Determination of The Intensity and Georeferencing of Urban Heat Islands in Temuco, Chile

Various methodologies to estimate the intensity of urban heat islands (UHI) have been proposed. However, there is no consensus on their combined or individual use. This study presents and analyzes the validation of the combined use of methodologies to capture data from UHI’a (mobile transects and xed stations) and the location of the UHI phenomenon by considering the city of Temuco, Chile as a case study. The database used for the xed station methodology consists of twenty-three stations belonging to the National Monitoring Network of Chile distributed throughout the city. For the mobile transect methodology the database was generated with three cars that crossed the city simultaneously taking one temperature record per minute. The results show that the two methodologies yield similar results, so it is possible to obtain the intensity of the IHU phenomenon using at least one of them. The results also showed that Temuco has a maximum intensity of the UHI phenomenon of 13°C, with this value being one of the highest recorded in UHI studies in the world. The UHI-related images make possible the visualization of the phenomenon at a certain time. For this, isotherms were deployed that link stations records with equal temperatures, using Delaunay’s triangulation as the theoretical base 51 , i.e.: (a) All the points must be connected to each other and form the greatest number of possible triangles without their corners overlapping. (b) The triangles are dened by joining the points closest to each other. (c) The formed triangles must be as regular as possible, maximizing smaller angles and minimizing the length of the sides 51 . Each isotherm is drawn with a different color, associating red with high temperatures and blue with low temperatures. The uctuation plot corresponds to the daily visualization (24 h) of the station temperatures that recorded the highest and lowest temperature input during the phenomenon. In turn, the results of the temperature by the ocial Chilean weather station located in the Maquehue sector are visualizing the behavior Temuco by making a point outside the city.


