DSpace community: 遙測科技碩士學位學程

運用機器學習與多來源衛星及地理空間資料進行印度尼西亞廖內南部季節性野火易發性建模;Seasonal Wildfire Susceptibility Modelling Using Machine Learning with Multi-Source Satellite and Geospatial Data in Southern Riau， Indonesia

title: 運用機器學習與多來源衛星及地理空間資料進行印度尼西亞廖內南部季節性野火易發性建模;Seasonal Wildfire Susceptibility Modelling Using Machine Learning with Multi-Source Satellite and Geospatial Data in Southern Riau， Indonesia abstract: 印尼廖內南部的森林與土地火災具有明顯的季節差異，並受到植被、水分、氣象、土地利用及泥炭地條件等多重因素的共同影響。本研究建立全年、旱季及雨季的野火易發性模型，並評估歷史環境變化指標是否能提供額外的預測資訊。研究整合2010至2024年的多來源衛星與地理空間資料，並將其統一至500公尺網格；燒毀區標籤則由MODIS MCD64A1 BurnDate產品建立。當期條件型與條件變化型兩種預測因子配置，分別使用四種機器學習演算法進行比較。模型採用1:5的火災與非火災樣本比例，以及分層70:30訓練–測試資料切分，並以PR-AUC作為主要評估指標。最終選定的模型為全年與旱季的條件變化型LightGBM模型，以及雨季的當期條件型LightGBM模型；其保留測試集的PR-AUC分別為0.921、0.905及0.966。歷史環境變化指標僅小幅提升全年與旱季模型的表現，且未改善雨季模型，顯示其預測效益具有季節差異。2024年的分析結果顯示，全年與旱季易發性地圖具有較高的空間一致性，而雨季的高易發性區域則較為局部。旱季的高與極高易發性區域面積最大，達8，581平方公里，占分析網格總面積的21.49%。SHAP分析顯示，全年模型的主要影響因子為植被狀況、土壤水分趨勢及降水；旱季模型主要受到降水、植被狀況及油棕土地利用背景的影響；雨季模型則主要受到風速、植被狀況及泥炭地背景的影響。使用2024年印尼國家警察之獨立火災紀錄進行驗證的結果顯示，全年、旱季及雨季分別有77.60%、76.79%及78.57%的匹配紀錄位於高與極高易發性區域。研究結果顯示，分季建模能揭示僅使用全年模型時可能被掩蓋的預測差異與空間差異。;Forest and land fires in Southern Riau， Indonesia， exhibit strong seasonal variation and are influenced by multiple interacting factors， including vegetation， moisture， meteorological conditions， land use， and peatland characteristics. This study developed wildfire susceptibility models for the all-season， dry-season， and rainy-season periods and assessed whether historical environmental-change indicators provided additional predictive information. Multi-source satellite and geospatial data from 2010 to 2024 were harmonized to a 500 m grid， while burned-area labels were derived from the MODIS MCD64A1 BurnDate product. Condition-only and condition-change predictor configurations were compared using four machine-learning algorithms. The modelling design employed a 1:5 fire-to-non-fire sampling ratio and a stratified 70:30 train–test split， with PR-AUC as the primary evaluation metric. The selected models were condition-change LightGBM for the all-season and dry-season periods and condition-only LightGBM for the rainy-season period， with held-out test PR-AUC values of 0.921， 0.905， and 0.966， respectively. Historical environmental-change indicators produced only modest improvements in the all-season and dry-season models and did not improve the rainy-season model， indicating that their predictive benefit was season-dependent. The 2024 results showed stronger spatial agreement between the all-season and dry-season susceptibility maps， whereas high-susceptibility areas during the rainy season were more localized. The dry-season High and Very High classes covered the largest area， reaching 8，581 km²， or 21.49% of the analytical grid extent. SHAP analysis identified vegetation condition， soil-moisture trend， and precipitation as the principal contributors to the all-season model; precipitation， vegetation condition， and oil-palm land-use context as the principal contributors to the dry-season model; and wind speed， vegetation condition， and peatland context as the principal contributors to the rainy-season model. Independent validation using 2024 POLRI fire records showed that 77.60%， 76.79%， and 78.57% of the matched records for the all-season， dry-season， and rainy-season periods， respectively， occurred within the High and Very High susceptibility classes. These findings demonstrate that separate seasonal models can reveal predictive and spatial differences that may be obscured when wildfire susceptibility is represented solely by an all-season model.

