gCentroid计算结果偏移：为何未落在SD边界预期中心位置？

Let me break down exactly why you're seeing this offset—it all boils down to how gCentroid calculates the center, versus what you’re intuitively expecting as the "true" center of your SD boundary.

• The core calculation logic of gCentroid
  When you pass a SpatialPointsDataFrame to rgeos::gCentroid, it computes the weighted mean centroid of all points in the dataset. By default, every point has equal weight, so the result is simply the average of all the x and y coordinates. If your boundary points aren’t evenly distributed (e.g., you sampled more points along one segment of the SD boundary), the centroid will naturally shift toward the area with the densest point cluster. This isn’t a bug—it’s the intended behavior for point-based centroid calculations.

• Your "expected center" is likely a different metric
  I’m guessing what you visualize as the "correct" center is either:

  1. The geometric midpoint of the boundary’s bounding box (the middle point between the farthest north/south/east/west points), or
  2. The area-weighted centroid of the polygon shape formed by your boundary points.

  gCentroid doesn’t compute either of these for a point dataset. For a polygon, it calculates the area-based centroid (which aligns with your intuition of a shape’s center), but for points, it’s purely a mean of the point coordinates.

• Quick way to verify this
  Plot your SpatialPointsDataFrame and overlay the calculated centroid. You’ll almost certainly see the centroid sitting right in the middle of the densest cluster of boundary points. For example, if your SD boundary has twice as many points on its eastern edge, the centroid will pull eastward.

• How to get the "expected" center instead
  If you want a center that matches the geometric middle of your SD boundary, try one of these approaches:

  1. Convert points to a polygon first: Use rgeos::gPolygonize() to turn your boundary points into a polygon object, then run gCentroid on that polygon. This will give you the area-based centroid, which isn’t affected by point density.
  2. Calculate the bounding box midpoint: Extract the bounding box of your points with bbox(c.), then compute the midpoint manually:

     bb <- bbox(c.)
     expected_center <- data.frame(
       x = (bb[1, 1] + bb[1, 2]) / 2,
       y = (bb[2, 1] + bb[2, 2]) / 2
     )
     

内容的提问来源于stack exchange，提问作者Rafael