Introduction
The development of urban centers and their expansion towards the peripheral areas have directly affected variations in the weather patterns of cities 1 . The transformation of natural surfaces that served to dissipate the energy radiating on surfaces and that could store and emit energy has produced a temperature increase in cities 2, 3 . This phenomenon is called urban heat islands (UHI), and their intensity is de ned as the temperature difference observed between urban zones and rural or adjacent sectors at a given moment 4,5,6 .
Increased power consumption for air conditioning to compensate for the high temperatures in buildings within the UHI generates polluting gases like carbon dioxide and nitrogen oxide, which are considered precursors of the ozone accelerating the formation of smog 7 . This directly contributes to global warming, forming a vicious circle 8 . According to the Intergovernmental Panel on Climate Change (IPCC), between 1983 and 2012 an increase in global temperature occurred 9, 10 that ranged between 0.6°C and 0.8°C, and it is expected that for the year 2100 the temperature will increase between 1.4°C and 5.8°C compared to 1990 temperatures 11 . A particular case is the city of Kuwait: since 1972 their temperatures have increased by 0.3°C and 0.8°C per year, with temperatures over 50°C becoming more frequent 12 .
The increase in global temperature has brought with it sustained periods of high temperatures called heat waves. During the presence of heat waves the UHI phenomenon is ampli ed, altering people's comfort and health 6,13 . In Europe in 2003, around 70,000 people died due to a heat wave 14 . In Russia and India in 2010 and 2015 respectively, more than 20,500 people died 14 , victims of heat waves, the consequences of which are augmented in UHI 14 .
UHI can be perceived in any season of the year and are visualized in both the day and night 15 . A study in Iași, Romania reported that the temperature difference between the city center and adjacent sectors varied between 3.0 and 4.0°C at night 16 . In Mexico City it was determined that the temperature difference in the day varies between 3.0 and 5.0°C in the rainy season 17 . In Akure, Nigeria during the dry season, the temperature difference between the rural and urban sectors was 4.4°C at 21:00 h 18 . In Toledo, Ohio there was a 2.0°C difference between the city center and its adjacent sectors throughout the year 19 . In 2002 it was determined that in Phoenix, Arizona, the temperature difference between the urban and rural sectors varied between 0.8°C and 5.4°C 20 . From this it is deduced that the UHI phenomenon is present in various cities regardless of their geographic location.
Several methodologies are used to determine the intensity of the UHI, including data capture using mobile transects, xed stations and satellite analyses 21 .
Data capture using mobile transects has been employed for the design of urban connection networks for more than 40 years 22 . It is prioritized when xed stations are insu cient to create an accurate representation of the phenomenon 22 . However, its measurement period is limited compared to the other methods. Martínez (2014) used this method to determine the intensity of the heat islands in the city of Alicante and concluded that the maximum intensity of the phenomenon was 4.5°C 23 . However, the measurement of the intensity only represents a static moment in the day, which does not visualize UHI behavior throughout the day.
The evaluation of UHI using xed stations is suitable for measuring temperature near the surface 17 .
Giannopoulou (2011) used this method to determine the characteristics of UHI in Athens in summer, concluding that the intensity of the phenomenon in the city reached 5°C 24 . However, this methodology can produce distorted results when there is a limited number of stations 25 . Erell and Williamson (2007) also mention that the xed networks are more expensive and prone to vandalism, which limits their acquisition and conservation in large numbers 26 . Stewart (2011) indicates that the satellite analyses provide more accurate results, taking measurements for long periods of time 27 . However, Konstantinov (2015) notes that the temperatures obtained via satellite are not always accurate, determining an error of measurement of 1.68°C in the city de Apatity 28 . Hardy and Nel (2015) used this method to determine the intensity of the UHI in Johannesburg, concluding that the temperature difference between the rural and urban sectors varied between 2°C and 3.5°C 29 .
However, this methodology does not analyze the air temperature in the city, but rather it is in uenced by the heat absorbed by the urban infrastructure and is detected by satellite images.
Although the intensity of the UHI phenomenon has been determined in different cities around the world, no comparison or validation of the results using combined methodologies has been found. This raises the need to evaluate simultaneously the phenomenon of atmospheric UHI in the same city with the different existing methodologies to determine the correlation of the results that these methodologies provide and have the certainty that their individual or combined use is both reliable and viable 30,31,32 .
Studies into the measurements of the intensity of the UHI phenomenon have been conducted in several countries. However, these studies have been carried out in countries whose boundary conditions, for example the typology of the houses, vegetation or climate are not extremely different. The case studies where the UHI can be substantive and the boundary conditions are extremely different are still pending. In this sense, countries like Chile represent a particular case in terms of UHI since the climate and vegetation of the cities vary according to their geographic location. In Chile studies have been conducted measuring the intensity of the UHI phenomenon by mobile transect and satellite images a 1,3,33,30 . Nevertheless, these studies are concentrated in the central zone, which is not representative due to the country's length and climatic variability.
The aim of this study is to measure and analyze the UHI intensity in Temuco, Chile using the xed station method contrasted with the mobile transect method. The objectives of this study are the validation of the combined use of methodologies in the data collection and georeferencing of the location of UHI in the city, the thermal imbalance of which affects people's quality of life.

Study Area
The city of Temuco is located at latitude -38.735932 and longitude -72.590409. It has a surface of 464 km 2 and is delimited by the Ñielol and Conunhueno hills with heights of 335 m a.s.l. and 360 m a.s.l. respectively. The Cautín River is located between the hills and crosses the city northeast to southwest.
According to the United Nations (UN) 42  This population growth has brought with it the construction of a large number of residential neighborhoods. This in turn entails the replacement of areas with vegetation for elements with the capacity to absorb the heat, resulting in an increase in temperatures in the city 45 .
Temuco is classi ed according to the Köppen climate classi cation as Csb (warm-summer Mediterranean climate). It is characterized as having cold winters and dry summers; mean temperatures exceed 10°C and its maximum and minimum temperatures range between 23°C and 1°C respectively 46 .
The temperature records from previous years have been taken by the Meteorological O ce of Chile using an o cial station located in a zone bordering the city in the Maquehue sector (Coord.: -38.76778, -72.59417). According to the temperature data recorded at the Maquehue station, 6.65% of the days between 2000-2019 reached maximum temperatures over 30°C, reaching 40.2°C in 2019 (February 15 -21:00 h). In the previous period between 1980-2000, only reached 3.2% of the days had maximum temperatures over 30°C; thus, the temperatures over 30°C have doubled in the last two decades. These increases in temperature even entail the annual average temperatures being higher than those stipulated in the Köppen classi cation. In Temuco the maximum daily temperatures appear in summer, with February being the month with approximately 50% of the highest temperatures, followed by January with 43.75% and March with 6.25%.

