latitude west coast city under winter and spring conditions. International Journal of Climatology 12:

14 992 R. A. SPRONKEN-SMITH Grimmond CSB, Oke TR Comparison of heat fluxes from summertime observations in the suburbs of four North American cities. Journal of Applied Meteorology 34: Grimmond CSB, Oke TR Aerodynamic properties of urban areas derived from analysis of surface form. Journal of Applied Meteorology 38: Grimmond CSB, Souch C Surface description for urban climate studies: a GIS based methodology. Geocarto International 9: Kondo J, Yamazawa H Aerodynamic roughness over an inhomogeneous ground surface. Boundary-Layer Meteorology 35: Penney AC Climate database (CLIDB). National Institute of Water and Atmospheric Research (NIWA) Technical Report 4. Raupach MR Simplified expressions for vegetation roughness length and zero-plane displacement as a function of canopy height and area index. Boundary-Layer Meteorology 71: Schmid HP Source areas for scalars and scalar fluxes. Boundary-Layer Meteorology 67: Schmid HP Experimental design for flux measurements: matching scales of observations and fluxes. Agricultural and Forest Meteorology 87: Spronken-Smith RA, Sturman AP, Wilton E. 22. The air pollution problem in Christchurch progress and prospects. Clean Air and Environmental Quality 36: Sturman AP, Tapper NJ The Weather and Climate of Australia and New Zealand. Oxford University Press: Melbourne; 476 pp. Sturman AP, Trewinnard AC, Gorman PA A study of atmospheric circulation over the South Island of New Zealand ( ). Weather and Climate 4: Tanner BD, Greene JP Measurements of sensible heat and water vapour fluxes using eddy correlation methods. Final Report to US Army Dugway Proving Grounds. Tapper NJ, Tyson PD, Owens IF, Hastie WJ Modelling the winter urban heat island over Christchurch, New Zealand. Journal of Applied Meteorology 2: Tennekes H Similarity relations, scaling laws and spectral dynamics. In Atmospheric Turbulence and Air Pollution Modelling, Nieuwstadt FTM, van Dop H (eds). D. Reidel: Dordrecht, Holland; Tomlinson AI, Sansom J Temperature normals for New Zealand for the period 1961 to 199. NIWA Science and Technology Series No. 4, Wellington, 18 pp. Van den Assem S Dispersion of air pollution in the Christchurch area. Unpublished PhD Thesis, University of Canterbury, New Zealand. Webb EK, Pearman GI, Leuning R Correction of flux measurements for density effects due to heat and water vapour transfer. Quarterly Journal of the Royal Meteorological Society 16: 85 1.