Mobile monitoring for urban air temperature at a microscale in Mississauga, Ontario, Canada, a unique region dominated by suburban sprawl, was completed via cycling. We sampled seven pre-determined routes across Mississauga, resulting in 3144 min of air temperature measurements between July and August 2022. We developed land use regression models to determine if stopping for 5-min periods every 20 min was beneficial compared to continuous collection. The model generated from the data captured while moving demonstrated the best performance, explaining 80 % of the spatial variability of air temperature in Mississauga. Regression kriging addressed issues of spatial autocorrelation in linear models, improving predictive performance (CV R2 = 0.83, CV RMSE = 0.95 °C, CV MAE = 0.74 °C). We used the regression kriging model from the data captured while moving to predict average, maximum, and 95th percentile air temperature at a 20 m-by-20 m spatial resolution across Mississauga. We also conducted one-way analysis of variance (ANOVA) tests between air temperature and marginalization and found that areas with higher levels of marginalization experience different air temperatures compared to areas with lower levels of marginalization. Our study supports mobile monitoring to access urban air temperature and improve predictive performance by integrating regression kriging.

中文翻译：

---

模拟郊区扩张中城市气温的微观空间分布


加拿大安大略省密西沙加是一个以郊区扩张为主的独特地区，通过骑自行车完成了对城市空气温度的微尺度移动监测。我们对密西沙加的 7 条预先确定的路线进行了采样，在 2022 年 7 月至 8 月期间测量了 3144 分钟的气温。我们开发了土地利用回归模型，以确定与连续收集相比，每 20 分钟停止 5 分钟是否有益。根据移动时捕获的数据生成的模型表现出最佳性能，解释了密西沙加 80% 的气温空间变异。回归克里金法解决了线性模型中的空间自相关问题，提高了预测性能（CV R2 = 0.83、CV RMSE = 0.95 °C、CV MAE = 0.74 °C）。我们根据移动过程中捕获的数据使用回归克里金模型来预测密西沙加 20 m x 20 m 空间分辨率的平均气温、最高气温和第 95 个百分点的气温。我们还在气温和边缘化之间进行了单向方差分析 (ANOVA) 测试，发现边缘化水平较高的地区与边缘化水平较低的地区相比，经历了不同的气温。我们的研究支持移动监测来获取城市气温并通过集成回归克里金法来提高预测性能。