Modelling Heat Islands - Parameters and Considerations

The Urban Heat Island, similar to most atmospheric phenomenon develops under a set of patterned conditions. It’s highly unlikely that the UHI would form randomly, but the main challenge associated with the UHI is that the conditions for formation are not easily quantifiable. 

In order to model a phenomenon, a thorough understanding of the characteristics of formation is required as the parameters that need to be incorporated into the model are highly dependant on them. Additionally, the conditions for mitigation or UHI decay also need to be understood as they should not be overlooked. Essentially, conditions for formation and depletion are two sides of the same coin, just sitting on opposite extremes. 

To date, the most important meteorological characteristics that have been determined to influence the UHI (positively or negatively) include:

• Wind speed
• Cloud coverage
• Relative Humidity (RH)
• Water vapour content (WVC)
• Atmospheric pressure ~ More specifically, Anticyclonic conditions

The reason WVC and RH are treated as two separate meteorological conditions is because there is a considerable difference in how the UHI responds to the actual vapour content of the atmosphere and simply how close the system is to saturation, irrespective of the vapour required for it to reach saturation. Regardless of the vapour content, the UHI appears to respond more strongly to how close the system is to saturation. This has not been tested in very warm regions and hence the effect of a very large water vapour content under average RH is still uncertain, which is why it is being considered.

It may seem that these conditions differ from Arnfield’s suggestions for UHI formation and that is particularly because Arnfield dealt primarily with optimising UHI formation whereas in the case of models, we are dealing with all parameters that have been shown to exhibit some form of influence on the UHI. 

Nocturnal UHI Statistical Model in Hamburg (Hoffmann et al., 2011)

A good example of a model developed to model the nocturnal UHI in relation to meteorological conditions is that of Hoffmann et al. (2011) in Hamburg. They worked to model the UHI whereby the parameters would be identified through the use of a generalised least squares method to form a statistical model. To identify the significant meteorological variables to include in their model, a regression based statistical model was developed to test each meteorological condition (X) under a 2-tailed t-test. The explained variance of R^2 was then used to test the strength of the relationship.

 ∆Tu−r =  aX + b

Note: ∆Tu−r refers to UHI intensity (Urban - Rural)

They concluded the most influential conditions were:

• Wind speed (Negative correlation) ~ This was discussed in a previous post (2 posts ago)
• Cloud coverage (Negative correlation)
• Relative Humidity (RH) (Positive correlation)

Wind was found to replace the air within the system, hence weakening the UHI. Cloud coverage prevented escaping heat from cooling the rural region, which then lowered the contrast between the urban and rural stations. High relative humidity meant more vapour would condense in the upper atmosphere releasing latent heat.

They decided to exclude the effects of WVC and Atmospheric pressure due to their observed weak influence on system as part of the study.

• Water vapour content (WVC) (0.2%)
• Atmospheric pressure (7.6%)

The linear regression model was then constructed:

∆Tu−r = aFF + bCC + cRH + d

Whereby:

FF = Windspeed

CC = Cloud Coverage

RH = Relative Humidity

a, b, c, d are fixed parameters to be determined through the least squares method.

The datasets used with respect to the meteorological observations were obtained from the German Meteorological Service (DWD), ERA40 and Regional Climate Model (RCM) results.

The figure below illustrates a frequency distribution of the UHI observed within Hamburg vs values calculated through the model. It was suggested within the paper that the overestimation of UHI was likely due to overestimations in the cloud coverage and relative humidity datasets that were obtained. Additionally, calibrations would need to be made to filter some of the biases in the system. Another factor to consider if applying a model such as this to future climates is the effect of increases in unstable climatic conditions that would rise due to warmer climates.

Figure 1: A frequency distribution plot of observed (Black Asterisk) and modelled UHI intensity using measurements (Black points) and ERA40 (Grey points) for the period of 1985 - 1999. Error bars indicate 95% confidence intervals due to unexplained variance. 

I highly recommend giving this paper a read for a more in depth understanding of the model’s design, capabilities and application. It should be noted that this is a city-scale model, which is ideal for contribution to climate models when appropriate. UHIs also form at smaller scales, each with their own set of approaches and challenges, which will be discussed in the next post. 


Probably...