再分析資料之西北太平洋熱帶氣旋客觀偵測與評估

Objective Detections and Evaluations of Western North Pacific Tropical Cyclones in the Reanalysis Datasets

林涵芳(Han-Fang Lin)

指導教授 ： 蔡孝忠

淡江大學/工學院/水資源及環境工程學系碩士班/碩士(2018年)

https://doi.org/10.6846/TKU.2018.00889

摘要

熱帶氣旋客觀偵測方法已經被廣泛運用於全球預報模式與氣候模擬之評估與校驗，典型之熱帶氣旋偵測條件有：(1)海平面氣壓場具有局部低壓、(2)熱帶氣旋中心之850 hPa相對渦度、(3)熱帶氣旋中心附近之最大表面風速、(4)高層暖心、(5)高低層氣溫距平差異以及(6)高低層相對渦度差異。
本研究利用2016年之ERA-Interim(ECMWF)與MERRA-2 (NASA)兩組再分析資料，分析數值模式內之熱帶氣旋特徵，採用兩種熱帶氣旋偵測方法：(1)熱帶氣旋直接比對與(2)熱帶氣旋客觀偵測。本研究以多變量羅吉斯迴歸(Multivariate Logistic Regression)改進熱帶氣旋客觀偵測模式，配合颱風最佳路徑資料(Best Tracks)進行校驗。藉由機率式與二元式等校驗方法，評估再分析資料於西北太平洋地區之熱帶氣旋客觀偵測結果，並分析兩組再分析資料之差異。
熱帶氣旋直接比對之結果顯示，MERRA-2之局部低壓中心偵測率(約99%)較ERA-Interim (約92%)高7%，而條件(2)至條件(4)之偵測率皆相似。然而，兩組再分析資料對於代表颱風垂直結構特徵之高低層氣溫距平差異條件，皆表現得不盡理想，條件(5)、(6)之偵測率較條件(2)-(4)約低了20至40%。熱帶氣旋客觀偵測之機率式校驗結果顯示，兩組再分析資料於2016年測試組之ROC曲線下面積(area under the Receiver Operating Characteristic curve)皆高達0.9，代表颱風客觀偵測模式具有良好的判別能力。二元式校驗結果顯示，兩組再分析資料之偵測率約為85%，然而，ERA-Interim之誤報數量(152)較MERRA-2再分析資料(311)約少了50%。最後，本研究將熱帶氣旋客觀偵測模式應用至2017年ERA-Interim。校驗結果顯示，2017年ERA-Interim具有較多的局部低壓中心缺漏個案；若要求路徑生命期需至少為24小時，熱帶氣旋之偵測率約為58%，誤報率約為40%。

關鍵字

熱帶氣旋 ； 客觀偵測 ； 羅吉斯迴歸 ； 再分析資料

並列摘要

Objective tropical cyclone (TC) detection methods have been widely used for tracking TCs in the global model forecasts and climate simulations. The typical detection criteria used by the TC tracking methods are: (i) local minimum in sea-level pressure (SLP)；(ii) 850 hPa relative vorticity at the TC center, (iii) maximum surface wind speed near the TC center；(iv) an upper-level warm core；(v) temperature anomaly at the upper level is higher than at the lower levels；and (vi) warm-core feature detection inferred from the difference of the relative vorticities between the lower and upper levels.
In this study, the atmospheric reanalysis datasets from the ERA-Interim (ECMWF) and MERRA-2 (NASA) are used to investigate the representativeness of TC features. Two TC identification methods are applied: (1) direct TC matching, and (2) objective TC detection. The Logistic Regression is used to improve the objective TC detection method. The results from two reanalysis datasets are compared with the best tracks by using the probabilistic and binary verification methods.
The result of the direct TC matching method shows that the probability of detection (POD) of the local minimum SLP centers is about 7% higher in the MERRA-2 (~99%) than that in the ERA-Interim (~92%), which may be due to the MERRA-2’s horizontal resolution ( ) is relatively higher. The PODs are also similar if the conditions 2-4 are analyzed. Both reanalysis datasets are somewhat less successfully in representing the TC’s vertical structure in terms of the differences of temperature anomalies between the lower and upper levels. The PODs are about 20-40% lower than those of the conditions 2-4. The result of the objective TC detection method shows that the area under the Receiver Operating Characteristic curve (AUC) for both datasets in 2016 are about 0.90. The binary verification result shows that the PODs are about 85% in both reanalysis datasets. However, the False Alarm number for the ERA-Interim (152) is about 50% lower than for the MERRA-2 (311). Finally, the objective TC detection method is applied to the ERA-Interim in 2017. The verification results show that there are more missing minimum pressure centers in the ERA-Interim in 2017. The POD is 58% and FAR is 40% if it is required that a TC track must last at least 24 hours.

並列關鍵字

Tropical cyclone ； Objective detections ； Logistic regression ； Reanalysis dataset

參考文獻

1. Agresti, A. (2002). Categorical Data Analysis (2nd ed.). Wiley.

2. Bengtsson, L., M. Botzet, and M. Esch (1995). Hurricane-type vortices in a general circulation model. Tellus, 47A, 175–196.

3. Bengtsson, L., K. I. Hodges, and M. Esch (2007a). Tropical cyclones in a T159 resolution global climate model: Comparison with observations and re-analyses. Tellus, 59A, 396–416.

4. Bengtsson, L., K. I. Hodges, M. Esch, N. Keenlyside, L. Kornblueh, J.-J. Luo, and T. Yamagata (2007b). How many tropical cyclones change in a warmer climate? Tellus, 59A, 539-561.

5. Camargo, S. J., and Zebiak, S. E. (2002). Improving the detection and tracking of tropical cyclones in atmospheric general circulation models. Weather and forecasting, 17(6), 1552-1162.