Bridging accuracy and efficiency: Advancing mean radiant temperature measurement in Urban Ecology

Bridging accuracy and efficiency: Advancing mean radiant temperature measurement in Urban Ecology | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Bridging accuracy and efficiency: Advancing mean radiant temperature measurement in Urban Ecology Aubrey Benson, Ben R. Crawford, John M. Frank, Ariane Middel, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6432428/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 16 Oct, 2025 Read the published version in International Journal of Biometeorology → Version 1 posted 4 You are reading this latest preprint version Abstract Extreme summertime heat is an increasing challenge for cities, highlighting the need for accurate, spatially meaningful methods to measure and map heat in ways that reflect human thermal experiences and inform land management decisions. Mean radiant temperature ( T mrt ) is a key metric for assessing urban heat at hyper-local scales, yet its measurement remains technically challenging. In this study, we apply the six-directional, gold standard method for measuring T mrt with globe thermometer-based approaches across multiple levels of spatial aggregation and develop a novel machine learning model trained on field data. Data were collected in a semi-arid city in Colorado, USA, over two summers. Using measurements from residential parcels, we show that aggregated globe thermometer data—collected using a low-cost, accessible sensor—can capture thermal patterns across landscapes with reasonable accuracy. Our findings also indicate that machine learning, combining six-directional and globe thermometer data, offers promising potential for improving both measurement accuracy and efficiency. These findings are particularly relevant for planners working at the scale of parcels, where heat adaptation strategies are commonly applied, and especially insightful for semi-arid cities and those increasingly experiencing arid summer conditions due to climate change. This work advances practical methods for integrating human thermal comfort into landscape planning for climate-resilient urban design. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Extreme weather and shifting climate regimes are increasingly impacting cities, challenging practitioners to understand how humans, flora, and fauna are experiencing temperature as they live and move across urban landscapes. Mean radiant temperature ( T mrt )—the weighted total of all incoming shortwave and longwave radiation for a reference individual or object—is emerging as a key tool for capturing hyper-local thermal dynamics (ISO 7726 1998 ). T mrt is especially useful for detecting and measuring fine-scale spatial and thermal heterogeneity, providing valuable information about how humans and other organisms experience heat in a city (Middel et al. 2019; 2021; Turner et al. 2022 ). This information can help inform strategies that promote human thermal comfort, manage heat exposure, and shape climate-responsive urban design. Despite its growing importance, obtaining T mrt in the field remains technically challenging. Scientists often rely on radiation modeling software like RayMan (Matzarakis et al. 2007 ; 2010 ), SOLWEIG (Lindberg et al. 2008 ), TUF-Pedestrian (Lachapelle et al. 2022 ), and ENVI-met (Bruse and Fleer 1998 ), which estimate T mrt based on assumptions about urban geometry and biophysical processes. While these models offer useful estimates of radiative loading, they must be validated by on-the-ground measurements to ensure accurate microclimate assessments across diverse urban landscapes (Crank et al. 2020 ). Despite advancements in T mrt measurement methods, the most accurate and reliable methods remain bulky, costly, and time-consuming to build (Rykaczewski et al. 2024 ). Thus, scientists and practitioners are challenged to determine when less resource-intensive methods suffice for their research and management objectives and how to implement them effectively. The highest cost, most accurate method for measuring T mrt to date is often referred to as the six-directional method, integral radiation measurement, or gold standard approach- referred to hereafter as the six-directional method. This approach requires measuring shortwave and longwave radiation upward, downward, and in four orthogonal horizontal directions to estimate the average incoming radiation for a typical standing human (Höppe 1992; Thorsson et al. 2007 ; Middel and Krayenhoff 2019 ; Aleksandrowicz and Pearlmutter, 2023 ; Cheung et al. 2024). The setup typically involves mounting 12 radiometers on two extended arms at 1.1 m height, making them rather difficult to build, maneuver, and mass-produce. Six-directional T mrt measurements have been shown to vary drastically with changes to ground and shade cover (Turner et al. 2022 ) and be strong predictors of spatio-temporal variation in human thermal comfort (Middel and Krayenhoff 2019 ), making the approach a valuable method for estimating the heat load on the human body. A popular alternative to the six-directional approach is the globe thermometer method, which estimates T mrt using air temperature ( T a ), wind speed ( \(\:{V}_{a}\) ), and globe temperature ( T g )- the internal temperature of a small globe sensor. This widely used method is easier to implement, but the conventional formula for obtaining T mrt (ISO 7726 1998 ) using the globe thermometer method yields results that differ from the six-directional approach. Discrepancies arise because T g relies on convection, which, in turn depends on accurate \(\:{V}_{a}\) measurements to correct for. Due to these inaccuracies, researchers have sought to fine-tune the globe method equation to varying environmental conditions and globe sensor types by developing alternative formulas (Lin and Matzarakis 2011 ; Chen et al. 2014 ; Vanos et al. 2021 ; Ouyang et al. 2022 ; Liu et al. 2022 ). All methods are developed under specific environmental conditions and may or may not apply to other geographic or temporal settings. Effective climate adaptation and urban heat mitigation depend on cities’ ability to measure and monitor thermal dynamics at hyper-local scales that reflect the human experience. To better understand how people experience temperature at the ground level and how thermal conditions affect urban trees and other non-human organisms, scientists must find more accessible measurement techniques that provide accurate results. This study addresses the need for practical and ecologically informative approaches to T mrt measurement that can support planning and policy decisions. Using a residential neighborhood in a semi-arid city as a test case, we evaluate the performance of low-cost globe thermometer readings (from a Kestrel 5400) compared to the six-directional, gold standard method (using a mobile biometeorological system called MaRTa). We used different globe thermometer conversion techniques from the literature and a novel machine learning approach. Specifically, we aim to 1) compare multiple Kestrel T mrt methods to MaRTa across different levels of spatial aggregation, 2) assess how varying radiation components influence discrepancies between MaRTa and Kestrel T mrt values for each of the methods, and 3) evaluate the impact of wind speed measurement accuracy on differences between the methods. Our work is motivated by the hyper-local spatial scale at which planners and land managers operate and humans experience temperature; planning policies—from tree planting specifications to landscape design guidelines—are implemented at fine scales, and decision-makers require data that reflects how people experience heat within them. Our study is the first to assess the performance of various T mrt methods in a semi-arid climate, and, while our fieldwork is grounded in a single city, the methodological insights apply to broader urban contexts, especially as cities in temperate zones begin to experience more arid summer conditions due to climate change. Materials and Methods Study Area The study was conducted in the medium-sized city of Fort Collins, Colorado, USA, in July and August 2023 and 2024 (Fig. 1 ). Located at the base of the Rocky Mountains at 1,500 m (5,000 ft) elevation, Fort Collins experiences a semi-arid climate with 300 days of sunshine and 37 cm (15 in) of annual precipitation (City of Fort Collins). Data were collected on front yards within a 1.64 km 2 (0.63 mi 2 ) neighborhood in conjunction with other research examining thermal dynamics of varying front yard types (Benson et al., in press; Witvliet et al., unpublished results). Weather Conditions The first summer of data collection occurred during the region’s wettest summer on record, while the second summer had more typical dry conditions (NOAA Online Weather Data 2023 ). Between June and August of 2023 (summer 1), the city experienced nearly three times the usual rainfall. The average summertime monthly rainfall from 1990 to 2022 was 44 mm, while in 2023, it was 129 mm per month. The minimum air temperature in the 2023 study period was 8 ℃, the maximum was 37 ℃, and the mean was 21 ℃. In 2024 (summer 2), the average summertime monthly rainfall was 25 mm, minimum air temperature was 8 ℃, maximum 39 ℃, and mean 22 ℃ (NOAA Online Weather Data 2023 ). Sample Point Selection We collected data on 32 front yards (Fig. 2 ) during the summers of 2023 and 2024, visiting yards in the morning and afternoon to capture diurnal variation. In 2023, each property was visited on two separate mornings and two afternoons at five points per yard. Morning measurements were taken at three properties sequentially at 08:30 h, 09:30 h, and 10:30 h, with the same properties revisited in the afternoon in the same order at 14:30 h, 15:30 h, and 16:30 h. During the first two weeks, houses were visited in a randomized order. Once all yards had been visited for a full day, the order was re-randomized for a second round of data collection. Data collection took place between July 17th and August 16th, 2023. In 2024, data collection followed a similar protocol between August 1st and August 14th, with a few modifications. Some houses from the 2023 sample were replaced with new ones, properties were visited on a single day rather than two, and data were collected from four points within each yard instead of five. In total, 896 data points were collected across the two years. Micrometeorology Measurements A Kestrel 5400 meter and the biometeorology MaRTa cart collected micrometeorology data at each sample point on the front yards. The Kestrel meter was mounted on a tripod at a height of 1.1 meters to match the height of the MaRTa sensors, and the two devices were positioned opposite each other on either side of the sample point (Fig. 3 ). After allowing the Kestrel to acclimate to its environment at each measurement point, meteorological data were collected and averaged over one minute. The MaRTa platform is equipped with 12 Apogee radiometers to measure shortwave and longwave radiation in six directions, the Campbell Scientific HygroVue10 for air temperature ( \(\:{T}_{a}\) ) and relative humidity housed within an Apogee TS-100 SS fan aspirated radiation shield, an Apogee Infrared Radiometer for surface temperature, and a Gill Ultrasonic Anemometer for measuring wind speed ( \(\:{V}_{a}\) ) (Fig. 4 ). Adapted from the MaRTy machine developed by Middel and Krayenhoff ( 2019 ), MaRTa calculates mean radiant temperature ( \(\:{T}_{mrt}\) ) using the following equation: $$\:{T}_{mrt}=\sqrt[4]{\frac{{\sum\:}_{i=1}^{6}{W}_{i}({a}_{k}{K}_{i}+{a}_{l}{L}_{i})}{{a}_{l}*\sigma\:}-}273.15\:K$$ 1 Where \(\:{K}_{i}\) = incoming shortwave radiation, \(\:{L}_{i}\) = incoming longwave radiation, \(\:{a}_{k}\:\) = human emissivity for shortwave radiation (0.7), \(\:{a}_{l}\) = human emissivity for longwave radiation (0.97), \(\:\sigma\:\) = the Stefan-Boltzmann constant (5.67 \(\:\:*\:\) 10 −8 ), and \(\:{W}_{i}\) = weighting parameters for the angles of a standing human (0.06 for upward and downward radiation; 0.22 for lateral radiation) (Middel and Krayenhoff 2019 ). Alongside the MaRTa cart, we used a Kestrel 5400, which features a 25 mm diameter black-coated copper globe. The micrometeorology metrics utilized from the Kestrel were globe temperature ( \(\:{T}_{g}\) ), ( \(\:{V}_{a}\) ), and \(\:{T}_{a}\) , which were used to estimate \(\:{T}_{mrt}\) . We avoided taking micrometeorology measurements during rain events, defined as periods of active rainfall, to prevent data for Benson et al. (in press) from being highly skewed by unusual weather conditions. Given the exceptionally rainy conditions of the summer, measurements were occasionally taken during light drizzling. Analysis Objective 1: Compare multiple Kestrel mean radiant temperature ( \(\:{\varvec{T}}_{\varvec{m}\varvec{r}\varvec{t}}\) ) methods to MaRTa across different levels of spatial aggregation By comparing to gold standard six-directional \(\:{T}_{mrt}\) values, we tested the accuracy of six methods for converting globe thermometer readings to \(\:{T}_{mrt}\) at the hyper-local level (Table 1 ). Each method’s accuracy was determined by extracting RMSE for the Kestrel’s \(\:{T}_{mrt}\) estimate compared to MaRTa’s \(\:{T}_{mrt}\) readings. Additionally, minimum, maximum, mean, and variation from the mean were considered for each \(\:{T}_{mrt}\) method to understand the spread of data points. Of the Kestrel-based methods, five are from the literature and one utilizes a novel machine learning approach. All analyses were performed in RStudio (RStudio Team 2020 ) using the “tidyverse” package (Wickham et al. 2019 ). Table 1 Summary of the globe thermometer methods for calculating mean radiant temperature ( \(\:{T}_{mrt}\) ) examined in this study, including location of the study and the type of globe thermometer(s) used, if available. Authors Location Globe Thermometer: Color | Diameter Thorsson et al. ( 2007 ) Various Gray | 38 mm Vanos et al. ( 2021 ) Tempe, Arizona, USA Black | 150 mm diameter \(\:\:\) Ouyang et al. ( 2022 ) Hong Kong Black | 40 mm Black | 25.4 mm Gray | 40 mm Liu et al. ( 2022 ) Tianjin, China Black | 70 mm Lin and Matzarakis ( 2011 ) N/A N/A This study Fort Collins, Colorado, USA Black | 25 mm The conventional method, as outlined by Thorsson et al. ( 2007 ), uses the following equation: $$\:{T}_{mrt}={[{\left({T}_{g}+273.15\right)}^{4}+\frac{1.1*{10}^{8}*{{V}_{a}}^{0.6}}{{\epsilon\:}_{g}*{D}^{0.4}}\left({T}_{g}-{T}_{a}\right)]}^{0.25}-273.15$$ 2 Where \(\:{T}_{g}\) is globe temperature (℃), \(\:{V}_{a}\) is wind speed (m/s), \(\:{T}_{a}\) is air temperature (℃), \(\:{\epsilon\:}_{g}\) is globe emissivity (0.95), and \(\:D\) is globe diameter. For the globe diameter, 150 mm was input because