Extended Abstract
Introduction
Climate change, driven by anthropogenic activities and amplified by urban morphology, is profoundly impacting human life. The increase in global temperatures, coupled with the urban heat island effect, is leading to the emergence of more frequent and intense heat-related risks. Bioclimatic indices, based on meteorological variables like temperature, humidity, and wind, are essential for evaluating thermal stress in urban environments and guiding sustainable management strategies. In the context of accelerating climate change, the study of future bioclimatic comfort under various emission scenarios (optimistic, moderate, and pessimistic) has become increasingly critical. While numerous studies have examined historical and future urban climatic comfort, many have focused on mean trends in time series data. However, climate change is known to increase the frequency and intensity of extreme weather events. Therefore, analyzing the impact of extreme high and low values of climatic variables on comfort indices is of paramount importance. Unlike traditional regression methods, quantile regression offers a powerful tool to investigate patterns of change across the entire distribution of a dataset—particularly the extremes. This approach can reveal how different quantiles (e.g., low, medium, and high) of independent climatic variables influence the dependent comfort index, providing crucial insights into the effects of severe weather (both hot and cold). Mazandaran Province, is a major tourist destination, renowned for its coastal and natural attractions. Babolsar, with the longest coastline in northern Iran, is a prime location for coastal tourism. As global temperatures rise, studying thermal stress and climatic comfort in these tourism-dependent areas becomes essential. Despite numerous studies on climate change and thermal comfort indices, a comprehensive, localized analysis of future thermal stress and comfort, especially for Iran's northern coastal regions, remains scarce. This study aims to fill this gap by applying multiple Shared Socioeconomic Pathways (SSPs) and quantile regression to analyze changes in bioclimatic comfort indices in Babolsar for both historical and future periods, offering an innovative and practical approach for adaptation planning.

Material and Methods
The study focuses on Babolsar (36°43'N, 52°39'30"E), a coastal city on the southern Caspian Sea with a humid temperate climate, average annual temperature of 17.8°C, and about 939 mm of precipitation. Two primary datasets were employed: (1) historical daily meteorological data (temperature, humidity, wind speed) from the Babolsar synoptic station (1987–2014), and (2) future climate projections (2020–2099) from the Coupled Model Intercomparison Project Phase 6 (CMIP6) under three SSP scenarios—optimistic (SSP1-2.6), moderate (SSP2-4.5), and pessimistic (SSP5-8.5). Among eight CMIP6 models evaluated for accuracy in Iran, the GFDL-ESM4 model was selected for its complete data, relatively high accuracy (R²=0.7), and appropriate resolution. Bias correction and downscaling to the station level were performed using bilinear interpolation in R. Two comfort indices were calculated: (a) Effective Temperature (ET), which integrates temperature and humidity (ET = T - 0.4(T - 10)(1 - RH/100)), with categories from "Very Hot" to "Very Cold"; and (b) Baker's Bioclimatic Index (Cp), designed for tourist comfort (Cp = (0.26 + 0.34V^0.632)(36.5 - T), where V is wind speed), with categories from "Hot, Unpleasant" to "Unbearable, Very Cold". The core analytical method was quantile regression, which estimates conditional quantiles (e.g., 5th, 50th, 95th percentiles) rather than just the conditional mean. This allows detection of significant trends across the full distribution (0.01 to 0.99 quantiles) for both historical and future periods.

Results and Discussion
The resuts revealed between 1987 and 2014, mean daily temperatures rose significantly across all quantiles, with the most pronounced warming (0.048°C/year) occurring on the coldest days. Concurrently, relative humidity declined—especially at lower quantiles—while wind speeds increased. The ET index refelected this warming trend, again strongest for cooler days, suggesting a reduction in cold extremes. Interestingly, Baker’s Index showed a historical increase in lower and medium quantiles, indicating a trend toward cooler days, which appeared beneficial. Future projections, however, diverge sharply by emissions scenario. Under the optimistic SSP1-2.6, temperatures continue rising in the near


future (2021–2060) but at a slower rate, with no significant trend expected by 2061–2100. In contrast, the moderate SSP2-4.5 and especially the pessimistic SSP5-8.5 project intensified warming, particularly for lower quantiles. Humidity projections are mixed under optimistic and moderate scenarios (with some high-quantile increases), but the pessimistic scenario foresees significant decreases, especially for medium and high quantiles. Wind speed trends are negligible under SSP1-2.6, slightly negative under SSP2-4.5, and complex under SSP5-8.5 (decreasing near-term, increasing far-term but with weaker slopes than historically). For the ET index, SSP1-2.6 shows near-term warming but far-term stabilization. Under SSP2-4.5 and particularly SSP5-8.5, strong increasing trends emerge across all quantiles, with far-future slopes reaching up to 0.56°C/year—signaling a dramatic rise in hot days. Starkly, Baker’s Index—historically trending toward cooler days—reverses under all future scenarios, showing decreasing trends across most quantiles, indicating a shift toward warmer conditions. This decline intensifies under SSP5-8.5 (slopes up to -0.062), implying a notable reduction in cool and cold days. Quantile regression revealed that ordinary least squares (OLS) mean trends often misrepresented median or extreme trends, underscoring the insufficiency of mean-only analyses. Under SSP5-8.5, upper ET quantiles intensify fastest near-term (0.3/year), while lower quantiles also rise dramatically far-term (0.07/year), eroding cooler days. For the comfort parameter (Cp), negative slopes under SSP5-8.5 are strongest for upper quantiles far-term, signaling a near-total loss of very cool days.

Conclusion
This study investigated historical and future climatic comfort using meteorological data and advanced statistical methods. Two primary datasets were utilized: historical daily weather records (temperature, humidity, wind speed) from 1987-2014 obtained from the local synoptic station, and future climate projections for 2020-2099 derived from the CMIP6 database. Among eight assessed models, the GFDL-ESM4 was selected for its accuracy (R²=0.7) and data completeness. Bias correction and downscaling were performed using bilinear interpolation in R. To evaluate human comfort, two bioclimatic indices were calculated. The Effective Temperature (ET) integrates temperature and humidity to classify comfort from "Very Hot" to "Very Cold." Baker's Bioclimatic Index (Cp) assesses tourist comfort by incorporating wind speed and temperature, with categories ranging from "Hot, Unpleasant" to "Unbearable, Very Cold." These indices were applied to both historical and future data under three Shared Socioeconomic Pathways (SSPs): a low-emissions (SSP1-2.6), moderate (SSP2-4.5), and high-emissions (SSP5-8.5) scenario. The core analytical method was quantile regression, which estimates relationships across different percentiles (e.g., 5th, 50th, 95th) of the dependent variable distribution. This approach is superior to ordinary least squares for climate studies because it captures changes in extreme values—critical for assessing climate impacts. The analysis identified significant trends across quantiles (0.01 to 0.99) for climatic variables and comfort indices over both historical and future periods, enabling a comprehensive understanding of how both average conditions and extremes may shift under different emission pathways.