基於物聯網標準與混合實境裝置建構物聯網虛實整合互動系統;Constructing an IoT Virtual-Physical Interaction System based on IoT Standards and Mixed Reality Devices

title: 基於物聯網標準與混合實境裝置建構物聯網虛實整合互動系統;Constructing an IoT Virtual-Physical Interaction System based on IoT Standards and Mixed Reality Devices abstract: 隨著物聯網(Internet of Things， IoT)與測繪技術快速發展，「數位孿生(Digital Twins)」這個概念也逐漸在這個資訊化時代引起各領域關注。數位孿生能夠有效描述真實物理世界物件在虛擬資訊世界中的數位表徵，持續提供觀測對象的動態與靜態資訊，因而廣泛應用於工業4.0、智慧城市、自動化家居等物聯網相關應用。數位孿生作為物理世界與虛擬世界的連結，其資料若能以符合直覺的方式呈現於使用者所處環境，便有潛力帶來身歷其境且直觀的操作體驗；然而目前主流物聯網系統多仰賴於二維介面操作，必須仰賴使用者事先在腦中建立螢幕上的抽象標籤與三維物理環境中實體裝置的對應關係，是相當不直覺且低效率的。此外，現行混合實境(Mixed Reality， MR)與物聯網(MR-IoT)系統及相關研究多採客製化介面，並將系統配置直接寫死(hard-coding)在底層程式碼中，導致新增裝置或場域遷移時皆需開發者重新介入，導致規模化部署及維護瓶頸。針對上述問題，本研究提出一套以開放標準為核心的MR-IoT數位孿生整合架構，並設計標準化設定檔規格(Deployment Schema)作為部署資訊的結構化描述機制；整合Microsoft HoloLens 2混合實境裝置空間映射與全息影像渲染能力、Azure Spatial Anchors視覺定位機制，以及開放地理空間協會SensorThings API感測與致動標準，並基於上述基礎建構三個核心功能組件：空間映射與錨點定位、IoT標準化通訊、及使用者介面。透過此架構，單一混合實境應用得以透過國際開放標準介面存取跨廠牌異質物聯網裝置，並將控制介面以全息影像精確疊加於實體裝置旁，以消除使用者於二維螢幕與三維物理空間之間的認知轉換負擔。在設定檔規格層面，本研究提出基於JSON之設定檔格式，採用樣板與實例分離設計。該規格將裝置屬性、控制面板配置與空間繫結等資訊集中於結構化文件當中，並透過識別碼與Open Geospatial Consortium (OGC)所提出的SensorThings API標準對應，同時填補既有標準於混合實境應用部署的空缺，藉由修改設定檔取代複雜程式碼改寫維護，從設計層面大幅降低系統建置成本。面對研究期間商用混合實境生態系的劇烈變動，本研究的設計亦印證了採用開放標準與跨平台架構的必要性。最終透過智慧家居情境建置原型系統，成功整合自製Arduino感測器與商用智慧燈泡，驗證所提架構的設備整合能力並討論設定檔部署之可行性。期望本研究能藉由降低部署成本與提升跨平台可移植性，為後續MR-IoT系統之異質裝置整合及跨場域規模化部署研究，奠定可靠之參考基礎。;With the rapid advancement of the Internet of Things (IoT) and geospatial technologies， the concept of Digital Twins has drawn growing attention across diverse fields in this information age. A digital twin effectively serves as the digital representation of real-world physical objects within a virtual information space， continuously providing both dynamic and static properties of the observed objects. Consequently， it has been widely adopted in IoT-related applications such as Industry 4.0， smart cities， and home automation. As a bridge between the physical and virtual worlds， digital twins hold the potential to deliver immersive and intuitive operational experiences—provided its data is presented intuitively within the user′s real-world environment. However， most mainstream IoT systems still rely on two-dimensional interfaces for interaction， requiring users to mentally establish mappings between abstract on-screen labels and physical devices in three-dimensional space—an approach that is both unintuitive and inefficient. Furthermore， existing Mixed Reality and IoT (MR-IoT) systems and related research commonly adopt custom interfaces with system configurations hard-coded directly into the underlying codebase. This means that adding new devices or migrating to a new venue requires repeated developer intervention， creating a significant bottleneck for scalable deployment and maintenance. To address these issues， this study proposes an open-standards-based MR-IoT Digital Twin integration architecture and introduces a standardized Deployment Schema as a structured mechanism for describing deployment information. The architecture integrates the spatial mapping and hologram rendering capabilities of Microsoft HoloLens 2， the visual positioning of Azure Spatial Anchors， and the sensing and actuation standards of the OGC SensorThings API (STA) to construct three core functional components: spatial mapping and anchor positioning， standardized IoT communication， and the user interface. Through this architecture， a single MR application can access heterogeneous IoT devices from different vendors via a unified international standard interface. Control panels are precisely overlaid on their corresponding physical devices as holograms， thereby eliminating the cognitive mapping burden imposed on users when bridging two-dimensional screens and three-dimensional physical space. At the deployment specification level， this study proposes a configuration file format based on JSON and JSON Schema， adopting a template-instance separation design. The Deployment Schema consolidates device attributes， control panel configurations， and spatial anchor bindings into a single structured document. It aligns with the STA standard through identifier references， filling the gap left by existing standards in MR application deployment. By enabling system updates through configuration file edits rather than complex code rewrites， the schema substantially reduces system construction and maintenance costs at the design level. In response to the significant shifts in the commercial MR ecosystem observed during the research period， the design approach taken in this study further affirms the necessity of adopting open standards and a cross-platform architecture. A prototype system was implemented in a smart home context， integrating a custom-built Arduino sensor and a commercial smart bulb， to demonstrate the device integration capability of the proposed architecture and to evaluate the feasibility of Deployment Schema-based deployment. It is hoped that this study， through reducing deployment costs and improving cross-platform portability， can provide a reliable reference foundation for future research on heterogeneous device integration and cross-venue scalable deployment in MR-IoT systems.