Procedure to contrast heat island location methodologies
It has been determined in the existing literature that UHI can be classi ed in two types: surface urban heat islands (UHI's) and atmospheric urban heat islands (UHI'a) 47 . The former is understood as the thermal difference between arti cial and natural surfaces, while the latter is understood as the difference in air temperature between the different zones in the city 47 . The methodologies used to determine the intensity of UHI vary according to their classi cation. Satellite image analysis is used to determine the intensity of UHI's, whereas the mobile transect or xed station methods can be used to determine the intensity of UHI'a 31 .
In the present study UHI'a were analyzed, so the data needed to evaluate their intensity were captured using mobile transects and xed stations. Both methodologies were based on the capture of air temperature, which made it possible to contrast both databases and calculate their Pearson's correlation coe cient, the interpretation of which is presented in Table 1. Additionally, to contrast and compare the two methodologies, the average difference, maximum temperature difference recorded by the two methodologies and the standard deviation between the two data sets were used. The method performed with mobile transects consists of roaming data collection through the city along a previously de ned route 49 . The temperatures of the cities were recorded each minute by two HOBO Pendant MX2201 sensors and a HOBO Pendant MX2202 sensor. These sensors have a rapid response to thermal variations and their measurement ranges cover from -20°C to 70°C with an accuracy of ±0.5°C.
Prior to the data collection, the three sensors were calibrated by capturing air temperatures with direct radiation and low shade. This measurement was used to verify that the temperature records captured by the three sensors were similar, with a maximum variation of ±0.15°C among them.
The sensors were positioned on the passenger side window of three cars that crossed the city simultaneously at an average speed of 30 km/h. For each data capture the corresponding coordinates were taken via GPS. The measurements were taken between 20:00 and 20:30 since in this period there is no in uence of direct solar radiation and the elements of the city begin to release their heat absorbed during the day. The data are downloaded via Bluetooth on the free platform HOBOmobile, available for Android, IOS and Windows.
To determine the route to take with the mobile transects, the UHI phenomenon was analyzed based on the data recorded by ReNaM in Temuco on November 27, 2017 and January 19, 2018. On these days, the highest annual temperatures were recorded, reaching 28.5°C and 31.4°C respectively. The high temperatures on these days generated UHI in four zones (Z-1, Z-2, Z-3 and Z-4) in the city ( Figure 7). The methodology used to locate the UHI at this stage is presented in chapter 2.3. It must be mentioned that for the determination of the mobile transect route a previous simulation of the UHI is necessarily required.
This helps to guarantee that the thermal pro le generated by the mobile transect more accurately represents the phenomenon. Based on the location of heat islands from previous years, the decision was made to perform three routes simultaneously ( Figure 8). The Transect 1 route (red) had a total distance of 10.50 km, crossing the city from north to south by the main avenue (Av. Caupolicán) passing through Z-1 and Z-2. The Transect 2 route (magenta) crossed the zones Z-3 and Z-4 including a total distance of 10.35 km from the western sector of the city in a northeast direction. The Transect 3 route (green) had a total distance of 10.73 km, initially crossing the city from north to south and then edging along the Cautín river in an east-west direction crossing Z-2. Additionally, it was de ned that the Transect 2 route begins where the Transect 3 ends in order determine if there was a temperature reduction rate over the period when the measurements were taken. It was ensured that all the trajectories of the mobile transects were within the isothermal map created based on the xed stations (chapter 2.3).

