重力異常（gravity anomaly）一般定義為大地水準面的實測重力值與正常地球橢球面正常重力值的差值。但大地水準面的重力值無法直接測量，所以需將地表測量的重力值歸算至大地水準面上。重力歸算主要包含二步驟：一為自由空間改正；一為去除測站與大地水準面層間質量引力的重力地形改正。傳統上重力地形改正是先作布格平板改正（將測站附近視為無限延伸的平坦地形），再移去及補上高出測站或低於水平面地形質量的引力。本研究為捨棄傳統重力地形改正方式，應用錐體斷面（地形剖面數值）直接進行大地水準面與測站層間質量引力的重力地形改正。

錐體斷面法為將測站視為原點，將地形劃分為等圓心角的扇形區塊，圓心角愈小表現的地形就愈真實，並以各扇形中心線的斷面（剖面）代表各扇形區塊的地形變化。再將各扇形斷面數據代入本研究所推導的錐體斷面法公式，得到重力地形改正。在與傳統漢默法計算方式比較，最主要差異為：(1)漢默法以測站水平面為基準面，本文方法為大地水準面基準面；(2) 漢默法計算的地形剖面為階梯式，本文方法地形剖面為錐體式，較契合真實地形。

本研究選擇台灣地區4點地形變化較大，且能代表不同類型地形特徵，分別為山頂2點、山谷1點及山腰1點進行實際地形測試，測試的方法分為A型和B型模版，A型和B型模版代表不同方位（相差扇形圓心角的一半）所計算的地形效應。由模板測試結果於地形12等分以上時，可達標準差在1.1mGal精度以內。錐體斷面法與傳統漢默法計算的結果比較，除玉山約12.6 mGal，其它3測試點誤差皆在4 mGal以內。因此，在台灣地區將地形12等分以上做重力地形改正是合理的。

The difference between actual gravity referring to the geoid and normal gravity estimated on a selected specific ellipsoid is generally defined as gravity anomaly. Because of the actual gravity on the geoid can not be obtained directly, we have to reduce the gravity measured on the physical surface of the earth to the geoid. Both of the two main procedures of gravity reduction are the free-air reduction and the terrain correction which removed the mass between the earth surface and the geoid. Traditionally, we always apply the Bouguer correction first and then remove the excess mass or fill it in the terrain correction. In this paper, we directly apply the terrain correction by means of the conical cross-section method.
The conical cross-section method assumes the surveying-site as an original point and then divides the neighbor topography into serial sector areas with equal central angle. The more we make the central angel smaller, the more we get the real expressed topography. The central lines of the sector are used to describe the variance of the topography within every sector area. By using the conical cross-section equations provided from this paper, we get the terrain correction. Comparing with the traditional Helmert's method, the principal differences are: (1) Helmert's method bases on the leveling-plate of the station, our method is on the geoid; (2) the cross-section of Helmert's method is trapezoid, our method is conical. The method gave from this paper is more suitable than Helmert's to fit the real topography.
In this paper, four specific points are selected from different sorts of topography. Two points are on the mountaintop. Another one is on the mountain valley, and the last one is on the mountainside. The plate A and plate B are used to calculate the topographic effects. From the calculated result, we found out the standard deviation could reach to 1.1 mGal by using the twelve equally divided parts of topography. Comparing results between conical cross-section and Helmert's method, the difference of the three specific points are within 4 mGal except the YuSan point of 12.6 mGal. Hence, it is reasonable to divide the topography into twelve equal parts for the terrain correction in Taiwan.

孔祥元、郭?明、劉宗泉，2002，大地測量學基礎，武漢大學出版社，武漢，265頁。

胡明城，2003，現代大地測量學的理論及其應用，測繪出版社，北京，476頁。

蔡坤哲，2001，重力地形修正之探討，國立中央大學地球物理研究所碩士論文，桃園，72頁。

陳國華，2004，整合TWVD2001水準及GPS資料改近台灣區域大地水準面模式已應用於GPS高程測量，國立成功大學測量及空間資訊學系博士論文，台南，154頁。

蕭宇伸，2002，嚴密正高改正，國立交通大學土木工程研究所碩士論文，新竹，67頁。

George B. Thomas,Jr.,1971，微積分，張海山，周本福合譯，中央圖書出版社，台北，632頁。

Hammer S., 1939, Terrain corrections for gravimeter stations,
Geophysics, 4, pp. 184-194.

Heiskanen, W. A. and H. Moritz, 1967, Physical Geodesy, W. H.
Freeman and Co., San Francisco, CA, U.S.A., 364 pages.

Nozaki K., 1981, A computer program for spherical terrain
correction, J. Geod. Soc. Japan, 27, pp.23-32.

Olivier, R. J., and Simard, R. G., 1981, Improvement of the conic prism model for terrain correction in rugged topography, Geophysics, 46, pp.1054-1056.

Yen,H.Y.,Y.H. Yeh and C.H. Chen, 1994, Study on gravity terrain correction of Taiwan, TAO, 1,2, pp.1-10.