Lewis P. Blunn, Flynn Ames, Hannah L. Croad, Adam Gainford, Ieuan Higgs, Mathew Lipson, Chun Hay Brian Lo
{"title":"城市热浪温度预测的机器学习纠偏和降尺度(从千米尺度到公顷尺度","authors":"Lewis P. Blunn, Flynn Ames, Hannah L. Croad, Adam Gainford, Ieuan Higgs, Mathew Lipson, Chun Hay Brian Lo","doi":"10.1002/met.2200","DOIUrl":null,"url":null,"abstract":"<p>The urban heat island (UHI) effect exacerbates near-surface air temperature (<i>T</i>) extremes in cities, with negative impacts for human health, building energy consumption and infrastructure. Using conventional weather models, it is both difficult and computationally expensive to simulate the complex processes controlling neighbourhood-scale variation of <i>T</i>. We use machine learning (ML) to bias correct and downscale <i>T</i> predictions made by the Met Office operational regional forecast model (UKV) to 100 m horizontal grid length over London, UK. A set of ML models (random forest, XGBoost, multiplayer perceptron) are trained using citizen weather station observations and UKV variables from eight heatwaves, along with high-resolution land cover data. The ML models improve the <i>T</i> mean absolute error (MAE) by up to 0.12°C (11%) relative to the UKV. They also improve the UHI diurnal and spatial representation, reducing the UHI profile MAE from 0.64°C (UKV) to 0.15°C. A multiple linear regression performs almost as well as the ML models in terms of <i>T</i> MAE, but cannot match the UHI bias correction performance of the ML models, only reducing the UHI profile MAE to 0.49°C. UKV latent heat flux is found to be the most important predictor of <i>T</i> bias. It is demonstrated that including more heatwaves and observation sites in training would reduce overfitting and improve ML model performance.</p>","PeriodicalId":49825,"journal":{"name":"Meteorological Applications","volume":null,"pages":null},"PeriodicalIF":2.3000,"publicationDate":"2024-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/met.2200","citationCount":"0","resultStr":"{\"title\":\"Machine learning bias correction and downscaling of urban heatwave temperature predictions from kilometre to hectometre scale\",\"authors\":\"Lewis P. Blunn, Flynn Ames, Hannah L. Croad, Adam Gainford, Ieuan Higgs, Mathew Lipson, Chun Hay Brian Lo\",\"doi\":\"10.1002/met.2200\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p>The urban heat island (UHI) effect exacerbates near-surface air temperature (<i>T</i>) extremes in cities, with negative impacts for human health, building energy consumption and infrastructure. Using conventional weather models, it is both difficult and computationally expensive to simulate the complex processes controlling neighbourhood-scale variation of <i>T</i>. We use machine learning (ML) to bias correct and downscale <i>T</i> predictions made by the Met Office operational regional forecast model (UKV) to 100 m horizontal grid length over London, UK. A set of ML models (random forest, XGBoost, multiplayer perceptron) are trained using citizen weather station observations and UKV variables from eight heatwaves, along with high-resolution land cover data. The ML models improve the <i>T</i> mean absolute error (MAE) by up to 0.12°C (11%) relative to the UKV. They also improve the UHI diurnal and spatial representation, reducing the UHI profile MAE from 0.64°C (UKV) to 0.15°C. A multiple linear regression performs almost as well as the ML models in terms of <i>T</i> MAE, but cannot match the UHI bias correction performance of the ML models, only reducing the UHI profile MAE to 0.49°C. UKV latent heat flux is found to be the most important predictor of <i>T</i> bias. It is demonstrated that including more heatwaves and observation sites in training would reduce overfitting and improve ML model performance.</p>\",\"PeriodicalId\":49825,\"journal\":{\"name\":\"Meteorological Applications\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":2.3000,\"publicationDate\":\"2024-05-17\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://onlinelibrary.wiley.com/doi/epdf/10.1002/met.2200\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Meteorological Applications\",\"FirstCategoryId\":\"89\",\"ListUrlMain\":\"https://onlinelibrary.wiley.com/doi/10.1002/met.2200\",\"RegionNum\":4,\"RegionCategory\":\"地球科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"METEOROLOGY & ATMOSPHERIC SCIENCES\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Meteorological Applications","FirstCategoryId":"89","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1002/met.2200","RegionNum":4,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"METEOROLOGY & ATMOSPHERIC SCIENCES","Score":null,"Total":0}
引用次数: 0
摘要
城市热岛(UHI)效应加剧了城市近地面的极端气温(T),对人类健康、建筑能耗和基础设施产生了负面影响。使用传统的天气模型模拟控制邻域尺度气温变化的复杂过程既困难又耗费计算成本。我们使用机器学习(ML)对英国伦敦上空 100 米水平网格长度的气象局业务区域预报模型(UKV)的气温预测进行偏差校正和降尺度预测。一组 ML 模型(随机森林、XGBoost、多人感知器)是利用市民气象站观测数据和八次热浪中的 UKV 变量以及高分辨率土地覆盖数据进行训练的。相对于 UKV,ML 模型将 T 平均绝对误差 (MAE) 最多提高了 0.12°C(11%)。它们还改善了 UHI 的昼夜和空间代表性,将 UHI 剖面 MAE 从 0.64°C(UKV)降至 0.15°C。就 T MAE 而言,多元线性回归与 ML 模型的表现几乎一样好,但无法与 ML 模型的 UHI 偏差校正性能相媲美,只能将 UHI 剖面 MAE 降低到 0.49°C。研究发现,UKV 潜热通量是预测 T 偏差的最重要因素。结果表明,在训练中加入更多的热浪和观测点可以减少过拟合,提高 ML 模型的性能。
Machine learning bias correction and downscaling of urban heatwave temperature predictions from kilometre to hectometre scale
The urban heat island (UHI) effect exacerbates near-surface air temperature (T) extremes in cities, with negative impacts for human health, building energy consumption and infrastructure. Using conventional weather models, it is both difficult and computationally expensive to simulate the complex processes controlling neighbourhood-scale variation of T. We use machine learning (ML) to bias correct and downscale T predictions made by the Met Office operational regional forecast model (UKV) to 100 m horizontal grid length over London, UK. A set of ML models (random forest, XGBoost, multiplayer perceptron) are trained using citizen weather station observations and UKV variables from eight heatwaves, along with high-resolution land cover data. The ML models improve the T mean absolute error (MAE) by up to 0.12°C (11%) relative to the UKV. They also improve the UHI diurnal and spatial representation, reducing the UHI profile MAE from 0.64°C (UKV) to 0.15°C. A multiple linear regression performs almost as well as the ML models in terms of T MAE, but cannot match the UHI bias correction performance of the ML models, only reducing the UHI profile MAE to 0.49°C. UKV latent heat flux is found to be the most important predictor of T bias. It is demonstrated that including more heatwaves and observation sites in training would reduce overfitting and improve ML model performance.
期刊介绍:
The aim of Meteorological Applications is to serve the needs of applied meteorologists, forecasters and users of meteorological services by publishing papers on all aspects of meteorological science, including:
applications of meteorological, climatological, analytical and forecasting data, and their socio-economic benefits;
forecasting, warning and service delivery techniques and methods;
weather hazards, their analysis and prediction;
performance, verification and value of numerical models and forecasting services;
practical applications of ocean and climate models;
education and training.