Fixed Stations Method
To conduct this study using the xed stations method, temperatures measured by the National Monitoring Network (ReNaM in Spanish) were used. The network has 23 intelligent sensors (Netatmo) represented in Figure 8 by black dots; these are installed in private homes distributed in various zones in Temuco.
The Netatmo weather stations are comprised of two devices (interior-exterior) of UV-resistant aluminum with a measurement range from -40°C to 65°C with an accuracy of ±0.3°C. The sensors installed on the outside are protected from rain and direct solar radiation to avoid deteriorations and a better accuracy in capturing data. The time interval with which the data are captured is thirty minutes, con gured in closed schedules so as not to have variations in the capture time, for example: 8:00-8:30-9:00 etc. The sensors were calibrated and validated by the Ministry of Housing and Urban Development (MINVU) in conjunction with the Chile Foundation.
Prior to the measurement of the intensity of the UHI with both methodologies, the temperature data from the HOBO and Netatmo sensors were contrasted. It was determined that the average variation between the sensors was 0.48°C. This implies that the possible temperature variations obtained with both methodologies at the same geographic point would not be heavily in uenced by the different devices used for the measurement.

Methodologies for UHI simulation and location in Temuco
The methodological framework used to produce images of urban zones with UHI and a uctuation plot that compares the maximum and minimum temperature in the city is based on the classic methodology of data export adapted for the UHI simulation context 50 . This methodology consists of (1) Analysis and data processing, (2) Methodology for data processing, (3) Sampling, dimensionality reduction and use of distance functions.  (Table 2). Finally, the image generation process is described using the previously mentioned data set. In the step, two types of images were generated: (1) Heat islands, and (2) Fluctuation plot. The UHI-related images make possible the visualization of the phenomenon at a certain time. The intensity of the phenomenon for every period of time was obtained using the equation: UHI = TUmax -TUoRmin. TUmax corresponds to the maximum urban temperature captured in the city, whereas TUoRmin corresponds to the minimum urban or rural temperature recorded in the city or its adjacent sectors. The intensity of the UHI phenomenon for Temuco was classi ed according to Table 3. Table 3 Classi cation and temperature ranges associated with UHI intensity based on 52 .

Classi cation Intensity Interpretation
Weak UHI ≤ 2°C There is no major difference between the temperature in the city and adjacent sectors There is a slight difference between the temperature in the city and adjacent sectors not perceptible by people The difference in temperature between the city and adjacent sectors is moderately perceptible by people There is a big difference in temperature between the city and adjacent sectors very perceptible by people Extremely strong 8°C < UHI The difference in temperature between the city and adjacent sectors is extreme and dangerous for people The stations used for the xed-station measurements record data every thirty minutes, whereas the sensors used in mobile transects record data every minute. Therefore, for the generation of temperature data with the same frequency as the mobile transects (one minute), the temperature data of each xed station were interpolated for thirty minutes.

