Air temperature evaluation

London Weather Centre (LWC)

Figure 5.4 shows the comparison for 48 hours (6th – 7th August 1998) of the air temperature for the lowest model grid level of the urbanised (METRAS+BEP) simulation and the traditional simulation with measurements from the LWC station. It is reiterated here that the LWC is situated in a highly urbanised grid cell, with an urban fraction of 95%. The model was initialised at 23:00 on the 5th August so the first two time steps were ignored in the comparison as they could be affected by model spin up.

0 5 10 15 20 25 30 35 00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 Time A ir t e mp eratur e ( o C) LWC METRAS BEP+METRAS

Figure 5.4: Diurnal cycle of air temperature (ºC) at the LWC site from August 6th _{to August 7}th ₁₉₉₈

according to the measurements at LWC (blue), the METRAS traditional simulation (pink) and the simulation with BEP (yellow).

The traditional simulation underestimates the maximum daytime temperature by 5.8 ºC, and also underestimates the night time minimum temperature by 1.3 ºC. During daytime the urbanised (METRAS+BEP) simulation shows a better agreement with the measurements, compared to the traditional simulation, but daily maximum temperatures are still underestimated by 3 ºC. This discrepancy may well be due to the fact the LWC measurement is made close to a roof top surface, which is likely to be in direct sunlight and characterised by a different heat capacity and albedo compared to the general city wide characteristics. Therefore in light winds and cloudless skies it is expected to be a much hotter surface during daytime (WMO 2006).

The comparison between the METRAS+BEP simulation and measurements is excellent between the night time hours of 22:00 and 06:00 compared to the traditional simulation due to the fact that during night time the traditional simulation, which does not compute the

radiation trapping in the street canyon, cools more than the BEP simulation. This was also found for the validation of BEP with data from the BUBBLE campaign in Roulet et al. (2005). Trusilova et al. (2008) also found, after implementing an urban canopy scheme into MM5, that the comparison with observations was best during the hours between 21:00 and 06:00.

In the morning hours the METRAS+BEP simulation heats up rather more rapidly than the measurements. This could be due to an increase in cloud cover which occurs at this time (06:00, 7th August), which is seen in the hourly cloud cover data for LWC but is not represented in the model due to the cloud subroutines being turned off.

The comparison for the near surface air temperatures is expected to be improved by the implementation of the BEP scheme, since it takes into account sources of energy in the urban area which the traditional approach neglects, such as the mechanisms of radiation trapping and shadowing in the canyon, and the basic anthropogenic heat treatment.

0 5 10 15 20 25 30 35 00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 Time Ai r temperature ( o C) LWC METRAS METRAS+BEP

Figure 5.5: Diurnal cycle of air temperature (ºC) at the LWC site from July 30th _{to July 31}st ₁₉₉₉

according to the measurements at LWC (blue), the METRAS traditional simulation (pink) and the simulation with BEP (yellow).

Figure 5.5 shows the air temperature comparison for the second detailed case study, the period of 30th-31st July 1999. This period was selected because it was used by Best (2005) to test the capability of the UK Met Office operational mesoscale model to reproduce expected urban phenomena. This second case study confirms the improvement in performance of the METRAS+BEP model for a highly urbanised location. The METRAS+BEP simulation shows a better comparison with LWC measurements during both daytime and night time hours. Again a cold bias is found for the day time peak temperature, and the METRAS+BEP model heats up too rapidly on the second day of simulation (31st July 1999). Interestingly for this period of simulation the results for the UK Met Office operational mesoscale model also showed that the air warms too quickly at dawn (Best 2005), giving a warm bias over this period. Neither the METRAS+BEP model, nor the traditional METRAS simulation fully represents the temperature variability during

the second day of simulation (31st July 1999), however this is likely to be due to an increase in cloud cover variability, which is not represented in either simulation.

London Heathrow Airport (LHR)

Figure 5.6 shows the same comparison for air temperature of the two simulations with the LHR station. 0 5 10 15 20 25 30 35 00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 Time Ai r temperature ( oC) LHR METRAS METRAS+BEP

Figure 5.6: Diurnal cycle of air temperature (ºC) at the LHR site from August 6th _{to August 7}th ₁₉₉₈

according to the measurements at LHR (blue), the traditional simulation (pink) and the simulation with BEP (yellow).