利用多個再分析資料集探討北歐海域海洋熱浪和海冰融化的時空變化;利用多個再分析資料集探討北歐海域海洋熱浪和海冰融化的時空變化

title: 利用多個再分析資料集探討北歐海域海洋熱浪和海冰融化的時空變化;利用多個再分析資料集探討北歐海域海洋熱浪和海冰融化的時空變化 abstract: 本研究整合海表溫度、海洋熱浪、多源海冰厚度、表面流場、表面風場與北極震盪指數，分析巴倫支–北歐海域（60–84°N、30°W–45°E）的長期時空變化及可能關聯。方法包括氣候與面積加權平均、線性趨勢、多源產品比較、東西分區及北極震盪年度相位分類。1985–2025年年平均海表溫度以約0.319 °C/decade升高；海洋熱浪綜合訊號於2000年後增強，2016–2018年及2023–2025年較明顯。三套長期海冰厚度產品均呈減薄方向：資料A於1996–2024年跨年冰季下降約−0.0668 m/decade，未達統計顯著；資料B與C在一般線性回歸下分別下降約−0.1332及−0.0291 m/decade，均達統計顯著。較厚海冰主要集中於格陵蘭東側、弗拉姆海峽及研究區北部。海冰厚度與海洋熱浪呈季節性空間對應，但不足以證明因果關係。表面流場與風場呈冬春較強、夏季較弱及秋季後再增強的循環，西側多數月份的平均流速與風速較高。不同北極震盪年度相位下均可能出現高海溫、較強海洋熱浪訊號及較低海冰厚度，顯示北極震盪可調節特定季節與年份，但並非長期暖化與海冰減薄的唯一驅動因素。整體變化反映氣候暖化、大西洋暖水、海冰回饋及大氣–海洋動力的共同作用。;This study integrates sea surface temperature， marine heatwaves， multi-source sea ice thickness， surface currents， surface winds， and the Arctic Oscillation to examine long-term environmental changes and their potential relationships in the Barents–Nordic Seas (60–84°N， 30°W–45°E). The analyses include climatological and area-weighted averaging， linear trend analysis， sea ice product intercomparison， east–west regional comparisons， and annual Arctic Oscillation phase classification. From 1985 to 2025， the annual mean sea surface temperature increased by approximately 0.319 °C per decade. Marine heatwave signals strengthened after 2000 and were most pronounced during 2016–2018 and 2023–2025， with summer–autumn hotspots in the northern Norwegian Sea， south of Svalbard， and western Barents Sea. All three long-term sea ice thickness products indicate thinning. Product A shows a non-significant decline of approximately −0.0668 m per decade during the 1996–2024 cross-year ice seasons. Products B and C decline by approximately −0.1332 and −0.0291 m per decade， respectively， and both trends are statistically significant under the ordinary linear regression used in this study. Despite product differences， thicker ice consistently occurs east of Greenland， in Fram Strait， and at higher latitudes. Sea ice thickness and marine heatwaves exhibit seasonal spatial correspondence， but the results do not establish direct causality. Surface currents and winds are stronger from winter to early spring， weaker in summer， and strengthen again in autumn， with higher mean speeds in the western region during most months. High sea surface temperatures， stronger marine heatwave signals， and low sea ice thickness occur under different annual Arctic Oscillation phase types. The Arctic Oscillation therefore modulates seasons and years but is not the sole driver of long-term ocean warming and sea ice thinning. Overall， changes reflect climate warming， Atlantic Water influence， sea ice feedbacks， surface circulation， wind forcing， and interannual Arctic Oscillation variability.