Results
Comparison of xed station and mobile transect methodologies Figures 3, 4 and 5 illustrate a comparison of the temperature records obtained using the mobile transect and xed station methodologies. These records were compared to determine how close the values are that these methodologies provide and to be able to validate their use separately or combined.
Transect 1 covered the route represented in red in Figure 2 of 10.5 km, where temperature data were captured every one minute. The route took thirty minutes, where thirty points were captured along the route. Figure 1 provides the graphical representation of the temperatures recorded by the transect contrasted with the temperatures obtained by isothermal maps given by the xed station methodology. It is observed that both methodologies yield similar temperature records. From record N°1 to record N°12 there is an increase in temperature, after which there is a decrease until record N°15. The similarity between the records obtained with the two methodologies is con rmed with the correlation coe cient of 0.55.
The greatest maximum temperature difference between the two methodologies at the same point reached 1.76°C, located at the moment at which the city is entered, where the change in temperature is considerable. However, the average temperature difference captured by the two methodologies along the entire route was 0.72°C, which implies that the classi cation of the UHI phenomenon is not altered by the points that have the greatest temperature difference. The standard deviation between the two temperature data sets is 0.5. This value is very close to the accuracy of the sensor, which implies that the dispersion of the data is not signi cant.
Transect 2 performed the route represented by magenta in Figure 2 which took 26 thirty-six minutes. The circuit covered 10.35 km, where 36 temperature records were captured. It might be considered that unlike Transect 1 the route ran through sectors with high tra c congestion, which acquired a larger amount of temperature data in a shorter distance. In Figure 2(a) it is possible to visualize with the xed station methodology that there are four zones in the city with high temperatures. These zones are not clearly re ected in the mobile transect methodology. However, it is noted that the intensity of both temperature records decrease until record N°29, when they begin to increase their intensity again. The similarity between the values is con rmed with the correlation coe cient of 0.61, a value classi ed as mean correlation.
The greatest maximum temperature difference captured by both methodologies at the same point was 2.91°C, whereas the average temperature difference captured by both methodologies at the same point was 1.6°C, a lower value than the ranges for which the UHI is de ned, which does not alter its classi cation. The standard deviation that yields this temperature data set is 0.65. This value indicates that despite analyzing the intensity of the phenomenon with the most distant record, its classi cation would not change.
Transect 3 performed the route represented in green in Figure 2 of 10.73 km, where data were captured for 30 minutes, creating a record of 30 points. Although the distance was greater than the other transects, the sector encompassed has few tra c lights, which makes the tra c steady and ow better. Figure 2(b) shows that the similarity in the two records is greater than Transects 1 and 2. With both methodologies it is observed that the journey goes between zones in the city that have a lower temperature to zones with a higher temperature. Figure 2(b) shows that both temperature intensities of the zones covered begin to decrease until record N°8, the point where the temperatures in the zones begin to increase. This similarity that the records have in the increase and decrease in their intensity in similar ranges is con rmed by the correlation coe cient of 0.62. This value is classi ed as mean correlation.
The greatest maximum temperature difference captured was 2.7°C, whereas the average temperature difference captured by both methodologies at the same point was 1.2°C, a lower value than the ranges for which the UHI intensity is de ned that does not alter its classi cation. The standard deviation obtained from this temperature data set is 0.91, a range in which the data tend to disperse.
Location of heat islands in Temuco Figure 3 shows the isothermal map (left) and uctuation plot (right) generated at the time and in the sector where the city reaches its maximum temperature.  (Figure 3, right) also shows that the maximum temperature in the city (Stn. 288) increases and decreases in short periods of time. This may be due to the fact that in the zone where the temperature peak is reached, there is abundant vegetation and most of the infrastructures are isolated houses, which allows for rapid air circulation and therefore reduces the temperature in the sector.
In Figure 3 it is observed that in addition to the zone with the maximum temperature (Z-3), there are two sectors that tend to have a higher temperature than the rest of the city (Z-1 and Z-4).