The LHR station is situated in a suburban model grid cell, with an urban fraction of 56%. The traditional METRAS simulation underestimates the daytime maximum temperature by 5.9 ºC, as well as failing to reproduce the timing of the peak temperature, but shows a good agreement with the measurements during night time. The METRAS+BEP simulation performs better than the traditional one during daytime, although maximum temperatures are still underestimated. During night time the METRAS+BEP simulation overestimates

the minimum temperature by 2.6 ºC. An explanation of the better night time temperature agreement between the measurements and the traditional simulation is that the local characteristics of the airport weather station (an extensive area of open flat concrete and grass with low albedo, and the additional influence of high heat fluxes due to transportation) are better represented by the roughness approach in the original METRAS model, rather than the urban scheme with its vertically distributed impact of the buildings. In this case the LHR surface station would not capture the average characteristics of the whole grid cell, and the comparison would be affected by the local surface characteristics. For the LHR site the METRAS+BEP simulation shows better agreement in terms of the increase in air temperature for the morning of 7th August than the urban LWC station. Figure 5.7 shows the comparison of the air temperature at LHR for the second case study, the period of 30th-31st July 1999. This corroborates the findings for the first case study. Again during daytime the METRAS+BEP simulation performs better than the traditional METRAS simulation, although there is still a cold bias in the peak daytime temperature. The minimum night time temperature is overestimated in both simulations.

0 5 10 15 20 25 30 35 00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 Time Air temp erat ure ( o C) LHR METRAS METRAS+BEP

Figure 5.7: Diurnal cycle of air temperature (ºC) at the LHR site from July 30th _{to July 31}st ₁₉₉₉

according to the LHR measurements (blue), the traditional METRAS simulation (pink) and the simulation with BEP (yellow).

St James’ Park (SJP) and Bracknell-Beaufort Park (BBP)

Data at SJP and BBP was only available for comparison for the first case study (6th-7th August 1998). Figure 5.8 shows the air temperature comparison for the SJP urban park site. Again the traditional simulation underestimates daytime maximum temperatures, and fails to capture the timing of the maximum, but represents night time minimum temperatures better than the METRAS+BEP simulation. The METRAS+BEP simulation also underestimates daytime maximum temperatures but does capture the timing of the peak temperature.

0 5 10 15 20 25 30 00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 Time Air temp erat ure ( oC) SJP METRAS METRAS+BEP

Figure 5.8: Diurnal cycle of air temperature (ºC) at the SJP site from August 6th _{to August 7}th ₁₉₉₈

according to the SJP measurements (blue), the traditional simulation (pink) and the simulation with BEP (yellow).

Figure 5.9 shows the air temperature comparison for the BBP site. The implementation of the urban canopy scheme should have no direct influence on a completely rural site located upstream of the city (Otte et al. 2004), however as this site is located outside the domain the measurements are compared with the closest grid cell, which has an urban fraction of 20% and is located close to the domain boundary. Differences between the measurements and the model results might also be affected by advection from nearby urban areas. Compared to LWC, LHR and SJP this is the grid cell with lowest urbanised fraction, and for this reason the difference between the traditional and METRAS+BEP air temperatures is much smaller as expected. For both simulations the comparison for the first 6 hours is excellent which probably reflects the fact that out of the four stations used for the validation the BBP station is closest to the Larkhill station whose surface data was used to initialise

the model. The smaller percentage of the urbanised fraction compared to the other locations determines the lack of spread of the results in the initial hours. An overestimation of nocturnal temperatures of up to 2 ºC for rural sites for simulations with and without BEP is also observed by Martilli (2003) for a validation of BEP for the city of Athens.

0 5 10 15 20 25 30 00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 08:00 12:00 16:00 20:00 00:00 04:00 Time Air temp erat ure ( oC) BBP METRAS METRAS+BEP

Figure 5.9: Diurnal cycle of air temperature (ºC) at the BBP site from August 6th _{to August 7}th ₁₉₉₈

according to the BBP measurements (blue), the traditional METRAS simulation (pink) and the simulation with BEP (yellow).