Abstract Indicators of landscape condition should be sensitive and specific to environmental change and provide early warning detection of ncipient changes. We assessed a suite of five spectral metrics derived from Landsat 5 TM imagery spanning a decade (1996–2007) to quantify ecosystem condition in the ridge-slough mosaic of the Everglades (South Florida, USA). These included the normalized difference vegetation index (NDVI), the same index using green instead of red band as the visible reference (NDVIg), the normalized difference water index (NDWI), the simple ratio of NIR and red bands (SR) and the moisture stress index (MSI). Mean and variance from pure ridge or slough pixels (i.e., those >30 m from a mapped patch edge) were quantified for twentyfour 2 × 5 km blocks across a gradient of hydrologic and ecological condition. Metrics were compared with field measures of landscape condition from block-scale soil elevation surveys, which capture dramatic spatial gradients between conserved and degraded locations. Elevationbased measures of landscape condition, validated as diagnostic in previous work, included soil elevation bi-modality (BISE), a binary measure of ecosystem condition, and the soil elevation standard deviation (SDSE), a continuous measure of condition. Spectral metric performance was assessed based on the strength (sensitivity) and shape (leading vs.lagging) of the relationship with elevation-based measures. We observed significant logistic regression slopes with BISE for only 3 metrics (mean ridge NDVI and NDVIg, mean ridge SR). In prediction of variance SDSE, the standard deviation of our VIs was more informative than the mean, and measures from sloughs more informative than ridges. The strongest predictions are variance in slough NDVIg and SR, suggesting that spatial heterogeneity in slough biomass is most informative for predicting the status of soil elevation variance. None of the vegetation metrics were leading indicators of change; a multivariate model using three VIs selected based on consistent performance across years (variance in Slough SR, slough NDVIg, and ridge NDWI) substantially improve predictions of SDSE and yielded more plausible prediction maps. These findings suggest that soil elevation changes from altered peat accretion dynamics precede visible changes in vegetation reflectance. While this constrains predictions of incipient ecosystem change, the reasonable performance of spectral metrics indicates that efficient monitoring of ridge-slough landscape health is possible as part of the ongoing Everglades restoration effort.

中文翻译：

---

大沼泽地脊-泥景观生态系统退化的远程检测

摘要 景观条件指标应对环境变化具有敏感性和特异性，并提供早期变化的预警检测。我们评估了一套源自 Landsat 5 TM 图像的 10 年（1996 年至 2007 年）的一套光谱指标，以量化大沼泽地（美国南佛罗里达州）的山脊-泥沼镶嵌中的生态系统状况。这些包括归一化差异植被指数 (NDVI)、使用绿色代替红色带作为可见参考的相同指数 (NDVIg)、归一化差异水指数 (NDWI)、近红外和红色带的简单比 (SR) 以及水分胁迫指数 (MSI)。对跨越水文和生态条件梯度的 24 个 2 × 5 公里块的纯山脊或泥浆像素（即那些距离映射斑块边缘 > 30 m 的像素）的均值和方差进行量化。将指标与块尺度土壤高程调查中景观条件的实地测量进行了比较，这些测量捕获了保护和退化位置之间的显着空间梯度。基于海拔的景观条件测量，在以前的工作中被验证为诊断，包括土壤海拔双峰 (BISE)，生态系统条件的二进制测量，以及土壤海拔标准偏差 (SDSE)，条件的连续测量。根据与基于海拔的测量的关系的强度（灵敏度）和形状（领先与滞后）评估光谱指标性能。我们仅观察到 BISE 对 3 个指标（平均脊 NDVI 和 NDVIg，平均脊 SR）的显着逻辑回归斜率。在方差 SDSE 的预测中，我们 VI 的标准偏差比平均值提供更多信息，来自泥沼的措施比山脊提供的信息更多。最强的预测是腐肉 NDVIg 和 SR 的方差，这表明腐肉生物量的空间异质性对于预测土壤海拔变化的状态最有帮助。没有一个植被指标是变化的领先指标；一个使用三个 VI 的多元模型，该模型基于多年来的一致性能（Slough SR、Slough NDVIg 和 ridge NDWI 的方差）显着提高了 SDSE 的预测并产生了更合理的预测图。这些发现表明，泥炭增生动态改变引起的土壤海拔变化先于植被反射率的可见变化。虽然这限制了对初期生态系统变化的预测，