透過整合盤古天氣與預報校正技術增強熱帶氣旋路徑預報;Enhanced Tropical Cyclone Track Forecasting Through Integration of Pangu-Weather and Forecast Correction

title: 透過整合盤古天氣與預報校正技術增強熱帶氣旋路徑預報;Enhanced Tropical Cyclone Track Forecasting Through Integration of Pangu-Weather and Forecast Correction abstract: 熱帶氣旋（TC）路徑預測對科學家與研究人員而言仍是一項挑戰；研究人員持續改進預報，以支援更準確的災害預警、疏散規劃與減災措施。熱帶氣旋對西北太平洋（WNP）地區影響顯著，但藉助人工智慧（AI）天氣預報模式（例如Pangu-Weather），可以產生數據驅動的預報，並有效地模擬複雜的大氣行為。本研究旨在透過引入Pangu-Weather目前尚未實現的預報校正和改進技術來提高熱帶氣旋路徑預測的準確性。本研究採用四步驟流程。首先，從ERA5再分析資料產生輸入數據，並將其轉換為NumPy陣列。接著，將 Pangu-Weather 作為預訓練的確定性大氣預報核心，以 6 小時間隔產生未來 7 天的大氣預報場；Pangu-Weather 本身並不直接輸出熱帶氣旋路徑。之後，利用 TempestExtremes（TE）從這些大氣預報場中擷取初始熱帶氣旋預報位置，再透過深層平均導引、平均海平面氣壓異常中心重新定位、厚度導引，以及 TE 錨定平滑與護欄方法進行後處理，以調整預報路徑的經緯度位置。最終產生確定性的 TC 路徑預報，並以 IBTrACS 作為參考資料，利用直接位置誤差（DPE）、均方根誤差（RMSE）及平均絕對誤差（MAE）評估原始路徑與後處理路徑的表現。本研究對2023年至2024年的43個TC進行了評估，結果顯示後處理總體上有所改善，但效果並不均勻。結果顯示，平均RMSE從478.74 km降至469.64 km，平均MAE從387.09 km降至380.47 km。在43個熱帶氣旋中，21個的均方根誤差（RMSE）有所改善，14個保持不變，8個有所惡化。例如，Ewiniar、Ampil、Bolaven和Soulik等熱帶氣旋的誤差降低幅度較大，而其他熱帶氣旋的改善幅度較小或略有惡化。這些結果表明，所提出的框架最好被理解為一種選擇性的改進層，可以改善許多合理的AI預測路徑，但並非通用的校正方案。;Tropical cyclone (TC) track prediction remains a challenge for scientists and researchers， who continually refine forecasts to support accurate disaster warnings， evacuation planning， and mitigation efforts. TCs significantly affect the Western North Pacific (WNP) region， but with the help of artificial intelligence (AI) weather forecasting models， such as Pangu-Weather， data-driven forecasts can be generated and can represent complex atmospheric behavior efficiently. This study seeks to improve TC track prediction by incorporating forecast correction and refinement techniques that are not yet implemented in Pangu-Weather. For this study， a four-step process is conducted. First， the inputs are generated from ERA5 reanalysis data and converted to NumPy arrays. Next， Pangu-Weather is used as a pretrained deterministic forecasting core to generate atmospheric forecast fields at 6-hour intervals over a 7-day horizon; it does not directly generate the TC track. TempestExtremes (TE) is then used to extract the initial TC forecast positions from these atmospheric fields， after which post-processing is applied using deep-layer-mean steering， mean sea level pressure anomaly recentering， thickness steering， and TE-anchored smoothing and guardrails. Finally， a deterministic TC track forecast is produced. The resulting tracks were evaluated against IBTrACS reference data using Direct Position Error (DPE)， Root Mean Squared Error (RMSE)， and Mean Absolute Error (MAE). A total of 43 TCs from 2023 to 2024 were evaluated and showed a net-positive but non-uniform response to post-processing. The results showed that the mean RMSE decreased from 478.74 km to 469.64 km， and the mean MAE decreased from 387.09 km to 380.47 km. Of the 43 TCs， 21 improved in RMSE， 14 remained neutral， and 8 degraded. Stronger reductions in errors were observed in cases such as Ewiniar， Ampil， Bolaven， and Soulik， whereas other storms showed weaker improvements or slight degradation. These findings indicate that the proposed framework is best interpreted as a selective refinement layer that can improve many plausible AI-derived tracks， but not as a universal correction scheme.