Abstract Coastal Louisiana has been facing a serious land loss problem over the past several decades, and extensive research has been undertaken to address the problem. However, the importance of landscape fragmentation on land loss has seldom been examined. This paper evaluates the effects of landscape fragmentation on land loss in the Lower Mississippi River Basin region. The research hypothesis is that the higher the degree of fragmentation in a locality, the greater the amount of land loss in the next time period. We used Landsat-TM data with a pixel size of 30 m × 30 m in 1996 and 2010 and transformed the images into either land or water pixels. We then calculated the fractal dimension and Moran's I spatial autocorrelation statistics and used them to represent the degree of landscape fragmentation. Four sample box sizes, including sizes of 101 × 101, 71 × 71, 51 × 51, and 31 × 31 pixels, were used to detect if there is a relationship between fragmentation and land loss at different neighborhood (context) scales. For each box size, 100 samples were randomly selected. To isolate the fragmentation effect so that it can be better evaluated, we used only sample boxes with a 50% land-water ratio. Regression results between fragmentation and land loss show that the R2 values for box sizes of 71 × 71, 51 × 51 and 31 × 31 were statistically significant (0.20, 0.45, 0.35; p

中文翻译：

---

景观破碎化对土地流失的影响

摘要 在过去的几十年里，路易斯安那州沿海地区一直面临着严重的土地流失问题，并且已经进行了广泛的研究来解决这个问题。然而，很少有人研究景观破碎化对土地流失的重要性。本文评估了景观破碎化对下密西西比河流域地区土地流失的影响。研究假设是一个地方的破碎程度越高，下一个时间段的土地流失量就越大。我们在 1996 年和 2010 年使用了像素大小为 30 m × 30 m 的 Landsat-TM 数据，并将图像转换为陆地或水域像素。然后我们计算了分形维数和 Moran's I 空间自相关统计量，并用它们来表示景观破碎程度。四种样品盒尺寸，包括大小为 101 × 101、71 × 71、51 × 51 和 31 × 31 的像素，用于检测在不同邻域（上下文）尺度下破碎与土地损失之间是否存在关系。对于每个盒子大小，随机选择 100 个样本。为了隔离碎片效应以便更好地对其进行评估，我们仅使用了具有 50% 土地-水比的样本框。破碎化与土地流失的回归结果表明，71×71、51×51和31×31的盒子大小的R2值具有统计学意义（0.20、0.45、0.35；p 我们只使用了 50% 地水比的样本箱。破碎化与土地流失的回归结果表明，71×71、51×51和31×31的盒子大小的R2值具有统计学意义（0.20、0.45、0.35；p 我们只使用了 50% 地水比的样本箱。破碎化和土地损失之间的回归结果表明，71×71、51×51和31×31的盒子大小的R2值具有统计学意义（0.20、0.45、0.35；p