of internal conversions the Kestrel uses when calculating \(\:{T}_{g}\) . The next method used a formula generated by Vanos et al. ( 2021 ) to correct for asymmetrical heating of globe thermometers. The corrected equation was developed under warm-hot and clear weather conditions in Tempe, Arizona, and resulted in the following: $$\:{T}_{mrt}=\:\sqrt[4]{{{(\overline{T}}_{s}+273.15)}^{4}+\left(0.24+2.08{{V}_{a}}^{0.5}+1.14{{V}_{a}}^{0.667}\right)\left({\overline{T}}_{s}-{T}_{a}\right){10}^{8}}-273.15$$ 3 Where, $$\:{\overline{T}}_{s}=1.345{T}_{g}-0.369{T}_{a}+0.725$$ Third, a method developed by Ouyang et al. ( 2022 ) was tested. The author’s formula was calibrated to a humid subtropical climate in Hong Kong, resulting in the following: $$\:{T}_{mrt}={[{\left({T}_{g}+273.15\right)}^{4}+\frac{0.678*{10}^{8}*{{V}_{a}}^{0.019}/{D}^{0.4}}{{\epsilon\:}_{g}}\left({T}_{g}-{T}_{a}\right)]}^{0.25}-273.15$$ 4 Fourth, a method outlined by Liu et al. ( 2022 ) was applied, where the authors used machine learning to alter the conventional formula, resulting in the following new equation: $$\:{T}_{mrt}={[{\left({T}_{g}+273.15\right)}^{4}+\frac{1.07*{10}^{8}*{{V}_{a}}^{0.251}}{{\epsilon\:}_{g}*{D}^{0.4}}\left({T}_{g}-{T}_{a}\right)]}^{0.25}-273.15$$ 5 The fifth method tested was proposed by Lin and Matzarakis ( 2011 ), where a constant wind speed value of 0.5 m/s was input into the conventional formula (Chen et al. 2014 ). Lastly, a random forest model was developed using a 70/30 split in the “randomForest” R package (Liaw and Wiener 2002 ); 70% of the MaRTa and Kestrel data pairs were randomly selected to train the model (n = 609), while the remaining 30% was used to test its accuracy (n = 258). The model used \(\:{T}_{g}\) , \(\:{T}_{a},\:\:\text{a}\text{n}\text{d}\:{V}_{a}\) from the Kestrel as input variables to build an ensemble of decision trees that predict \(\:{T}_{mrt}\) measured by MaRTa. In addition to inspecting individual points of Kestrel and MaRTa data, we also investigated the performance of globe thermometer methods at the front yard level by averaging all points per yard (five points of data were collected per yard in 2023, four points were collected in 2024). For each method of Kestrel-derived \(\:{T}_{mrt}\) , we aggregated points at the front yard level and then compared them to MaRTa aggregations at the same level by 1) comparing each method individually to MaRTa \(\:{T}_{mrt}\) and extracting RMSE, and 2) examining differences in averages and deviation from the mean between all methods. We excluded the random forest model from the front yard level analysis, as its 70/30 train-test split was applied across all points regardless of yard boundaries, resulting in incomplete or missing yard-level data that prevented consistent averaging. Objective 2: Assess how varying radiation components influence discrepancies between MaRTa and Kestrel mean radiant temperature ( \(\:{\varvec{T}}_{\varvec{m}\varvec{r}\varvec{t}}\) ) values Because high incoming solar radiation is known to skew globe thermometer data, we isolated MaRTa radiation components to examine impacts on ∆ \(\:{T}_{mrt}.\) First, differences in \(\:{T}_{mrt}\) between MaRTa and Kestrel were calculated for each data point ( ∆ \(\:{T}_{mrt}\) = \(\:MaRTa\:{T}_{mrt}-\:Kestrel\:{T}_{mrt}\) ; n = 867). Then, for each method of Kestrel \(\:{T}_{mrt}\) , we examined relationships between total incoming shortwave and longwave radiation on ∆ \(\:{T}_{mrt}.\) Total shortwave and longwave radiation were calculated as a weighted total according to the six-directional \(\:{T}_{mrt}\) equation. Objective 3: Evaluate the impact of wind speed measurement accuracy on differences between the methods We performed a sensitivity analysis on \(\:{V}_{a}\) measurement accuracy for the Kestrel-derived \(\:{T}_{mrt}\) methods to determine the amount of error introduced using a low-cost wind speed sensor. For each method, \(\:{V}_{a}\:\) readings from MaRTa’s sonic anemometer were input instead of those from the Kestrel’s impeller and change in RMSE was noted. The Lin and Matzarakis ( 2011 ) method was excluded from this analysis because it does not require \(\:{V}_{a}\) inputs from the Kestrel (it uses a stable wind speed value of 0.5 m/s). For the random forest model method utilizing \(\:{T}_{g}\) , \(\:{T}_{a},\:\:\text{a}\text{n}\text{d}\:{V}_{a}\) to predict MaRTa \(\:{T}_{mrt}\) , we replaced \(\:{V}_{a}\) with MaRTa \(\:{V}_{a}\) to see if that improved the model performance. We also extracted feature importance for each variable in both random forest model versions. Results Objective 1: Compare multiple Kestrel mean radiant temperature ( \(\:{T}_{mrt}\) ) methods to MaRTa across different levels of spatial aggregation We calculated and compared multiple globe thermometer methods for estimating \(\:{T}_{mrt}\) at two levels of spatial aggregation to evaluate the efficacy of each method on our data and observe changes in discrepancies incurred by shifting the level of spatial aggregation. By comparing each Kestrel \(\:{T}_{mrt}\) method to the gold-standard, six-directional MaRTa machine, we found that the random forest model performed best at the single-point level, compared to the other methods selected from the literature (Fig. 5 ). Out of the literature-based methods, the Lin and Matzarkis (2011) method worked best on our data and the Vanos et al. ( 2021 ) method performed the poorest. Furthermore, discrepancies between Kestrel and MaRTa \(\:{T}_{mrt}\) decreased when examining \(\:{T}_{mrt}\) aggregated at the front yard level. Specifically, each of the methods decreased in RMSE when scaling up from the hyper-local, individual point level. Overall, the globe thermometer methods showed closer agreement with MaRTa \(\:{T}_{mrt}\) at lower ranges than at the higher ranges of \(\:{T}_{mrt}\) . MaRTa showed less variation in \(\:{T}_{mrt}\) compared to the globe thermometer methods, apart from the machine learning method (Fig. 6 ). The Vanos et al. ( 2021 ) method had the most variation in values, while the random forest model had the least. Objective 2: Assess how varying radiation components influence discrepancies between MaRTa and Kestrel mean radiant temperature ( \(\:{\varvec{T}}_{\varvec{m}\varvec{r}\varvec{t}}\) ) values High solar radiation via sun exposure is known to cause errors in globe thermometer readings; we assessed the impact of shortwave radiation on our observed Kestrel \(\:{T}_{mrt}\) by deconstructing radiation components of MaRTa \(\:{T}_{mrt}\) and comparing that to ∆ \(\:{T}_{mrt}\) , where ∆ \(\:{T}_{mrt}\:\) = \(\:MaRTa\:{T}_{mrt}-\:Kestrel\:{T}_{mrt}\) . As total shortwave radiation increased, all six methods of Kestrel \(\:{T}_{mrt}\) increased in error, while incoming total longwave radiation appeared to have little influence on ∆ \(\:{T}_{mrt}\) (Fig. 7 ). Objective 3: Evaluate the impact of wind speed ( \(\:{\varvec{V}}_{\varvec{a}}\) ) measurement accuracy on differences between the methods For each globe thermometer method estimating \(\:{T}_{mrt}\) , we investigated error reduction incurred from use using a low-cost \(\:{V}_{a}\) sensor and found a slight reduction in RMSE for most methods (Table 2 ). On average, replacing the Kestrel impeller with MaRTa’s sonic anemometer resulted in a reduction in RMSE of 0.78. The random forest model, which uses a distinct approach from the other methods to determine \(\:{T}_{mrt}\) , showed little improvement from this alteration. This likely reflects the low influence of \(\:{V}_{a}\) in the model overall. Feature importance values—indicators of each variable’s relative contributions—were low for both Kestrel \(\:{V}_{a}\) (7.55) and MaRTa \(\:{V}_{a}\) (5.74), where a lower value indicates less importance in the model. This is compared with \(\:{T}_{g}\) (68.06 for the Kestrel model, 77.93 for the MaRTa model) and \(\:{T}_{a}\) (32.81 and 31.28, respectively). Table 2 Impact of using MaRTa’s sonic anemometer in place of the Kestrel 5400’s impeller for calculating \(\:{T}_{mrt}\) at the individual point level. The second column shows the change in RMSE associated with each method, while the third column shows the resulting RMSE for swapping in MaRTa wind speed ( \(\:{V}_{a}\) ) for each Kestrel method. Method Difference in RMSE using MaRTa \(\:{\varvec{V}}_{\varvec{a}}\) Resulting RMSE using MaRTa \(\:{\varvec{V}}_{\varvec{a}}\) Thorsson et al. ( 2007 ) -0.87 8.72 Vanos et al. ( 2021 ) -1.96 9.74 Ouyang et al. ( 2022 ) -0.99 8.13 Liu et al. ( 2022 ) -1.14 7.91 Lin & Matzarakis ( 2011 ) -- -- Random Forest Model + 0.27 6.76 Discussion Challenges to the Kestrel Approach: Wind Speed, Globe Type Solar Radiation For conventional globe thermometer-based methods, it has long been known that wind speed ( \(\:{V}_{a}\) ) dramatically influences \(\:{T}_{mrt}\) calculations, as the conventional method was originally developed for indoor environments with minimal air movement (ISO 7726 1998 ). Our analysis showed that a simple and effective way to smooth out Kestrel \(\:{T}_{mrt}\) inaccuracies was by applying a stable, low \(\:{V}_{a}\) constant to the Thorsson et al. ( 2007 ) method, as recommended by Lin and Matzarakis ( 2011 ) and Chen et al. ( 2014 ). Applying \(\:{V}_{a}\) from the MaRTa machine’s sonic anemometer to the various Kestrel \(\:{T}_{mrt}\) calculations also improved our models, but to a lesser degree than the Lin and Mazarakis (2011) method. This could be partially due to MaRTa’s placement slightly away from the Kestrel, meaning the two instruments may have experienced distinct \(\:{V}_{a}\) . In the case of the random forest model, performance did not improve with the use of MaRTa \(\:{V}_{a}\) , likely because \(\:{V}_{a}\) was not a particularly important feature to the model in the first place. Unlike the other methods, which are based on formulas where \(\:{V}_{a}\) directly influences the outcome, the random forest model uses data-driven decision trees and does not require any specific predefined relationships between inputs and the response. If a variable does not contribute substantially to improving predictive accuracy—either because it has limited variation or is less informative relative to other inputs—the model will assign it less importance. In this case, \(\:{V}_{a}\) contributed less than \(\:{T}_{g}\) and \(\:{T}_{a}\) , which more strongly predicted \(\:{T}_{mrt}\) across the dataset. Another factor influencing our globe-based \(\:{T}_{mrt}\) results is the globe sensor type. Each of the methods we examined were developed using globes that vary in size, material, and color. The Kestrel 5400 used in this study features a 25 mm black-coated copper globe, which equilibrates to its environment more quickly than larger globes but is also more susceptible to convective error (Kántor et al. 2015 ). Notably, the Kestrel internally converts its \(\:{T}_{g}\) readings to simulate those of a standard 150 mm globe thermometer. We expect that some of the discrepancy between our results and those reported in other studies arises from mismatches in globe sensor types. For example, Vanos et al. ( 2021 ) developed a \(\:{T}_{mrt}\) equation using data from a 150 mm globe. While the Kestrel’s output is adjusted to match this standard, its actual globe is much smaller. Therefore, we expect that had we used a physical 150 mm globe, the results from our Vanos et al. ( 2021 ) test would have more closely aligned with the six-directional \(\:{T}_{mrt}\) . The radiation decomposition analysis (Objective 3) added to our understanding of observed Kestrel \(\:{T}_{mrt}\) error, indicating that the globe thermometer tended to overpredict \(\:{T}_{mrt}\) under conditions of high shortwave radiation, which is consistent with Chen et al. ( 2014 ) and Liu et al. 2022 , but contrary to Thorsson et al. ( 2007 ). This discrepancy is likely due to our and Chen et al.’s ( 2014 ) use of a black globe sensor, which may “overheat” and measure hotter temperatures, whereas Thorsson et al. used a gray globe, which is closer in albedo to the human skin. This implies that solar radiation drives differences between globe thermometer and six-directional \(\:{T}_{mrt}\) , suggesting that the black globe thermometer readings become unstable under conditions with high shortwave radiation. Spatial Aggregation of Globe Thermometer-Derived Mean Radiant Temperature Reveals Thermal Patterns in Heterogeneous Landscapes Our results show that Kestrel-derived (globe thermometer) \(\:{T}_{mrt}\) aligns more closely with MaRTa (the gold standard for \(\:{T}_{mrt}\) measurements) when averaging across four to five points per yard, compared to its performance at the single-point level. RMSE values between Kestrel and MaRTa \(\:{T}_{mrt}\) were lower at the yard level, while hyper-local (single point) discrepancies were more pronounced. The front yards we studied ranged from 22 to 246 m² in area—roughly the size of a pocket park or green roof. These findings suggest that, in the absence of access to a six-directional platform, averaging multiple measurements across small urban spaces can capture broad thermal patterns with reasonable accuracy. The focus of urban ecologists is understanding how ecosystem processes interact with landscape heterogeneity, rather than isolated data points. Our findings indicate that averaging Kestrel \(\:{T}_{mrt}\) yields a practical and reliable approach for microclimate assessment in front yards, even if the Kestrel struggles to pinpoint extremely fine-scale variation. Indeed, Benson et al. (in press), using the same dataset, shows that aggregated T mrt measurements detect landscape thermal patterns where air temperature ( T a ) lacks sensitivity. These findings highlight that, despite its limitations, the Kestrel can effectively link urban landscape structure to biophysical outcomes when spatially aggregating. In addition to spatial aggregation and other previously discussed factors influencing the accuracy of the globe thermometer method, another relevant consideration is the temporal resolution of \(\:{T}_{mrt}\) data. Thorsson et al. ( 2007 ) recommended averaging globe thermometer readings over five-minute periods to reduce inconsistences. In our study, we recorded data at one-minute intervals, consecutively across the yard, such that the final \(\:{T}_{mrt}\) value for each front yard reflects a five-minute average. Thus, our method involves both temporal and spatial aggregation: the former through consecutive point measurements, and the latter through sampling across multiple locations within a yard. While we acknowledge that both types of aggregation contribute to error reduction, we expect spatial aggregation to be the dominant factor. This is because \(\:{T}_{mrt}\) variation across space is likely greater than variation across the short time span needed to complete measurements on a single yard. That being said, we cannot fully disentangle the contributions of spatial versus temporal averaging in this design, and the role of temporal smoothing warrants further attention in future work. Machine Learning for Refined Mean Radiant Temperature Field Methods Ours is the first study to incorporate a machine learning technique to predict \(\:{T}_{mrt}\) based on Kestrel data. By using MaRTa (six-directional) as a highly-accurate \(\:{T}_{mrt}\) reference, we were able to evaluate the performance of Kestrel metrics and found that this approach holds promise in bolstering Kestrel-based field methods. However, more data across diverse landscapes and under a range of weather conditions would be needed to develop a robust model for use beyond this study. Localized research may benefit from pursuing tailored machine learning models developed under a range of local weather and landscape conditions. Although this approach still relies on a six-directional machine for model training, its role shifts from being the primary \(\:{T}_{mrt}\) data source to serving as a tool for building models at key moments during a field campaign. We suspect that our random forest model would be further strengthened by incorporating shade into the model. In machine learning models built to predict human thermal experience, Guo et al. ( 2024 ) split their micrometeorology data into two sets based on presence of shade, while AlKhaled et al. ( 2024 ) included shade presence as a variable in their models to predict \(\:{T}_{mrt}\) based on air temperature and a variety of non-physical variables. Because shade presence is a key factor influencing \(\:{T}_{mrt}\) and \(\:{T}_{g}\) , combining one of these approaches with ours would almost certainly reduce our model error. Conclusions Our findings reinforce the six-directional method as a critical tool for collecting reliable \(\:{T}_{mrt}\) data while demonstrating how the globe thermometer approach can be adapted for urban ecology research. Fine-scale \(\:{T}_{mrt}\) data provide scientists across disciplines with information about the influence of urban landscape heterogeneity on biophysical outcomes—critical information for cities striving to adapt to challenges such as densification and extreme weather conditions. Although precise \(\:{T}_{mrt}\) measurements currently require specialized six-directional platforms like MaRTa, the accuracy of the globe thermometer method can be improved. Modifying the ISO Standard 7726 equation and averaging clusters of measurement points within a study area can reduce error. Further data are needed to evaluate the efficacy of alternative equations under diverse environmental conditions. However, our results suggest a simple improvement may lie in incorporating a low, constant V a into the conventional equation, as proposed by Lin and Matzarakis ( 2011 ). Researchers with access to six-directional \(\:{T}_{mrt}\) platforms should consider leveraging machine learning to develop localized adaptations of the globe thermometer method. However, ongoing validation against six-directional measurements would be essential, and models could not be applied beyond the range of conditions in which they were trained. A third promising technique has recently emerged that utilizes small cylinders to estimate T mrt . Developed by Rykaczewski et al. ( 2024 ), this cost-effective approach builds on the three-globe radiometer-anemometer concept developed by Nakayoshi et al. (2014), measuring shortwave and longwave radiation fluxes and the convection coefficient to estimate T mrt . While this technique shows promise as an affordable and accurate alternative to both six-directional and globe thermometer methods, it is still undergoing refinement (Rykaczewski et al. 2025 ) and was not suitable for application in our study. As cities grow more complex and climate pressures intensify, the demand has never been greater for scientists to produce results that are both adaptable and reliable at hyper-local scales. Our work bridges methodological rigor with pragmatic field applications, carving a path forward for urban ecologists seeking to capture spatial and thermal complexity of urban landscapes. Rather than choosing between accuracy and efficiency, we argue for an integrative approach—where methodological innovation, like machine learning and spatial aggregation, enhances the accessibility and ecological relevance of T mrt measurement. By refining the methods we use to sense and study the urban landscape, we sharpen our ability to ask and answer more meaningful questions about how humans and other organisms experience cities. Declarations Declaration of Funding This research was funded by the Faculty of Forestry at the University of British Columbia. Meteorological equipment were funded by the USDA Forest Service Denver Urban Field Station. Data Availability Data will be made available upon request. Competing Interests The authors declare no competing interests. Author Contributions Conceptualization: Aubrey Benson, Melissa McHale. Methodology: Aubrey Benson, John Frank, Melissa McHale. Data collection: Aubrey Benson, John Frank, George Valentine. Analysis: Aubrey Benson, John Frank, George Valentine, Melissa McHale. Writing-original draft: Aubrey Benson. Writing-review and editing: Ben Crawford, John Frank, Ariane Middel, George Valentine, Travis Warziniack, Melissa McHale. Acknowledgements The authors would like to thank the homeowners who generously allowed us to use their yards for our research, as well as all those who provided local support during our data collection. This work is supported through a collaboration with the USDA Forest Service Denver Urban Field Station, who are studying urban heat in the Front Range of Colorado. Finally, we extend our gratitude to all reviewers for providing invaluable feedback on manuscripts. References Aleksandrowicz, O., & Pearlmutter, D. (2023). The significance of shade provision in reducing street-level summer heat stress in a hot Mediterranean climate. Landscape and Urban Planning , 229 , 104588. https://doi.org/10.1016/J.LANDURBPLAN.2022.104588 AlKhaled, S. R., Middel, A., Shaeri, P., Buo, I., & Schneider, F. A. (2024). WebMRT: An online tool to predict summertime mean radiant temperature using machine learning. Sustainable Cities and Society , 115 . https://doi.org/10.1016/j.scs.2024.105861 Bao, Y., Gao, M., Luo, D., & Zhou, X. (2022). The Influence of Plant Community Characteristics in Urban Parks on the Microclimate. Forests , 13 (9). https://doi.org/10.3390/f13091342 Bruse, M., & Fleer, H. (1998). Simulating surface-plant-air interactions inside urban environments with a three dimensional numerical model. In Environmental Modelling & Software (Vol. 13). http://www.geographie.ruhr- Cao, J., Zhou, W., Zheng, Z., Ren, T., & Wang, W. (2021). Within-city spatial and temporal heterogeneity of air temperature and its relationship with land surface temperature. Landscape and Urban Planning , 206 . https://doi.org/10.1016/j.landurbplan.2020.103979 Chen, Y. C., Lin, T. P., & Matzarakis, A. (2014). Comparison of mean radiant temperature from field experiment and modelling: a case study in Freiburg, Germany. Theoretical and Applied Climatology , 118 (3), 535–551. https://doi.org/10.1007/s00704-013-1081-z Chow, W. T. L., Akbar, S. N. A. B. A., Heng, S. L., & Roth, M. (2016). Assessment of measured and perceived microclimates within a tropical urban forest. Urban Forestry and Urban Greening , 16 , 62–75. https://doi.org/10.1016/j.ufug.2016.01.010 Coutts, A. M., White, E. C., Tapper, N. J., Beringer, J., & Livesley, S. J. (2016). Temperature and human thermal comfort effects of street trees across three contrasting street canyon environments. Theoretical and Applied Climatology , 124 (1–2), 55–68. https://doi.org/10.1007/s00704-015-1409-y Crank, P. J., Middel, A., Wagner, M., Hoots, D., Smith, M., & Brazel, A. (2020). Validation of seasonal mean radiant temperature simulations in hot arid urban climates. Science of the Total Environment , 749 . https://doi.org/10.1016/j.scitotenv.2020.141392 Fort Collins Facts . (n.d.). City of Fort Collins. https://www.fcgov.com/visitor/fcfacts (accessed on 12 January 2023) Guo, R., Yang, B., Guo, Y., Li, H., Li, Z., Zhou, B., Hong, B., & Wang, F. (2024). Machine learning-based prediction of outdoor thermal comfort: Combining Bayesian optimization and the SHAP model. Building and Environment , 254 . https://doi.org/10.1016/j.buildenv.2024.111301 Ibsen, P. (2025). [Unpublished Manuscript]. United States Geological Survey. ISO 7726. (1998). Ergonomics of the thermal environment - Instruments for Measuring physical quantities. Jin, L., Schubert, S., Salim, M. H., & Schneider, C. (2020). Impact of air conditioning systems on the outdoor thermal environment during summer in Berlin, Germany. International Journal of Environmental Research and Public Health , 17 (13), 1–21. https://doi.org/10.3390/ijerph17134645 Kántor, N., Kovács, A., & Lin, T. P. (2015). Looking for simple correction functions between the mean radiant temperature from the “standard black globe” and the “six-directional” techniques in Taiwan. Theoretical and Applied Climatology , 121 (1–2), 99–111. https://doi.org/10.1007/s00704-014-1211-2 Lachapelle, J. A., Krayenhoff, E. S., Middel, A., Meltzer, S., Broadbent, A. M., & Georgescu, M. (2022). A microscale three-dimensional model of urban outdoor thermal exposure (TUF-Pedestrian). International Journal of Biometeorology , 66 (4), 833–848. https://doi.org/10.1007/s00484-022-02241-1 A. Liaw and M. Wiener (2002. Classification and Regression by randomForest. R News 2(3), 18-22. Lin, T. P., & Matzarakis A. (2011). Estimation of outdoor mean radiant temperature by field experiment and modelling for human-biometeorology use. The 11th Annual Meeting of the European Meteorological Society . Lindberg, F., Holmer, B., & Thorsson, S. (2008). SOLWEIG 1.0 - Modelling spatial variations of 3D radiant fluxes and mean radiant temperature in complex urban settings. International Journal of Biometeorology , 52 (7), 697–713. https://doi.org/10.1007/s00484-008-0162-7 Liu, K., You, W., Chen, X., & Liu, W. (2022). Study on the Influence of Globe Thermometer Method on the Accuracy of Calculating Outdoor Mean Radiant Temperature and Thermal Comfort. Atmosphere , 13 (5). https://doi.org/10.3390/atmos13050809 Matzarakis, A., Rutz, F., & Mayer, H. (2007). Modelling radiation fluxes in simple and complex environments - Application of the RayMan model. International Journal of Biometeorology , 51 (4), 323–334. https://doi.org/10.1007/s00484-006-0061-8 Matzarakis, A., Rutz, F., & Mayer, H. (2010). Modelling radiation fluxes in simple and complex environments: Basics of the RayMan model. International Journal of Biometeorology , 54 (2), 131–139. https://doi.org/10.1007/s00484-009-0261-0 Ecosystem Health and Sustainability (Vol. 4, Issue 5, pp. 115–131). Taylor and Francis Ltd. https://doi.org/10.1080/20964129.2018.1480905 Middel, A., Häb, K., Brazel, A. J., Martin, C. A., & Guhathakurta, S. (2014). Impact of urban form and design on mid-afternoon microclimate in Phoenix Local Climate Zones. In Landscape and Urban Planning (Vol. 122, p. 16). Elsevier BV. https://doi.org/10.1016/j.landurbplan.2013.11.004 Middel, A., Selover, N., Hagen, B., & Chhetri, N. (2016). Impact of shade on outdoor thermal comfort—a seasonal field study in Tempe, Arizona. International Journal of Biometeorology , 60 (12), 1849–1861. https://doi.org/10.1007/s00484-016-1172-5 Middel, A., & Krayenhoff, E. S. (2019). Micrometeorological determinants of pedestrian thermal exposure during record-breaking heat in Tempe, Arizona: Introducing the MaRTy observational platform. The Science of the Total Environment , 687 , 137–151. https://doi.org/10.1016/j.scitotenv.2019.06.085 Nakayoshi, M., Kanda, M., & de Dear, R. (2015). Globe Anemo-radiometer. Boundary-Layer Meteorology , 155 (2), 209–227. https://doi.org/10.1007/s10546-014-0003-7 NOAA Online Weather Data (2023). National Oceanic and Atmospheric Administration. https://www.weather.gov/wrh/climate?wfo=bou (accessed on 27 November 2023). Ouyang, W., Liu, Z., Lau, K., Shi, Y., & Ng, E. (2022). Comparing different recalibrated methods for estimating mean radiant temperature in outdoor environment. Building and Environment , 216 . https://doi.org/10.1016/j.buildenv.2022.109004 RStudio Team. (2020). RStudio: Integrated Development for R. RStudio, PBC, Boston, MA. Rykaczewski, K., Joshi, A., Viswanathan, S. H., Guddanti, S. S., Sadeghi, K., Gupta, M., Jaiswal, A. K., Kompally, K., Pathikonda, G., Barlett, R., Vanos, J. K., & Middel, A. (2024). A simple three-cylinder radiometer and low-speed anemometer to characterize human extreme heat exposure. International Journal of Biometeorology . https://doi.org/10.1007/s00484-024-02646-0 Rykaczewski, K., Joshi, A., Viswanathan, S. H., Parkerson, E., Gupta, M., Park, M., ... & Middel, A. (2025). Advanced human heat exposure sensing using two cylinder anemometer and radiometer: introducing CARla. International Journal of Biometeorology, 1-14. Thorsson, S., Lindberg, F., Eliasson, I., & Holmer, B. (2007). Different methods for estimating the mean radiant temperature in an outdoor urban setting. International Journal of Climatology , 27 (14), 1983–1993. https://doi.org/10.1002/JOC.1537 Turner, K. V., Rogers, M. L., Zhang, Y., Middel, A., Schneider, F. A., Ocón, J. P., Seeley, M., & Dialesandro, J. (2022). More than surface temperature: mitigating thermal exposure in hyper-local land system. Journal of Land Use Science , 17 (1), 79–99. https://doi.org/10.1080/1747423X.2021.2015003 University of Utah, Department of Atmospheric Sciences. (2024). MesoWest weather data (Station KFNL). https://mesowest.utah.edu/cgi-bin/droman/meso_base_dyn.cgi?stn=KFNL Vanos, J. K., Rykaczewski, K., Middel, A., Vecellio, D. J., Brown, R. D., & Gillespie, T. J. (2021). Improved methods for estimating mean radiant temperature in hot and sunny outdoor settings. International Journal of Biometeorology , 65 (6), 967–983. https://doi.org/10.1007/s00484-021-02131-y Wang, J., Zhou, W., & Jiao, M. (2022). Location matters: planting urban trees in the right places improves cooling. Frontiers in Ecology and the Environment , 20 (3), 147–151. https://doi.org/10.1002/fee.2455 Water Sources . (n.d.). City of Fort Collins. https://www.fcgov.com/utilities/water-sources. Accessed on 13 February 2023. Wickham H, Averick M, Bryan J, Chang W, McGowan LD, François R, Grolemund G, Hayes A, Henry L, Hester J, Kuhn M, Pedersen TL, Miller E, Bache SM, Müller K, Ooms J, Robinson D, Seidel DP, Spinu V, Takahashi K, Vaughan D, Wilke C, Woo K, Yutani H (2019). “Welcome to the tidyverse.” Journal of Open Source Software, *4*(43), 1686. doi:10.21105/joss.01686 . Cite Share Download PDF Status: Published Journal Publication published 16 Oct, 2025 Read the published version in International Journal of Biometeorology → Version 1 posted Reviewers agreed at journal 20 May, 2025 Reviewers invited by journal 29 Apr, 2025 Editor assigned by journal 16 Apr, 2025 First submitted to journal 15 Apr, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6432428","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":449537620,"identity":"3f878b4b-4368-4b1c-b15f-f9c5598d73fd","order_by":0,"name":"Aubrey Benson","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6ElEQVRIiWNgGAWjYFCCHDApByIkGBAkYS3GDGxgxQbEa0lsIFoLf3vuwccFf2zS58/vMbzxcccfBoPbzQ8YftTg1iJx5l2y8cy2tNwNx3iMLWeeMWAwuHPMgLHnGB5rbuSYSfM2HM7dwMYDZLQBtdzIYWAG+QwXkL+RY/6b58/hdPk2oJa/cC3/cGsBKjBj5mE7nMBwDKiFEaaFsQ23FkOgX4DuSTPccCyt2LK3zZhHEuiXg719uLXIHc89+Jnnj428fPPhjTd+tsnJ8d1ufvjgxzc83kcHPCDiAAkaRsEoGAWjYBRgAQBYbE4yPipi+AAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0009-0006-5518-1461","institution":"The University of British Columbia","correspondingAuthor":true,"prefix":"","firstName":"Aubrey","middleName":"","lastName":"Benson","suffix":""},{"id":449537621,"identity":"b4fe24e0-d1bc-4457-8c0b-93c5467a8d08","order_by":1,"name":"Ben R. Crawford","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Ben","middleName":"R.","lastName":"Crawford","suffix":""},{"id":449537622,"identity":"ac55e43f-2fd7-44c8-9d0d-171997f40d10","order_by":2,"name":"John M. Frank","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"John","middleName":"M.","lastName":"Frank","suffix":""},{"id":449537623,"identity":"6273903f-2155-4fb3-8926-1481ec09922c","order_by":3,"name":"Ariane Middel","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Ariane","middleName":"","lastName":"Middel","suffix":""},{"id":449537624,"identity":"6579fa93-1a0d-4cca-900d-9dd03181dc63","order_by":4,"name":"George P. Valentine","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"George","middleName":"P.","lastName":"Valentine","suffix":""},{"id":449537625,"identity":"09ffb333-a2f1-4e72-a9a9-9c3fa89072e5","order_by":5,"name":"Travis Warziniack","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Travis","middleName":"","lastName":"Warziniack","suffix":""},{"id":449537626,"identity":"e20feee9-0ac8-42f1-a333-3ef75de1ebc4","order_by":6,"name":"Melissa R. McHale","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Melissa","middleName":"R.","lastName":"McHale","suffix":""}],"badges":[],"createdAt":"2025-04-12 06:00:51","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6432428/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6432428/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00484-025-03038-8","type":"published","date":"2025-10-16T15:56:50+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":81970490,"identity":"b3dac3d1-ba97-44ff-9a0c-12e100fcae4f","added_by":"auto","created_at":"2025-05-05 12:23:33","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1072600,"visible":true,"origin":"","legend":"\u003cp\u003eFort Collins is a city in northern Colorado, USA, at the base of the Rocky Mountains.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6432428/v1/a4074317aeda4c52ee1805af.png"},{"id":81970460,"identity":"afa88c84-74a1-4595-95dc-5bed0391d319","added_by":"auto","created_at":"2025-05-05 12:23:32","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":106015,"visible":true,"origin":"","legend":"\u003cp\u003eSystematic sampling design for the front yards studied. A two-by-two grid was first established, followed by placing five sample points. For the second year of data collection, in 2024, only four points were measured, omitting the central point in each yard.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6432428/v1/08067ea82d8769f425cb6c61.png"},{"id":81970523,"identity":"9a9cb13f-429d-43aa-836c-275ccb1f28bb","added_by":"auto","created_at":"2025-05-05 12:23:34","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":924017,"visible":true,"origin":"","legend":"\u003cp\u003eThe MaRTa machine (right) and Kestrel (left) were positioned opposite each other over each sample point (central yellow/orange “x”) and recorded data for one minute.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6432428/v1/3a8b5878d2324959c5d713db.png"},{"id":81970488,"identity":"e79c5caa-016e-4c8e-8b1d-6803f2c9bb1b","added_by":"auto","created_at":"2025-05-05 12:23:33","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":724845,"visible":true,"origin":"","legend":"\u003cp\u003eThe MaRTa machine is equipped with shortwave and longwave radiation sensors in six directions and sensors for air temperature, surface temperature, relative humidity, and wind velocity.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6432428/v1/b341bde42be18f21c2ce312e.png"},{"id":81970405,"identity":"4d040984-bb2a-4209-b636-f083c98d3a91","added_by":"auto","created_at":"2025-05-05 12:23:31","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":606839,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of Kestrel (globe thermometer) methods for calculating mean radiant temperature (Tmrt), compared to MaRTa (six-directional). Blue panels depict all points from all yards included in the study, while orange panels depict points aggregated at the front yard level. Solid black lines denote linear lines of best fit. Black dashed lines denote a theoretical slope of 1.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6432428/v1/6d1589824a9a2ff82858f9f1.png"},{"id":81970881,"identity":"e43ed784-3e76-4381-bef9-ea100bd4d8ac","added_by":"auto","created_at":"2025-05-05 12:31:34","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":133724,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of Tmrt methods, including MaRTa (6-directional), (A.) conventional method, (B.) Vanos et al. (2021), (C.) Ouyang et al. (2022), (D.) Liu et al. (2022), (E.) Lin and Matzarakis (2011), and (F.) a random forest model. Red labels indicate maximum values for each dataset; blue labels indicate minimum values; black labels represent means.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-6432428/v1/265138341502b05d0064bb2e.png"},{"id":81970515,"identity":"fda58165-bf6d-4480-9bc9-cb27c91e7c95","added_by":"auto","created_at":"2025-05-05 12:23:34","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":145441,"visible":true,"origin":"","legend":"\u003cp\u003eRelationships between radiation and the difference between Kestrel and MaRTa mean radiant temperature (∆Tmrt =MaRTa Tmrt-Kestrel Tmrt) values for multiple Kestrel T_mrt methods. Blue panels depict total incoming shortwave (SW) radiation, while orange panels depict total incoming longwave (LW) radiation. Radiation is weighted according to the six-directional method T_mrt equation (0.06 weighting for upward and downward radiation; 0.22 for lateral radiation). Note that shortwave panels use a semi-log scale.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-6432428/v1/63a420d2735820f4a41228a6.png"},{"id":93955907,"identity":"0f2ff8dc-273b-4d51-9f26-2299616df7e8","added_by":"auto","created_at":"2025-10-20 16:06:09","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4427954,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6432428/v1/a1f22a66-ec82-4837-b9b9-950edf2a4d21.pdf"}],"financialInterests":"","formattedTitle":"Bridging accuracy and efficiency: Advancing mean radiant temperature measurement in Urban Ecology","fulltext":[{"header":"Introduction","content":"\u003cp\u003eExtreme weather and shifting climate regimes are increasingly impacting cities, challenging practitioners to understand how humans, flora, and fauna are experiencing temperature as they live and move across urban landscapes. Mean radiant temperature (\u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e)\u0026mdash;the weighted total of all incoming shortwave and longwave radiation for a reference individual or object\u0026mdash;is emerging as a key tool for capturing hyper-local thermal dynamics (ISO 7726 \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e1998\u003c/span\u003e). \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e is especially useful for detecting and measuring fine-scale spatial and thermal heterogeneity, providing valuable information about how humans and other organisms experience heat in a city (Middel et al. 2019; 2021; Turner et al. \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). This information can help inform strategies that promote human thermal comfort, manage heat exposure, and shape climate-responsive urban design.\u003c/p\u003e \u003cp\u003eDespite its growing importance, obtaining \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e in the field remains technically challenging. Scientists often rely on radiation modeling software like RayMan (Matzarakis et al. \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2010\u003c/span\u003e), SOLWEIG (Lindberg et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2008\u003c/span\u003e), TUF-Pedestrian (Lachapelle et al. \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), and ENVI-met (Bruse and Fleer \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1998\u003c/span\u003e), which estimate \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e based on assumptions about urban geometry and biophysical processes. While these models offer useful estimates of radiative loading, they must be validated by on-the-ground measurements to ensure accurate microclimate assessments across diverse urban landscapes (Crank et al. \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Despite advancements in \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e measurement methods, the most accurate and reliable methods remain bulky, costly, and time-consuming to build (Rykaczewski et al. \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Thus, scientists and practitioners are challenged to determine when less resource-intensive methods suffice for their research and management objectives and how to implement them effectively.\u003c/p\u003e \u003cp\u003eThe highest cost, most accurate method for measuring \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e to date is often referred to as the six-directional method, integral radiation measurement, or gold standard approach- referred to hereafter as the six-directional method. This approach requires measuring shortwave and longwave radiation upward, downward, and in four orthogonal horizontal directions to estimate the average incoming radiation for a typical standing human (H\u0026ouml;ppe 1992; Thorsson et al. \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Middel and Krayenhoff \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Aleksandrowicz and Pearlmutter, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Cheung et al. 2024). The setup typically involves mounting 12 radiometers on two extended arms at 1.1 m height, making them rather difficult to build, maneuver, and mass-produce. Six-directional \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e measurements have been shown to vary drastically with changes to ground and shade cover (Turner et al. \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) and be strong predictors of spatio-temporal variation in human thermal comfort (Middel and Krayenhoff \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), making the approach a valuable method for estimating the heat load on the human body.\u003c/p\u003e \u003cp\u003eA popular alternative to the six-directional approach is the globe thermometer method, which estimates \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e using air temperature (\u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003ea\u003c/em\u003e\u003c/sub\u003e), wind speed (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e), and globe temperature (\u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003eg\u003c/em\u003e\u003c/sub\u003e)- the internal temperature of a small globe sensor. This widely used method is easier to implement, but the conventional formula for obtaining \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e (ISO 7726 \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e1998\u003c/span\u003e) using the globe thermometer method yields results that differ from the six-directional approach. Discrepancies arise because \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003eg\u003c/em\u003e\u003c/sub\u003e relies on convection, which, in turn depends on accurate \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e measurements to correct for. Due to these inaccuracies, researchers have sought to fine-tune the globe method equation to varying environmental conditions and globe sensor types by developing alternative formulas (Lin and Matzarakis \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Chen et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Vanos et al. \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Ouyang et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Liu et al. \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). All methods are developed under specific environmental conditions and may or may not apply to other geographic or temporal settings.\u003c/p\u003e \u003cp\u003eEffective climate adaptation and urban heat mitigation depend on cities\u0026rsquo; ability to measure and monitor thermal dynamics at hyper-local scales that reflect the human experience. To better understand how people experience temperature at the ground level and how thermal conditions affect urban trees and other non-human organisms, scientists must find more accessible measurement techniques that provide accurate results. This study addresses the need for practical and ecologically informative approaches to \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e measurement that can support planning and policy decisions. Using a residential neighborhood in a semi-arid city as a test case, we evaluate the performance of low-cost globe thermometer readings (from a Kestrel 5400) compared to the six-directional, gold standard method (using a mobile biometeorological system called MaRTa). We used different globe thermometer conversion techniques from the literature and a novel machine learning approach. Specifically, we aim to 1) compare multiple Kestrel \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e methods to MaRTa across different levels of spatial aggregation, 2) assess how varying radiation components influence discrepancies between MaRTa and Kestrel \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e values for each of the methods, and 3) evaluate the impact of wind speed measurement accuracy on differences between the methods.\u003c/p\u003e \u003cp\u003eOur work is motivated by the hyper-local spatial scale at which planners and land managers operate and humans experience temperature; planning policies\u0026mdash;from tree planting specifications to landscape design guidelines\u0026mdash;are implemented at fine scales, and decision-makers require data that reflects how people experience heat within them. Our study is the first to assess the performance of various \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e methods in a semi-arid climate, and, while our fieldwork is grounded in a single city, the methodological insights apply to broader urban contexts, especially as cities in temperate zones begin to experience more arid summer conditions due to climate change.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003eStudy Area\u003c/h2\u003e\n \u003cp\u003eThe study was conducted in the medium-sized city of Fort Collins, Colorado, USA, in July and August 2023 and 2024 (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). Located at the base of the Rocky Mountains at 1,500 m (5,000 ft) elevation, Fort Collins experiences a semi-arid climate with 300 days of sunshine and 37 cm (15 in) of annual precipitation (City of Fort Collins). Data were collected on front yards within a 1.64 km\u003csup\u003e2\u003c/sup\u003e (0.63 mi\u003csup\u003e2\u003c/sup\u003e) neighborhood in conjunction with other research examining thermal dynamics of varying front yard types (Benson et al., in press; Witvliet et al., unpublished results).\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eWeather Conditions\u003c/h3\u003e\n\u003cp\u003eThe first summer of data collection occurred during the region\u0026rsquo;s wettest summer on record, while the second summer had more typical dry conditions (NOAA Online Weather Data \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e). Between June and August of 2023 (summer 1), the city experienced nearly three times the usual rainfall. The average summertime monthly rainfall from 1990 to 2022 was 44 mm, while in 2023, it was 129 mm per month. The minimum air temperature in the 2023 study period was 8 ℃, the maximum was 37 ℃, and the mean was 21 ℃. In 2024 (summer 2), the average summertime monthly rainfall was 25 mm, minimum air temperature was 8 ℃, maximum 39 ℃, and mean 22 ℃ (NOAA Online Weather Data \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eSample Point Selection\u003c/h3\u003e\n\u003cp\u003eWe collected data on 32 front yards (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e) during the summers of 2023 and 2024, visiting yards in the morning and afternoon to capture diurnal variation. In 2023, each property was visited on two separate mornings and two afternoons at five points per yard. Morning measurements were taken at three properties sequentially at 08:30 h, 09:30 h, and 10:30 h, with the same properties revisited in the afternoon in the same order at 14:30 h, 15:30 h, and 16:30 h. During the first two weeks, houses were visited in a randomized order. Once all yards had been visited for a full day, the order was re-randomized for a second round of data collection. Data collection took place between July 17th and August 16th, 2023. In 2024, data collection followed a similar protocol between August 1st and August 14th, with a few modifications. Some houses from the 2023 sample were replaced with new ones, properties were visited on a single day rather than two, and data were collected from four points within each yard instead of five. In total, 896 data points were collected across the two years.\u003c/p\u003e\n\u003ch3\u003eMicrometeorology Measurements\u003c/h3\u003e\n\u003cp\u003eA Kestrel 5400 meter and the biometeorology MaRTa cart collected micrometeorology data at each sample point on the front yards. The Kestrel meter was mounted on a tripod at a height of 1.1 meters to match the height of the MaRTa sensors, and the two devices were positioned opposite each other on either side of the sample point (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e). After allowing the Kestrel to acclimate to its environment at each measurement point, meteorological data were collected and averaged over one minute.\u003c/p\u003e\n\u003cp\u003eThe MaRTa platform is equipped with 12 Apogee radiometers to measure shortwave and longwave radiation in six directions, the Campbell Scientific HygroVue10 for air temperature (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{a}\\)\u003c/span\u003e\u003c/span\u003e) and relative humidity housed within an Apogee TS-100 SS fan aspirated radiation shield, an Apogee Infrared Radiometer for surface temperature, and a Gill Ultrasonic Anemometer for measuring wind speed (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e) (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e). Adapted from the MaRTy machine developed by Middel and Krayenhoff (\u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e), MaRTa calculates mean radiant temperature (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e) using the following equation:\u003c/p\u003e\n\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e$$\\:{T}_{mrt}=\\sqrt[4]{\\frac{{\\sum\\:}_{i=1}^{6}{W}_{i}({a}_{k}{K}_{i}+{a}_{l}{L}_{i})}{{a}_{l}*\\sigma\\:}-}273.15\\:K$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{K}_{i}\\)\u003c/span\u003e\u003c/span\u003e = incoming shortwave radiation, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{L}_{i}\\)\u003c/span\u003e\u003c/span\u003e = incoming longwave radiation, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{a}_{k}\\:\\)\u003c/span\u003e\u003c/span\u003e= human emissivity for shortwave radiation (0.7), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{a}_{l}\\)\u003c/span\u003e\u003c/span\u003e = human emissivity for longwave radiation (0.97), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003e = the Stefan-Boltzmann constant (5.67\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:*\\:\\)\u003c/span\u003e\u003c/span\u003e10\u003csup\u003e\u0026minus;8\u003c/sup\u003e), and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{W}_{i}\\)\u003c/span\u003e\u003c/span\u003e = weighting parameters for the angles of a standing human (0.06 for upward and downward radiation; 0.22 for lateral radiation) (Middel and Krayenhoff \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAlongside the MaRTa cart, we used a Kestrel 5400, which features a 25 mm diameter black-coated copper globe. The micrometeorology metrics utilized from the Kestrel were globe temperature (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{g}\\)\u003c/span\u003e\u003c/span\u003e), (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e), and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{a}\\)\u003c/span\u003e\u003c/span\u003e, which were used to estimate \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eWe avoided taking micrometeorology measurements during rain events, defined as periods of active rainfall, to prevent data for Benson et al. (in press) from being highly skewed by unusual weather conditions. Given the exceptionally rainy conditions of the summer, measurements were occasionally taken during light drizzling.\u003c/p\u003e\u003ch3\u003eAnalysis\u003c/h3\u003e\u003cp\u003e\u003cem\u003eObjective 1: Compare multiple Kestrel mean radiant temperature (\u003c/em\u003e \u003cspan class=\"InlineEquation\"\u003e\u0026nbsp;\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{T}}_{\\varvec{m}\\varvec{r}\\varvec{t}}\\)\u003c/span\u003e\u0026nbsp;\u003c/span\u003e \u003cem\u003e) methods to MaRTa across different levels of spatial aggregation\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003eBy comparing to gold standard six-directional \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e values, we tested the accuracy of six methods for converting globe thermometer readings to \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e at the hyper-local level (Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). Each method\u0026rsquo;s accuracy was determined by extracting RMSE for the Kestrel\u0026rsquo;s \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e estimate compared to MaRTa\u0026rsquo;s \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e readings. Additionally, minimum, maximum, mean, and variation from the mean were considered for each \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e method to understand the spread of data points. Of the Kestrel-based methods, five are from the literature and one utilizes a novel machine learning approach. All analyses were performed in RStudio (RStudio Team \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e) using the \u0026ldquo;tidyverse\u0026rdquo; package (Wickham et al. \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eSummary of the globe thermometer methods for calculating mean radiant temperature (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e) examined in this study, including location of the study and the type of globe thermometer(s) used, if available.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"3\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAuthors\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLocation\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eGlobe Thermometer: Color | Diameter\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eThorsson et al. (\u003cspan class=\"CitationRef\"\u003e2007\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eVarious\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGray | 38 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eVanos et al. (\u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTempe, Arizona, USA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBlack | 150 mm diameter\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOuyang et al. (\u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHong Kong\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBlack | 40 mm\u003c/p\u003e\n \u003cp\u003eBlack | 25.4 mm\u003c/p\u003e\n \u003cp\u003eGray | 40 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLiu et al. (\u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTianjin, China\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBlack | 70 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLin and Matzarakis (\u003cspan class=\"CitationRef\"\u003e2011\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eN/A\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eN/A\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eThis study\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFort Collins, Colorado, USA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBlack | 25 mm\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eThe conventional method, as outlined by Thorsson et al. (\u003cspan class=\"CitationRef\"\u003e2007\u003c/span\u003e), uses the following equation:\u003c/p\u003e\n \u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e$$\\:{T}_{mrt}={[{\\left({T}_{g}+273.15\\right)}^{4}+\\frac{1.1*{10}^{8}*{{V}_{a}}^{0.6}}{{\\epsilon\\:}_{g}*{D}^{0.4}}\\left({T}_{g}-{T}_{a}\\right)]}^{0.25}-273.15$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{g}\\)\u003c/span\u003e\u003c/span\u003e is globe temperature (℃), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e is wind speed (m/s), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{a}\\)\u003c/span\u003e\u003c/span\u003e is air temperature (℃), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\epsilon\\:}_{g}\\)\u003c/span\u003e\u003c/span\u003e is globe emissivity (0.95), and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:D\\)\u003c/span\u003e\u003c/span\u003e is globe diameter. For the globe diameter, 150 mm was input because of internal conversions the Kestrel uses when calculating \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{g}\\)\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e\n \u003cp\u003eThe next method used a formula generated by Vanos et al. (\u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e) to correct for asymmetrical heating of globe thermometers. The corrected equation was developed under warm-hot and clear weather conditions in Tempe, Arizona, and resulted in the following:\u003c/p\u003e\n \u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e$$\\:{T}_{mrt}=\\:\\sqrt[4]{{{(\\overline{T}}_{s}+273.15)}^{4}+\\left(0.24+2.08{{V}_{a}}^{0.5}+1.14{{V}_{a}}^{0.667}\\right)\\left({\\overline{T}}_{s}-{T}_{a}\\right){10}^{8}}-273.15$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003eWhere,\u003c/p\u003e\n \u003cdiv id=\"Equa\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e$$\\:{\\overline{T}}_{s}=1.345{T}_{g}-0.369{T}_{a}+0.725$$\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003eThird, a method developed by Ouyang et al. (\u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e) was tested. The author\u0026rsquo;s formula was calibrated to a humid subtropical climate in Hong Kong, resulting in the following:\u003c/p\u003e\n \u003cdiv id=\"Equ4\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equ4\" name=\"EquationSource\"\u003e$$\\:{T}_{mrt}={[{\\left({T}_{g}+273.15\\right)}^{4}+\\frac{0.678*{10}^{8}*{{V}_{a}}^{0.019}/{D}^{0.4}}{{\\epsilon\\:}_{g}}\\left({T}_{g}-{T}_{a}\\right)]}^{0.25}-273.15$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e4\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003eFourth, a method outlined by Liu et al. (\u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e) was applied, where the authors used machine learning to alter the conventional formula, resulting in the following new equation:\u003c/p\u003e\n \u003cdiv id=\"Equ5\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equ5\" name=\"EquationSource\"\u003e$$\\:{T}_{mrt}={[{\\left({T}_{g}+273.15\\right)}^{4}+\\frac{1.07*{10}^{8}*{{V}_{a}}^{0.251}}{{\\epsilon\\:}_{g}*{D}^{0.4}}\\left({T}_{g}-{T}_{a}\\right)]}^{0.25}-273.15$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e5\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003eThe fifth method tested was proposed by Lin and Matzarakis (\u003cspan class=\"CitationRef\"\u003e2011\u003c/span\u003e), where a constant wind speed value of 0.5 m/s was input into the conventional formula (Chen et al. \u003cspan class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eLastly, a random forest model was developed using a 70/30 split in the \u0026ldquo;randomForest\u0026rdquo; R package (Liaw and Wiener \u003cspan class=\"CitationRef\"\u003e2002\u003c/span\u003e); 70% of the MaRTa and Kestrel data pairs were randomly selected to train the model (n\u0026thinsp;=\u0026thinsp;609), while the remaining 30% was used to test its accuracy (n\u0026thinsp;=\u0026thinsp;258). The model used \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{g}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{a},\\:\\:\\text{a}\\text{n}\\text{d}\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e from the Kestrel as input variables to build an ensemble of decision trees that predict \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e measured by MaRTa.\u003c/p\u003e\n \u003cp\u003eIn addition to inspecting individual points of Kestrel and MaRTa data, we also investigated the performance of globe thermometer methods at the front yard level by averaging all points per yard (five points of data were collected per yard in 2023, four points were collected in 2024). For each method of Kestrel-derived \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e, we aggregated points at the front yard level and then compared them to MaRTa aggregations at the same level by 1) comparing each method individually to MaRTa \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e and extracting RMSE, and 2) examining differences in averages and deviation from the mean between all methods. We excluded the random forest model from the front yard level analysis, as its 70/30 train-test split was applied across all points regardless of yard boundaries, resulting in incomplete or missing yard-level data that prevented consistent averaging.\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eObjective 2: Assess how varying radiation components influence discrepancies between MaRTa and Kestrel mean radiant temperature (\u003c/em\u003e \u003cspan class=\"InlineEquation\"\u003e\u0026nbsp;\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{T}}_{\\varvec{m}\\varvec{r}\\varvec{t}}\\)\u003c/span\u003e\u0026nbsp;\u003c/span\u003e \u003cem\u003e) values\u003c/em\u003e\u003c/p\u003e\n \u003cp\u003eBecause high incoming solar radiation is known to skew globe thermometer data, we isolated MaRTa radiation components to examine impacts on \u003cem\u003e∆\u003c/em\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}.\\)\u003c/span\u003e\u003c/span\u003e First, differences in \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e between MaRTa and Kestrel were calculated for each data point (\u003cem\u003e∆\u003c/em\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e\u003cem\u003e=\u003c/em\u003e \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:MaRTa\\:{T}_{mrt}-\\:Kestrel\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e; n\u0026thinsp;=\u0026thinsp;867). Then, for each method of Kestrel \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e, we examined relationships between total incoming shortwave and longwave radiation on \u003cem\u003e∆\u003c/em\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}.\\)\u003c/span\u003e\u003c/span\u003e Total shortwave and longwave radiation were calculated as a weighted total according to the six-directional \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e equation.\u003c/p\u003e\n \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003cp\u003eObjective 3: Evaluate the impact of wind speed measurement accuracy on differences between the methods\u003c/p\u003e\n \u003cp\u003eWe performed a sensitivity analysis on \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e measurement accuracy for the Kestrel-derived \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e methods to determine the amount of error introduced using a low-cost wind speed sensor. For each method, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\:\\)\u003c/span\u003e\u003c/span\u003ereadings from MaRTa\u0026rsquo;s sonic anemometer were input instead of those from the Kestrel\u0026rsquo;s impeller and change in RMSE was noted. The Lin and Matzarakis (\u003cspan class=\"CitationRef\"\u003e2011\u003c/span\u003e) method was excluded from this analysis because it does not require \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e inputs from the Kestrel (it uses a stable wind speed value of 0.5 m/s). For the random forest model method utilizing \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{g}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{a},\\:\\:\\text{a}\\text{n}\\text{d}\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e to predict MaRTa \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e, we replaced \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e with MaRTa \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e to see if that improved the model performance. We also extracted feature importance for each variable in both random forest model versions.\u003c/p\u003e\n \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cem\u003eObjective 1: Compare multiple Kestrel mean radiant temperature (\u003c/em\u003e \u003cspan class=\"InlineEquation\"\u003e\u0026nbsp;\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u0026nbsp;\u003c/span\u003e \u003cem\u003e) methods to MaRTa across different levels of spatial aggregation\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eWe calculated and compared multiple globe thermometer methods for estimating \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e at two levels of spatial aggregation to evaluate the efficacy of each method on our data and observe changes in discrepancies incurred by shifting the level of spatial aggregation. By comparing each Kestrel \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e method to the gold-standard, six-directional MaRTa machine, we found that the random forest model performed best at the single-point level, compared to the other methods selected from the literature (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e). Out of the literature-based methods, the Lin and Matzarkis (2011) method worked best on our data and the Vanos et al. (\u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e) method performed the poorest. Furthermore, discrepancies between Kestrel and MaRTa \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e decreased when examining \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e aggregated at the front yard level. Specifically, each of the methods decreased in RMSE when scaling up from the hyper-local, individual point level. Overall, the globe thermometer methods showed closer agreement with MaRTa \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e at lower ranges than at the higher ranges of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e\n\u003cp\u003eMaRTa showed less variation in \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e compared to the globe thermometer methods, apart from the machine learning method (Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e). The Vanos et al. (\u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e) method had the most variation in values, while the random forest model had the least.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eObjective 2: Assess how varying radiation components influence discrepancies between MaRTa and Kestrel mean radiant temperature (\u003c/em\u003e \u003cspan class=\"InlineEquation\"\u003e\u0026nbsp;\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{T}}_{\\varvec{m}\\varvec{r}\\varvec{t}}\\)\u003c/span\u003e\u0026nbsp;\u003c/span\u003e \u003cem\u003e) values\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eHigh solar radiation via sun exposure is known to cause errors in globe thermometer readings; we assessed the impact of shortwave radiation on our observed Kestrel \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e by deconstructing radiation components of MaRTa \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e and comparing that to ∆\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e, where ∆\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\:\\)\u003c/span\u003e\u003c/span\u003e= \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:MaRTa\\:{T}_{mrt}-\\:Kestrel\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e. As total shortwave radiation increased, all six methods of Kestrel \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e increased in error, while incoming total longwave radiation appeared to have little influence on ∆\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e (Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eObjective 3: Evaluate the impact of wind speed (\u003c/em\u003e \u003cspan class=\"InlineEquation\"\u003e\u0026nbsp;\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{V}}_{\\varvec{a}}\\)\u003c/span\u003e\u0026nbsp;\u003c/span\u003e \u003cem\u003e) measurement accuracy on differences between the methods\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eFor each globe thermometer method estimating \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e, we investigated error reduction incurred from use using a low-cost \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e sensor and found a slight reduction in RMSE for most methods (Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). On average, replacing the Kestrel impeller with MaRTa\u0026rsquo;s sonic anemometer resulted in a reduction in RMSE of 0.78. The random forest model, which uses a distinct approach from the other methods to determine \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e, showed little improvement from this alteration. This likely reflects the low influence of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e in the model overall. Feature importance values\u0026mdash;indicators of each variable\u0026rsquo;s relative contributions\u0026mdash;were low for both Kestrel \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e (7.55) and MaRTa \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e (5.74), where a lower value indicates less importance in the model. This is compared with \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{g}\\)\u003c/span\u003e\u003c/span\u003e (68.06 for the Kestrel model, 77.93 for the MaRTa model) and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{a}\\)\u003c/span\u003e\u003c/span\u003e (32.81 and 31.28, respectively).\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eImpact of using MaRTa\u0026rsquo;s sonic anemometer in place of the Kestrel 5400\u0026rsquo;s impeller for calculating \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e at the individual point level. The second column shows the change in RMSE associated with each method, while the third column shows the resulting RMSE for swapping in MaRTa wind speed (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e) for each Kestrel method.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"3\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMethod\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eDifference in RMSE using MaRTa \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{V}}_{\\varvec{a}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eResulting RMSE using MaRTa \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{V}}_{\\varvec{a}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eThorsson et al. (\u003cspan class=\"CitationRef\"\u003e2007\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-0.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8.72\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eVanos et al. (\u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-1.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9.74\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOuyang et al. (\u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8.13\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLiu et al. (\u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-1.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7.91\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLin \u0026amp; Matzarakis (\u003cspan class=\"CitationRef\"\u003e2011\u003c/span\u003e\u003cem\u003e)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e--\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e--\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eRandom Forest Model\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e+\u0026thinsp;0.27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6.76\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eChallenges to the Kestrel Approach: Wind Speed, Globe Type Solar Radiation\u003c/h2\u003e \u003cp\u003eFor conventional globe thermometer-based methods, it has long been known that wind speed (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e) dramatically influences \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e calculations, as the conventional method was originally developed for indoor environments with minimal air movement (ISO 7726 \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e1998\u003c/span\u003e). Our analysis showed that a simple and effective way to smooth out Kestrel \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e inaccuracies was by applying a stable, low \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e constant to the Thorsson et al. (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2007\u003c/span\u003e) method, as recommended by Lin and Matzarakis (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2011\u003c/span\u003e) and Chen et al. (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Applying \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e from the MaRTa machine\u0026rsquo;s sonic anemometer to the various Kestrel \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e calculations also improved our models, but to a lesser degree than the Lin and Mazarakis (2011) method. This could be partially due to MaRTa\u0026rsquo;s placement slightly away from the Kestrel, meaning the two instruments may have experienced distinct \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eIn the case of the random forest model, performance did not improve with the use of MaRTa \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e, likely because \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e was not a particularly important feature to the model in the first place. Unlike the other methods, which are based on formulas where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e directly influences the outcome, the random forest model uses data-driven decision trees and does not require any specific predefined relationships between inputs and the response. If a variable does not contribute substantially to improving predictive accuracy\u0026mdash;either because it has limited variation or is less informative relative to other inputs\u0026mdash;the model will assign it less importance. In this case, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{V}_{a}\\)\u003c/span\u003e\u003c/span\u003e contributed less than \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{g}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{a}\\)\u003c/span\u003e\u003c/span\u003e, which more strongly predicted \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e across the dataset.\u003c/p\u003e \u003cp\u003eAnother factor influencing our globe-based \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e results is the globe sensor type. Each of the methods we examined were developed using globes that vary in size, material, and color. The Kestrel 5400 used in this study features a 25 mm black-coated copper globe, which equilibrates to its environment more quickly than larger globes but is also more susceptible to convective error (K\u0026aacute;ntor et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Notably, the Kestrel internally converts its \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{g}\\)\u003c/span\u003e\u003c/span\u003e readings to simulate those of a standard 150 mm globe thermometer. We expect that some of the discrepancy between our results and those reported in other studies arises from mismatches in globe sensor types. For example, Vanos et al. (\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) developed a \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e equation using data from a 150 mm globe. While the Kestrel\u0026rsquo;s output is adjusted to match this standard, its actual globe is much smaller. Therefore, we expect that had we used a physical 150 mm globe, the results from our Vanos et al. (\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) test would have more closely aligned with the six-directional \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eThe radiation decomposition analysis (Objective 3) added to our understanding of observed Kestrel \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e error, indicating that the globe thermometer tended to overpredict \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e under conditions of high shortwave radiation, which is consistent with Chen et al. (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2014\u003c/span\u003e) and Liu et al. \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2022\u003c/span\u003e, but contrary to Thorsson et al. (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). This discrepancy is likely due to our and Chen et al.\u0026rsquo;s (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2014\u003c/span\u003e) use of a black globe sensor, which may \u0026ldquo;overheat\u0026rdquo; and measure hotter temperatures, whereas Thorsson et al. used a gray globe, which is closer in albedo to the human skin. This implies that solar radiation drives differences between globe thermometer and six-directional \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e, suggesting that the black globe thermometer readings become unstable under conditions with high shortwave radiation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eSpatial Aggregation of Globe Thermometer-Derived Mean Radiant Temperature Reveals Thermal Patterns in Heterogeneous Landscapes\u003c/h2\u003e \u003cp\u003eOur results show that Kestrel-derived (globe thermometer) \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e aligns more closely with MaRTa (the gold standard for \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e measurements) when averaging across four to five points per yard, compared to its performance at the single-point level. RMSE values between Kestrel and MaRTa \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e were lower at the yard level, while hyper-local (single point) discrepancies were more pronounced. The front yards we studied ranged from 22 to 246 m\u0026sup2; in area\u0026mdash;roughly the size of a pocket park or green roof. These findings suggest that, in the absence of access to a six-directional platform, averaging multiple measurements across small urban spaces can capture broad thermal patterns with reasonable accuracy.\u003c/p\u003e \u003cp\u003eThe focus of urban ecologists is understanding how ecosystem processes interact with landscape heterogeneity, rather than isolated data points. Our findings indicate that averaging Kestrel \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e yields a practical and reliable approach for microclimate assessment in front yards, even if the Kestrel struggles to pinpoint extremely fine-scale variation. Indeed, Benson et al. (in press), using the same dataset, shows that aggregated \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e measurements detect landscape thermal patterns where air temperature (\u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003ea\u003c/em\u003e\u003c/sub\u003e) lacks sensitivity. These findings highlight that, despite its limitations, the Kestrel can effectively link urban landscape structure to biophysical outcomes when spatially aggregating.\u003c/p\u003e \u003cp\u003eIn addition to spatial aggregation and other previously discussed factors influencing the accuracy of the globe thermometer method, another relevant consideration is the temporal resolution of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e data. Thorsson et al. (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2007\u003c/span\u003e) recommended averaging globe thermometer readings over five-minute periods to reduce inconsistences. In our study, we recorded data at one-minute intervals, consecutively across the yard, such that the final \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e value for each front yard reflects a five-minute average. Thus, our method involves both temporal and spatial aggregation: the former through consecutive point measurements, and the latter through sampling across multiple locations within a yard. While we acknowledge that both types of aggregation contribute to error reduction, we expect spatial aggregation to be the dominant factor. This is because \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e variation across space is likely greater than variation across the short time span needed to complete measurements on a single yard. That being said, we cannot fully disentangle the contributions of spatial versus temporal averaging in this design, and the role of temporal smoothing warrants further attention in future work.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eMachine Learning for Refined Mean Radiant Temperature Field Methods\u003c/h2\u003e \u003cp\u003eOurs is the first study to incorporate a machine learning technique to predict \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e based on Kestrel data. By using MaRTa (six-directional) as a highly-accurate \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e reference, we were able to evaluate the performance of Kestrel metrics and found that this approach holds promise in bolstering Kestrel-based field methods. However, more data across diverse landscapes and under a range of weather conditions would be needed to develop a robust model for use beyond this study. Localized research may benefit from pursuing tailored machine learning models developed under a range of local weather and landscape conditions. Although this approach still relies on a six-directional machine for model training, its role shifts from being the primary \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e data source to serving as a tool for building models at key moments during a field campaign.\u003c/p\u003e \u003cp\u003eWe suspect that our random forest model would be further strengthened by incorporating shade into the model. In machine learning models built to predict human thermal experience, Guo et al. (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) split their micrometeorology data into two sets based on presence of shade, while AlKhaled et al. (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) included shade presence as a variable in their models to predict \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e based on air temperature and a variety of non-physical variables. Because shade presence is a key factor influencing \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{g}\\)\u003c/span\u003e\u003c/span\u003e, combining one of these approaches with ours would almost certainly reduce our model error.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eOur findings reinforce the six-directional method as a critical tool for collecting reliable \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e data while demonstrating how the globe thermometer approach can be adapted for urban ecology research. Fine-scale \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e data provide scientists across disciplines with information about the influence of urban landscape heterogeneity on biophysical outcomes\u0026mdash;critical information for cities striving to adapt to challenges such as densification and extreme weather conditions.\u003c/p\u003e\n\u003cp\u003eAlthough precise \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e measurements currently require specialized six-directional platforms like MaRTa, the accuracy of the globe thermometer method can be improved. Modifying the ISO Standard 7726 equation and averaging clusters of measurement points within a study area can reduce error. Further data are needed to evaluate the efficacy of alternative equations under diverse environmental conditions. However, our results suggest a simple improvement may lie in incorporating a low, constant \u003cem\u003eV\u003c/em\u003e\u003csub\u003e\u003cem\u003ea\u003c/em\u003e\u003c/sub\u003e into the conventional equation, as proposed by Lin and Matzarakis (\u003cspan class=\"CitationRef\"\u003e2011\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eResearchers with access to six-directional \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{T}_{mrt}\\)\u003c/span\u003e\u003c/span\u003e platforms should consider leveraging machine learning to develop localized adaptations of the globe thermometer method. However, ongoing validation against six-directional measurements would be essential, and models could not be applied beyond the range of conditions in which they were trained.\u003c/p\u003e\n\u003cp\u003eA third promising technique has recently emerged that utilizes small cylinders to estimate \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e. Developed by Rykaczewski et al. (\u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e), this cost-effective approach builds on the three-globe radiometer-anemometer concept developed by Nakayoshi et al. (2014), measuring shortwave and longwave radiation fluxes and the convection coefficient to estimate \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e. While this technique shows promise as an affordable and accurate alternative to both six-directional and globe thermometer methods, it is still undergoing refinement (Rykaczewski et al. \u003cspan class=\"CitationRef\"\u003e2025\u003c/span\u003e) and was not suitable for application in our study.\u003c/p\u003e\n\u003cp\u003eAs cities grow more complex and climate pressures intensify, the demand has never been greater for scientists to produce results that are both adaptable and reliable at hyper-local scales. Our work bridges methodological rigor with pragmatic field applications, carving a path forward for urban ecologists seeking to capture spatial and thermal complexity of urban landscapes. Rather than choosing between accuracy and efficiency, we argue for an integrative approach\u0026mdash;where methodological innovation, like machine learning and spatial aggregation, enhances the accessibility and ecological relevance of \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e measurement. By refining the methods we use to sense and study the urban landscape, we sharpen our ability to ask and answer more meaningful questions about how humans and other organisms experience cities.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eDeclaration of Funding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research was funded by the Faculty of Forestry at the University of British Columbia. Meteorological equipment were funded by the USDA Forest Service Denver Urban Field Station.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData will be made available upon request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConceptualization: Aubrey Benson, Melissa McHale. Methodology: Aubrey Benson, John Frank, Melissa McHale. Data collection: Aubrey Benson, John Frank, George Valentine. Analysis: Aubrey Benson, John Frank, George Valentine, Melissa McHale. Writing-original draft: Aubrey Benson. Writing-review and editing: Ben Crawford, John Frank, Ariane Middel, George Valentine, Travis Warziniack, Melissa McHale.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors would like to thank the homeowners who generously allowed us to use their yards for our research, as well as all those who provided local support during our data collection. This work is supported through a collaboration with the USDA Forest Service Denver Urban Field Station, who are studying urban heat in the Front Range of Colorado. Finally, we extend our gratitude to all reviewers for providing invaluable feedback on manuscripts.