以氣象站數據建立營建工地熱暴露評估即時校正模式之研究

Abstract

本研究旨在建立一套將氣象站數據轉化為營建工地綜合溫度熱指數（WBGT）之即時校正模式。近年因氣候變遷與氣溫持續升高，戶外作業族群面臨更高的熱壓力暴露風險。其中營造業因長時間曝曬與高體力勞動，熱傷害發生率遠高於其他產業，長期暴露不僅會引發中暑等急性疾病，更與慢性腎臟病等長期健康損害有顯著關聯。 然而，現行預警機制普遍存在雙重局限：首先在指標代表性上，多數指引僅採用涵蓋氣溫與濕度的熱指數（HI），無法全面反映風速與輻射對人體熱平衡的綜合影響；其次在空間代表性上，直接以氣象站的觀測值代表工地現場，亦忽略了施工現地因材料蓄熱與通風受阻所造成的局部氣候落差。這兩大問題，導致營建工地的現場熱風險遭到嚴重低估。 本研究採取物理模型轉換與現場實測校正之混合式策略。首先，整合工地現場儀器量測資料與中央氣象署鄰近測站數據，比較工地現場之HI、氣象站之HI，以及工地現場之WBGT、氣象站推估之WBGT之差異。接著，本研究採取兩階段校正策略：首先透過Liljegren物理能量平衡模型，結合工地經緯度與時序參數，計算出氣象站推估之WBGT，並進一步以現場實測資料建立迴歸校正式，使氣象站推估結果更貼近營建工地現場之實際熱暴露情形。 研究結果顯示，工地現場之HI較氣象站之HI約高2°C，且兩者在熱危害分級上呈現明顯差異；氣象站之HI多集中於第1至第2級，而工地現場之HI則主要落於第3級。WBGT 分析亦顯示，工地現場之WBGT較氣象站推估之WBGT約高1°C，雖然兩者觀察值範圍相近，但此差距已足以使部分資料落入較高風險分級。統計分析顯示，HI與WBGT在氣象站資料與工地現場資料之間皆具有顯著差異（p < 0.05），說明僅依賴氣象站資料可能造成熱危害風險低估，證實現地校正之必要性。 在模型建立方面，Liljegren物理模型推估之氣象站推估WBGT與工地現場之WBGT具有高度相關性（R²=0.9524），顯示其具備良好的預測潛力。然而，模型仍存在約1°C之低估誤差，可能影響熱危害分級判斷。因此，本研究進一步導入現地校正機制，並將此兩階段架構建置為Excel即時估計工具，以提升氣象站資料應用於營建工地熱危害管理之準確性與實務可行性。

This study aims to develop a real-time correction model that converts meteorological data into the Wet Bulb Globe Temperature (WBGT) specific to construction sites. Due to climate change and continuously rising temperatures in recent years, outdoor worker populations face higher risks of heat stress exposure. Among these, the construction industry suffers from a significantly higher incidence of heat injury than other industries due to prolonged sun exposure and high physical labor. Prolonged exposure not only triggers acute diseases such as heat stroke but also correlates significantly with long-term health damage such as chronic kidney disease.

However, existing early-warning mechanisms generally suffer from dual limitations. First, in terms of index representativeness, guidelines in Taiwan only adopt the Heat Index (HI), which covers temperature and humidity, failing to comprehensively reflect the combined effects of wind speed and radiation on human thermal balance. Second, in terms of spatial representativeness, directly using weather station observations to represent construction sites also overlooks the local climate gaps caused by material heat retention and obstructed ventilation at the construction sites. These two major problems lead to a severe underestimation of the onsite heat risks at construction sites.

This study adopted a hybrid strategy combining physical model conversion with field-based correction. Field measurements from a construction site were integrated with meteorological data from a nearby Central Weather Administration station to compare differences between construction-site HI and weather-station HI, as well as construction-site WBGT and weather-station-estimated WBGT. The Liljegren physical energy balance model was then applied to estimate weather-station-estimated WBGT using weather station data, site-specific geographic coordinates, and temporal parameters. A regression-based correction equation was further developed using field measurements to adjust the weather-station-estimated WBGT toward the actual thermal exposure conditions at the construction site.

The results showed that construction-site HI was approximately 2°C higher than weather-station HI, resulting in clear differences in heat hazard classification. Weather-station HI values were mainly distributed in Categories 1 and 2, whereas construction-site HI values were predominantly classified as Category 3. WBGT analysis also showed that construction-site WBGT was approximately 1°C higher than weather-station-estimated WBGT. Although the observed ranges of the two WBGT datasets were comparable, this difference was sufficient to shift part of the results into a higher risk category. Statistical analyses confirmed significant differences between construction-site and weather-station values for both HI and WBGT (p < 0.05), indicating that reliance solely on weather station data may underestimate heat hazard risks at construction sites.

For model development, the weather-station-estimated WBGT derived using the Liljegren physical model showed a strong correlation with construction-site WBGT (R² = 0.9524), demonstrating its strong predictive potential. However, an underestimation of approximately 1°C remained, which may affect heat hazard classification. Therefore, this study further incorporated field-based correction and implemented the two-stage framework into an Excel-based real-time estimation tool to improve the accuracy and practical applicability of weather station data for construction site heat hazard management.