Urban heat island (UHI) is a widely investigated phenomenon in the field of urban climate characterized by warming of urban areas relative to its surrounding rural environs. Being able to understand the mechanism behind the UHI formation of a city in relation to global climate change can be very useful especially when planning for mitigation and adaptation strategies. However, the lack of UHI studies, especially long-term, of many cities such as in developing countries makes it difficult to generalize the mechanism for UHI formation. Thus, there is a growing demand for studies that focus on the simultaneous analyses of UHI and its trends throughout the world.
On the other hand, there is a long-standing history of constructing global surface temperature datasets to understand the global climate since the late 1800s by Köppen (IPCC Fourth Assessment Report, 2007). Since then, datasets have been constructed in grid formats with spacing ranging from 0.5 to 5.0 latitudinal and longitudinal degrees. During its construction, homogeneity adjustments were conducted to temperature measurements that do not necessarily represent the regional grid or can potentially contaminate it (Hansen et al., 2010). This led to the common reduction of weighted contribution (Rohde et al., 2013) or filtering out of temperature observations at urban areas prior to the averaging of all temperature readings within the grid.
Given the need to quantify the urban effect to temperature for multiple urban areas and the unavoidable dampening of its influence to the construction of global surface temperature datasets, we propose a rapid yet uniform method to analyze the UHI globally through the aid of a global surface temperature dataset.
II. Methodology
To estimate the UHI, we utilized the source data and the gridded outputs of the Berkeley Earth Surface Temperature (BEST) Dataset (berkeleyearth.org; Rohde et al., 2013). The advantage of BEST is that the absolute temperatures of the grids are provided unlike other datasets that provide only anomalies. The reduced weighting of urban temperatures’ influence to the gridded outputs (spacing: 1.0°~100 km) due to homogeneity adjustments implies that gridded surface temperatures represent temperatures of the “rural” environment. The vast quality-controlled station data readily-accessible and compiled for the construction of BEST comprises monthly statistics of mean daily average temperature (Tavg), mean daily minimum temperature (Tmin), and mean daily maximum temperature (Tmax) which includes “urban” areas. Urban stations were collected based on the population exceeding 50,000 within a buffer region of 10 km in diameter from the LandScan 2013™ High Resolution global Population Dataset. Next, the trends of monthly temperatures from urban stations with monthly samples greater than 500 from the period of 1960 – 2009 were estimated for both the urban station and its encompassing BEST grid in terms of °C/century. Here, we did not distinguish between seasons or amount of precipitation. Urban stations were screened further by accepting trends falling within an acceptable P-value less than 0.05. Subtracting the trends of Tmin, Tmax, and Tavg between the assumed urban stations and its surrounding rural grids results to 3 trends of UHI; UHITmin, UHITmax, and UHITavg. Using ArcGIS and Python, around 200 UHI trends coming from various urban stations spread throughout the world were calculated.
III. Summary of Results and Recommendation
The largest difference in trends on average between urban and rural were found in the Tmin which represent nighttime temperatures, followed by Tmin, and then Tmax, which likewise represents daytime peak temperatures. UHIT was found to increase as urban areas are situated farther from the equator, such as in the mid-latitudes, where UHIT were found to be largest, especially for UHITmin. The urban stations can be grouped in terms of their climate zones (tropical, dry, temperate, and continental; see attached figure). Excluding continental climates due to the lack in station count, UHIT in dry climates were found to be largest in both UHITmin and UHITavg. On the other hand, least UHIT was found in tropical climates. UHITmin was also found to quadratically increase with the diurnal temperature range (DTR), approximated by the difference in the average of all monthly maximum and corresponding minimum temperatures of the urban station being analyzed. Dry (arid) climates tend to have a wider DTR which is more intense at the rural areas due to the higher thermal inertia at urban areas which leads to larger UHI intensities. Finally, UHITmin was found to logarithmically decrease with increasing wind speed (average of monthly wind speeds from ERA Interim 2001 – 2010; see attached figure) for cities located in temperate and dry climate regimes. However, wind speed dependence of UHIT was not seen in cities at tropical zones possibly due to the large synoptic influence (e.g. monsoonal winds following rainfall and intertropical convergence generating lines of cumulus clouds).
The large spread of UHIT from the analyses was due to the local, micro-scale condition surround the station, currently not included in the analyses. Nevertheless, the evidence of UHI as well as the trends’ dependence to global-scale indicators such as climate regimes and other synoptic-scale factors such as wind speed or DRT was found.
From this study, we recognized the need for more temperature observations in both urban and rural areas, consideration of surrounding local condition (sky-view factor, building morphology) of the urban stations in the analyses, as well as efforts to possibly reproduce the findings of this study in numerical weather modelling after overcoming the problem of realistic parameter distribution for the surface boundary at urban areas (global urban climatology).
Acknowledgment: This research was supported by the Environment Research and Technology Development Fund (S-14) of the Ministry of the Environment, Japan.
IV. References
Hansen J, Ruedy R, Sato M, Lo K. 2010. Global surface temperature change. Reviews of Geophysics 48: RG4004.
Rohde R, Muller R, Jacobsen R, Perlmutter S, Rosenfeld A, Wurtele J, Curry J, Wickham C, Mosher S. 2013. Berkeley Earth Temperature Averaging Process. Geoinfor Geostat: An Overview 1:2
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner