Impact of biometeorological variable uncertainties in rational planning: examples from a tropical urban environment, Singapore.

Heat stress and thermal performance indices are used to guide smart city planning based on whether they exceed predefined threshold values determined from numerical thermal sensation scales. The numerical thermal sensation scales are usually generated from inferential statistics (usually a linear regression) based on heat stress index values and air temperature sensation votes gathered during biometeorological field campaigns. However, uncertainties in the input variables and parameters used to calculate the heat stress values are often overlooked, potentially leading to inaccurate threshold values and hence planning recommendations. Additionally, the choice of regression model is frequently not justified, even though some models are more sensitive to individual variability in input variables than others.

 

This poster assesses the sensitivity of the modified Physiological Equivalent Temperature (mPET) and the Outdoor Thermal Comfort Autonomy (OTCA) indices to uncertainties in body anthropometric and meteorological input variables. Data from a 2018 and 2019 biometeorological survey in the Tanjong Pagar district of Singapore are used. This work compares two classical statistical approaches (ordinary least square regression vs. ordinal logistic regression) to compute numerical boundaries for thermal sensation scales based on the collected air temperature sensation votes and computed heat stress values during the survey. The results show: i) the metabolic heat rate is the main source of uncertainty in mPET; ii) the mPET clothing insulation scheme and body mass index have minimal impact; iii) inaccurate estimates of globe temperature or simultaneous errors in atmospheric humidity (underestimation) and air temperature (overestimation) significantly impact mPET values; and iv) the probabilistic approach is more robust to input data variability than the deterministic approach when calculating thermal sensation scale boundaries.