Discussion
The results of the comparison of the temperature records obtained using the mobile transect and xed station methodologies show similar UHI intensities in Temuco. The correlation coe cients of the three circuits varied between 0.55 and 0.62, being classi ed as a moderate correlation. It may be inferred from this that both methodologies can analyze UHI intensity, obtaining similar classi cations. This would also imply that both methodologies deliver reliable results used either individually or complementarily.
The time to perform the three routes with the mobile transects was approximately thirty minutes. The transect 2 route began at point -38.756761,-72.646664, where a temperature of 20.68°C was recorded. On the other hand, the transect 3 route included at the end of its run the temperature record at the same point as the start of transect 2. The temperature recorded at the same geographical point, but with a 30-minute delay for transect 3, was 20.80°C. This indicates that the temperature of the city in the time period when the measurements were taken (thirty minutes) using mobile transects had no signi cant variations. However, it must be considered that, were the run time to be extended, there could be a temperature variation in the city, which would involve making time corrections to the recorded values.
It has been noted that the xed station methodology is expensive, and the stations are subject to vandalism 26 . Therefore, in terms of installation and maintenance costs, using mobile transects makes more sense. Nevertheless, it must be considered that the mobile transect methodology can classify the UHI intensity, but only considering thermal pro les. By contrast, the xed station methodology permits an overall analysis of the UHI phenomenon in the city. Here it is also inferred that for the choice of routes for the mobile transects, it is essential to know beforehand the location of the zones where the UHI are located, which is why the xed station methodology would be a suitable complement.
The temperature difference of each measurement point in Temuco and the control station located in Maquehue on December 4, 2019 at 14:00 h ranged between 0.2°C and 13°C. This demonstrates that the temperature in the entire city is higher than that of the neighboring rural sectors. The maximum intensity of the UHI phenomenon in Temuco was 13°C, which is why it is classi ed as extremely strong (Table 3). This value was recorded when the temperature at the Maquehue control point was 25.4°C.
Using a linear projection it is expected that in heat wave scenarios (temperatures higher than 30°C for more than 3 consecutive days), the temperatures of the zones identi ed as UHI in Temuco exceed 43°C. In Figure 5, a temperature comparison appears for two heat wave episodes (1 and 2) that occurred in February 2020 between the Maquehue station (MS1 and MS2) and the zones with the highest temperature in Temuco (Z-1 to Z-4 in Figure 8). Here, it is observed rst that the maximum temperature in Temuco occurs between 15:00 and 16:00 h, which represents a behavior that does not follow the common pattern of the UHI phenomenon. The maximum temperature peak in most of the cases studied was reached between 19:00 and 20:00 h. However, in the case of Temuco at that time the temperature in the city dropped, whereas outside the city (Maquehue) the temperature reached its daily maximum.
Additionally, the maximum temperatures recorded in the city had an average increase of 26°C in a short period of time (between 10 and 15 h). The same speed of temperature change appeared from 16 h where it reached on average 5.3 °C per hour. This is higher than the maximum acceptable temperature change rate (3 °C/h) so that the human body does not suddenly feel hot or cold. This suggests the need to conduct more detailed studies on the behavior of the phenomenon in Temuco. Analyzing from the materiality of constructions, direction, etc., several factors may explain more precisely the behavior of the UHI phenomenon in Temuco.
In the temperature pro les in Figure 5, there is a high correlation between the temperature recorded at the Maquehue station and the temperature change in the city. Thus, using linear regressions, per hour equations were determined to describe the temperature pro le of the zones with the highest temperature in the city (Z1, Z2, Z3 and Z4) based on the temperature at Maquehue. Temperature pro les were modeled for a third heat wave that occurred in March 2020 and using an example, the real temperature and the modeled one in zone 3 (Z-3) were compared from the temperature at the Maquehue station ( Figure 6). Figure 6 shows that the values obtained from the modeling are very close to the real temperature pro les that occurred during that same heat wave event. Here, it was determined that the average real and modeled temperature differences in the 72-hour period reached 1°C and the correlation coe cient was 0.98. This strong correlation suggests it is possible to determine the temperature in the different zones of the city from the temperature of the weather station located outside the city (Maquehue). This would also imply that it is possible to know the past pro les and the prediction for the future of heat wave events based on climate change scenarios. Nevertheless, additional studies are needed to verify these hypotheses and in that sense they represent a continuation of the work presented here.
It has been studied that the UHI phenomenon has direct in uence on the power consumption of buildings in cities 35-38 . Hinkel (2003) and Silva (2014) determined UHI occur not only in high-temperature seasons, but it is also possible to detect them in low-temperature seasons 39,22 . Stewart and Oke (2012) determined that in low-temperature seasons UHI can cause a drop in power consumption for heating 40 .
Therefore, it is estimated that the occurrence of UHI in winter in Temuco contributes to the reduced power requirement for heating. However, Capelli (2005) maintain that in high-temperature seasons, there is an increase in energy demand for air conditioning 41 , which is why an equalization in total energy demand is also expected.

Conclusions
In this work the combined use of methodologies to capture temperature records was compared and contrasted and in addition the urban heat islands (UHI) of Temuco, Chile were simulated and located.    Simulation of the UHI phenomenon in Temuco using xed station methodology (a) and uctuation plot (b) on December 4, 2019 between sectors that have the maximum and minimum temperature at 14:00 h. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.  Comparison of temperature for two heat wave episodes between Maquehue station (MS1 and MS2) and the zones with the highest temperature in Temuco (Z1, Z2, Z3, Z4).

Figure 6
Comparison of real and modeled temperature pro les for zone 3 in Temuco based on the temperatures of the Maquehue station (MS) located outside the city. designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 8
Route to capture data with the mobile transect method (red, magenta and green lines) and xed stations distributed in different points of the city (black points). Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.