We use drone images of a post-hurricane mangrove in south-west Mexico to apply landscape segmentation and evaluate direct disturbance effects on forest cover and indirect effects on two closely related bird species. Hurricane Carlotta made landfall in June 2012. We obtained images at ten sites comprising two mangrove types in August 2015 and made standardised counts of Rufous-naped Wren and Banded Wren at these sites in 2011 and 2015. Superpixels were extracted from drone images using an energy-driven sampling algorithm. The attributes of a sub-sample of representative superpixels were encoded as classifier inputs for feature extraction, and supervised classification was then attained with a random decision forest. The scale of classified superpixels was congruent with micropatches in wren habitats. We used percentages derived from superpixel classes of live trees and dead wood to calculate post-hurricane fractional vegetation cover. We applied generalised linear mixed models to relate live tree cover to pre- and post-hurricane frequency indices for both wren species. White mangrove sites produced markedly higher fractional cover values than red mangrove sites. Whereas the Rufous-naped Wren was not associated with live tree cover, Banded Wren occupancy was higher in areas with greater cover. Segmentation methods combined with bird monitoring constitute a tool to analyse how faunal populations respond to or are impacted by hurricane-induced changes to wetland vegetation.

中文翻译：

---

集成遥感和图像处理以测试飓风后红树林生态系统中的干扰效应

我们使用墨西哥西南部飓风后红树林的无人机图像来应用景观分割并评估对森林覆盖的直接干扰影响和对两种密切相关的鸟类的间接影响。飓风卡洛塔于 2012 年 6 月登陆。我们于 2015 年 8 月在包括两种红树林类型的 10 个地点获得了图像，并在 2011 年和 2015 年对这些地点的棕枕鹪鹩和带状鹪鹩进行了标准化计数。使用能量从无人机图像中提取了超像素驱动采样算法。代表性超像素的子样本的属性被编码为用于特征提取的分类器输入，然后通过随机决策森林实现监督分类。分类超像素的尺度与鹪鹩栖息地中的微斑一致。我们使用来自活树和死木的超像素类的百分比来计算飓风后植被覆盖率。我们应用广义线性混合模型将活树覆盖与两种鹪鹩物种的飓风前和飓风后频率指数联系起来。白色红树林站点产生的分数覆盖值明显高于红色红树林站点。虽然棕枕鹪鹩与活树覆盖率无关，但在覆盖率较大的地区，带状鹪鹩的入住率更高。分割方法与鸟类监测相结合，构成了一种分析动物种群如何应对飓风引起的湿地植被变化或受其影响的工具。我们应用广义线性混合模型将活树覆盖与两种鹪鹩物种的飓风前和飓风后频率指数联系起来。白色红树林站点产生的分数覆盖值明显高于红色红树林站点。虽然棕枕鹪鹩与活树覆盖率无关，但在覆盖率较大的地区，带状鹪鹩的入住率更高。分割方法与鸟类监测相结合，构成了一种分析动物种群如何应对飓风引起的湿地植被变化或受其影响的工具。我们应用广义线性混合模型将活树覆盖与两种鹪鹩物种的飓风前和飓风后频率指数联系起来。白色红树林站点产生的分数覆盖值明显高于红色红树林站点。虽然棕枕鹪鹩与活树覆盖率无关，但在覆盖率较大的地区，带状鹪鹩的入住率更高。分割方法与鸟类监测相结合，构成了一种分析动物种群如何应对飓风引起的湿地植被变化或受其影响的工具。