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eAleksandrowicz, O., \u0026amp; Pearlmutter, D. (2023). The significance of shade provision in reducing street-level summer heat stress in a hot Mediterranean climate. \u003cem\u003eLandscape and Urban Planning\u003c/em\u003e, \u003cem\u003e229\u003c/em\u003e, 104588. https://doi.org/10.1016/J.LANDURBPLAN.2022.104588\u003c/li\u003e\n \u003cli\u003eAlKhaled, S. R., Middel, A., Shaeri, P., Buo, I., \u0026amp; Schneider, F. A. (2024). WebMRT: An online tool to predict summertime mean radiant temperature using machine learning. \u003cem\u003eSustainable Cities and Society\u003c/em\u003e, \u003cem\u003e115\u003c/em\u003e. https://doi.org/10.1016/j.scs.2024.105861\u003c/li\u003e\n \u003cli\u003eBao, Y., Gao, M., Luo, D., \u0026amp; Zhou, X. (2022). The Influence of Plant Community Characteristics in Urban Parks on the Microclimate. \u003cem\u003eForests\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(9). https://doi.org/10.3390/f13091342\u003c/li\u003e\n \u003cli\u003eBruse, M., \u0026amp; Fleer, H. (1998). Simulating surface-plant-air interactions inside urban environments with a three dimensional numerical model. In \u003cem\u003eEnvironmental Modelling \u0026amp; Software\u003c/em\u003e (Vol. 13). http://www.geographie.ruhr-\u003c/li\u003e\n \u003cli\u003eCao, J., Zhou, W., Zheng, Z., Ren, T., \u0026amp; Wang, W. (2021). Within-city spatial and temporal heterogeneity of air temperature and its relationship with land surface temperature. \u003cem\u003eLandscape and Urban Planning\u003c/em\u003e, \u003cem\u003e206\u003c/em\u003e. https://doi.org/10.1016/j.landurbplan.2020.103979\u003c/li\u003e\n \u003cli\u003eChen, Y. C., Lin, T. P., \u0026amp; Matzarakis, A. (2014). Comparison of mean radiant temperature from field experiment and modelling: a case study in Freiburg, Germany. \u003cem\u003eTheoretical and Applied Climatology\u003c/em\u003e, \u003cem\u003e118\u003c/em\u003e(3), 535\u0026ndash;551. https://doi.org/10.1007/s00704-013-1081-z\u003c/li\u003e\n \u003cli\u003eChow, W. T. L., Akbar, S. N. A. B. A., Heng, S. L., \u0026amp; Roth, M. (2016). Assessment of measured and perceived microclimates within a tropical urban forest. \u003cem\u003eUrban Forestry and Urban Greening\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e, 62\u0026ndash;75. https://doi.org/10.1016/j.ufug.2016.01.010\u003c/li\u003e\n \u003cli\u003eCoutts, A. M., White, E. C., Tapper, N. J., Beringer, J., \u0026amp; Livesley, S. J. (2016). Temperature and human thermal comfort effects of street trees across three contrasting street canyon environments. \u003cem\u003eTheoretical and Applied Climatology\u003c/em\u003e, \u003cem\u003e124\u003c/em\u003e(1\u0026ndash;2), 55\u0026ndash;68. https://doi.org/10.1007/s00704-015-1409-y\u003c/li\u003e\n \u003cli\u003eCrank, P. J., Middel, A., Wagner, M., Hoots, D., Smith, M., \u0026amp; Brazel, A. (2020). Validation of seasonal mean radiant temperature simulations in hot arid urban climates. \u003cem\u003eScience of the Total Environment\u003c/em\u003e, \u003cem\u003e749\u003c/em\u003e. https://doi.org/10.1016/j.scitotenv.2020.141392\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eFort Collins Facts\u003c/em\u003e. (n.d.). City of Fort Collins. https://www.fcgov.com/visitor/fcfacts (accessed on 12 January 2023)\u003c/li\u003e\n \u003cli\u003eGuo, R., Yang, B., Guo, Y., Li, H., Li, Z., Zhou, B., Hong, B., \u0026amp; Wang, F. (2024). Machine learning-based prediction of outdoor thermal comfort: Combining Bayesian optimization and the SHAP model. \u003cem\u003eBuilding and Environment\u003c/em\u003e, \u003cem\u003e254\u003c/em\u003e. https://doi.org/10.1016/j.buildenv.2024.111301\u003c/li\u003e\n \u003cli\u003eIbsen, P. (2025). [Unpublished Manuscript]. \u003cem\u003eUnited States Geological Survey.\u0026nbsp;\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003eISO 7726. (1998). Ergonomics of the thermal environment - Instruments for Measuring physical quantities.\u003c/li\u003e\n \u003cli\u003eJin, L., Schubert, S., Salim, M. H., \u0026amp; Schneider, C. (2020). Impact of air conditioning systems on the outdoor thermal environment during summer in Berlin, Germany. \u003cem\u003eInternational Journal of Environmental Research and Public Health\u003c/em\u003e, \u003cem\u003e17\u003c/em\u003e(13), 1\u0026ndash;21. https://doi.org/10.3390/ijerph17134645\u003c/li\u003e\n \u003cli\u003eK\u0026aacute;ntor, N., Kov\u0026aacute;cs, A., \u0026amp; Lin, T. P. (2015). Looking for simple correction functions between the mean radiant temperature from the \u0026ldquo;standard black globe\u0026rdquo; and the \u0026ldquo;six-directional\u0026rdquo; techniques in Taiwan. \u003cem\u003eTheoretical and Applied Climatology\u003c/em\u003e, \u003cem\u003e121\u003c/em\u003e(1\u0026ndash;2), 99\u0026ndash;111. https://doi.org/10.1007/s00704-014-1211-2\u003c/li\u003e\n \u003cli\u003eLachapelle, J. A., Krayenhoff, E. S., Middel, A., Meltzer, S., Broadbent, A. M., \u0026amp; Georgescu, M. (2022). A microscale three-dimensional model of urban outdoor thermal exposure (TUF-Pedestrian). \u003cem\u003eInternational Journal of Biometeorology\u003c/em\u003e, \u003cem\u003e66\u003c/em\u003e(4), 833\u0026ndash;848. https://doi.org/10.1007/s00484-022-02241-1\u003c/li\u003e\n \u003cli\u003eA. Liaw and M. Wiener (2002. Classification and Regression by randomForest. R News 2(3), 18-22.\u003c/li\u003e\n \u003cli\u003eLin, T. P., \u0026amp; Matzarakis A. (2011). Estimation of outdoor mean radiant temperature by field experiment and modelling for human-biometeorology use. \u003cem\u003eThe 11th Annual Meeting of the European Meteorological Society\u003c/em\u003e.\u003c/li\u003e\n \u003cli\u003eLindberg, F., Holmer, B., \u0026amp; Thorsson, S. (2008). SOLWEIG 1.0 - Modelling spatial variations of 3D radiant fluxes and mean radiant temperature in complex urban settings. \u003cem\u003eInternational Journal of Biometeorology\u003c/em\u003e, \u003cem\u003e52\u003c/em\u003e(7), 697\u0026ndash;713. https://doi.org/10.1007/s00484-008-0162-7\u003c/li\u003e\n \u003cli\u003eLiu, K., You, W., Chen, X., \u0026amp; Liu, W. (2022). Study on the Influence of Globe Thermometer Method on the Accuracy of Calculating Outdoor Mean Radiant Temperature and Thermal Comfort. \u003cem\u003eAtmosphere\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(5). https://doi.org/10.3390/atmos13050809\u003c/li\u003e\n \u003cli\u003eMatzarakis, A., Rutz, F., \u0026amp; Mayer, H. (2007). Modelling radiation fluxes in simple and complex environments - Application of the RayMan model. \u003cem\u003eInternational Journal of Biometeorology\u003c/em\u003e, \u003cem\u003e51\u003c/em\u003e(4), 323\u0026ndash;334. https://doi.org/10.1007/s00484-006-0061-8\u003c/li\u003e\n \u003cli\u003eMatzarakis, A., Rutz, F., \u0026amp; Mayer, H. (2010). Modelling radiation fluxes in simple and complex environments: Basics of the RayMan model. \u003cem\u003eInternational Journal of Biometeorology\u003c/em\u003e, \u003cem\u003e54\u003c/em\u003e(2), 131\u0026ndash;139. https://doi.org/10.1007/s00484-009-0261-0\u0026nbsp;\u003cem\u003eEcosystem Health and Sustainability\u003c/em\u003e (Vol. 4, Issue 5, pp. 115\u0026ndash;131). Taylor and Francis Ltd. https://doi.org/10.1080/20964129.2018.1480905\u003c/li\u003e\n \u003cli\u003eMiddel, A., H\u0026auml;b, K., Brazel, A. J., Martin, C. A., \u0026amp; Guhathakurta, S. (2014). Impact of urban form and design on mid-afternoon microclimate in Phoenix Local Climate Zones. In \u003cem\u003eLandscape and Urban Planning\u003c/em\u003e (Vol. 122, p. 16). Elsevier BV. https://doi.org/10.1016/j.landurbplan.2013.11.004\u003c/li\u003e\n \u003cli\u003eMiddel, A., Selover, N., Hagen, B., \u0026amp; Chhetri, N. (2016). Impact of shade on outdoor thermal comfort\u0026mdash;a seasonal field study in Tempe, Arizona. \u003cem\u003eInternational Journal of Biometeorology\u003c/em\u003e, \u003cem\u003e60\u003c/em\u003e(12), 1849\u0026ndash;1861. https://doi.org/10.1007/s00484-016-1172-5\u003c/li\u003e\n \u003cli\u003eMiddel, A., \u0026amp; Krayenhoff, E. S. (2019). Micrometeorological determinants of pedestrian thermal exposure during record-breaking heat in Tempe, Arizona: Introducing the MaRTy observational platform. \u003cem\u003eThe Science of the Total Environment\u003c/em\u003e, \u003cem\u003e687\u003c/em\u003e, 137\u0026ndash;151. https://doi.org/10.1016/j.scitotenv.2019.06.085\u003c/li\u003e\n \u003cli\u003eNakayoshi, M., Kanda, M., \u0026amp; de Dear, R. (2015). Globe Anemo-radiometer. \u003cem\u003eBoundary-Layer Meteorology\u003c/em\u003e, \u003cem\u003e155\u003c/em\u003e(2), 209\u0026ndash;227. https://doi.org/10.1007/s10546-014-0003-7\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eNOAA Online Weather Data\u003c/em\u003e (2023). National Oceanic and Atmospheric Administration. https://www.weather.gov/wrh/climate?wfo=bou (accessed on 27 November 2023).\u003c/li\u003e\n \u003cli\u003eOuyang, W., Liu, Z., Lau, K., Shi, Y., \u0026amp; Ng, E. (2022). Comparing different recalibrated methods for estimating mean radiant temperature in outdoor environment. \u003cem\u003eBuilding and Environment\u003c/em\u003e, \u003cem\u003e216\u003c/em\u003e. https://doi.org/10.1016/j.buildenv.2022.109004\u003c/li\u003e\n \u003cli\u003eRStudio Team. (2020). \u003cem\u003eRStudio: Integrated Development for R.\u0026nbsp;\u003c/em\u003eRStudio, PBC, Boston, MA.\u003c/li\u003e\n \u003cli\u003eRykaczewski, K., Joshi, A., Viswanathan, S. H., Guddanti, S. S., Sadeghi, K., Gupta, M., Jaiswal, A. K., Kompally, K., Pathikonda, G., Barlett, R., Vanos, J. K., \u0026amp; Middel, A. (2024). A simple three-cylinder radiometer and low-speed anemometer to characterize human extreme heat exposure. \u003cem\u003eInternational Journal of Biometeorology\u003c/em\u003e. https://doi.org/10.1007/s00484-024-02646-0\u003c/li\u003e\n \u003cli\u003eRykaczewski, K., Joshi, A., Viswanathan, S. H., Parkerson, E., Gupta, M., Park, M., ... \u0026amp; Middel, A. (2025). Advanced human heat exposure sensing using two cylinder anemometer and radiometer: introducing CARla. International Journal of Biometeorology, 1-14.\u003c/li\u003e\n \u003cli\u003eThorsson, S., Lindberg, F., Eliasson, I., \u0026amp; Holmer, B. (2007). Different methods for estimating the mean radiant temperature in an outdoor urban setting. \u003cem\u003eInternational Journal of Climatology\u003c/em\u003e, \u003cem\u003e27\u003c/em\u003e(14), 1983\u0026ndash;1993. https://doi.org/10.1002/JOC.1537\u003c/li\u003e\n \u003cli\u003eTurner, K. V., Rogers, M. L., Zhang, Y., Middel, A., Schneider, F. A., Oc\u0026oacute;n, J. P., Seeley, M., \u0026amp; Dialesandro, J. (2022). More than surface temperature: mitigating thermal exposure in hyper-local land system. \u003cem\u003eJournal of Land Use Science\u003c/em\u003e, \u003cem\u003e17\u003c/em\u003e(1), 79\u0026ndash;99. https://doi.org/10.1080/1747423X.2021.2015003\u003c/li\u003e\n \u003cli\u003eUniversity of Utah, Department of Atmospheric Sciences. (2024). \u003cem\u003eMesoWest weather data\u003c/em\u003e (Station KFNL). https://mesowest.utah.edu/cgi-bin/droman/meso_base_dyn.cgi?stn=KFNL\u003c/li\u003e\n \u003cli\u003eVanos, J. K., Rykaczewski, K., Middel, A., Vecellio, D. J., Brown, R. D., \u0026amp; Gillespie, T. J. (2021). Improved methods for estimating mean radiant temperature in hot and sunny outdoor settings. \u003cem\u003eInternational Journal of Biometeorology\u003c/em\u003e, \u003cem\u003e65\u003c/em\u003e(6), 967\u0026ndash;983. https://doi.org/10.1007/s00484-021-02131-y\u003c/li\u003e\n \u003cli\u003eWang, J., Zhou, W., \u0026amp; Jiao, M. (2022). Location matters: planting urban trees in the right places improves cooling. \u003cem\u003eFrontiers in Ecology and the Environment\u003c/em\u003e, \u003cem\u003e20\u003c/em\u003e(3), 147\u0026ndash;151. https://doi.org/10.1002/fee.2455\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eWater Sources\u003c/em\u003e. (n.d.). City of Fort Collins. https://www.fcgov.com/utilities/water-sources. Accessed on 13 February 2023.\u003c/li\u003e\n \u003cli\u003eWickham H, Averick M, Bryan J, Chang W, McGowan LD, Fran\u0026ccedil;ois R, Grolemund G, Hayes A, Henry L, Hester J, Kuhn M, Pedersen TL, Miller E, Bache SM, M\u0026uuml;ller K, Ooms J, Robinson D, Seidel DP, Spinu V, Takahashi K, Vaughan D, Wilke C, Woo K, Yutani H (2019). \u0026ldquo;Welcome to the tidyverse.\u0026rdquo; Journal of Open Source Software, *4*(43), 1686. doi:10.21105/joss.01686 \u0026lt;https://doi.org/10.21105/joss.01686\u0026gt;.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-biometeorology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ijbm","sideBox":"Learn more about [International Journal of Biometeorology](http://link.springer.com/journal/484)","snPcode":"484","submissionUrl":"https://www.editorialmanager.com/ijbm/default2.aspx","title":"International Journal of Biometeorology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6432428/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6432428/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eExtreme summertime heat is an increasing challenge for cities, highlighting the need for accurate, spatially meaningful methods to measure and map heat in ways that reflect human thermal experiences and inform land management decisions. Mean radiant temperature (\u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e) is a key metric for assessing urban heat at hyper-local scales, yet its measurement remains technically challenging. In this study, we apply the six-directional, gold standard method for measuring \u003cem\u003eT\u003c/em\u003e\u003csub\u003e\u003cem\u003emrt\u003c/em\u003e\u003c/sub\u003e with globe thermometer-based approaches across multiple levels of spatial aggregation and develop a novel machine learning model trained on field data. Data were collected in a semi-arid city in Colorado, USA, over two summers. Using measurements from residential parcels, we show that aggregated globe thermometer data\u0026mdash;collected using a low-cost, accessible sensor\u0026mdash;can capture thermal patterns across landscapes with reasonable accuracy. Our findings also indicate that machine learning, combining six-directional and globe thermometer data, offers promising potential for improving both measurement accuracy and efficiency. These findings are particularly relevant for planners working at the scale of parcels, where heat adaptation strategies are commonly applied, and especially insightful for semi-arid cities and those increasingly experiencing arid summer conditions due to climate change. This work advances practical methods for integrating human thermal comfort into landscape planning for climate-resilient urban design.\u003c/p\u003e","manuscriptTitle":"Bridging accuracy and efficiency: Advancing mean radiant temperature measurement in Urban Ecology","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-05 12:23:19","doi":"10.21203/rs.3.rs-6432428/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"","date":"2025-05-20T15:10:08+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-04-29T07:51:59+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-04-16T08:22:53+00:00","index":"","fulltext":""},{"type":"submitted","content":"International Journal of Biometeorology","date":"2025-04-15T23:09:19+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-biometeorology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ijbm","sideBox":"Learn more about [International Journal of Biometeorology](http://link.springer.com/journal/484)","snPcode":"484","submissionUrl":"https://www.editorialmanager.com/ijbm/default2.aspx","title":"International Journal of Biometeorology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"e3078c75-69b8-4bd6-8bcf-dc75c42ce5f1","owner":[],"postedDate":"May 5th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-10-20T15:59:29+00:00","versionOfRecord":{"articleIdentity":"rs-6432428","link":"https://doi.org/10.1007/s00484-025-03038-8","journal":{"identity":"international-journal-of-biometeorology","isVorOnly":false,"title":"International Journal of Biometeorology"},"publishedOn":"2025-10-16 15:56:50","publishedOnDateReadable":"October 16th, 2025"},"versionCreatedAt":"2025-05-05 12:23:19","video":"","vorDoi":"10.1007/s00484-025-03038-8","vorDoiUrl":"https://doi.org/10.1007/s00484-025-03038-8","workflowStages":[]},"version":"v1","identity":"rs-6432428","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6432428","identity":"rs-6432428","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}