A technique for establishing reference nutrient concentrations across watersheds affected by humans
A technique for establishing reference nutrient concentrations across watersheds affected by humans
复制标题
DOI：
10.4319/lom.2004.2.333
复制
发表时间：
2004-10-01
影响因子：
2.7
通讯作者：
Oakes, RM
中科院分区：
文献类型：
--
作者：
Dodds, WK;Oakes, RM
建立河流和溪流的参考营养条件对于评估人类对水生生态系统的影响以及保护水质和生物完整性是必要的。已经提出了几种方法：（1）从一个区域或数据集中的所有河流和溪流的统计分布中提取样本，（2）参考溪流方法，以及（3）从现有参考溪流中建模河流网络。我们提出了一个额外的统计方法来估计的影响，人为的土地利用对营养盐浓度。首先，我们量化的区域变化，通过使用协方差分析，其中总氮或总磷是因变量，地区是分类预测，和人为土地利用的百分比（e。例如，在一个实施例中，耕地、城市用地）为协变量。如果没有显著的区域效应，或者如果存在显著的区域效应，则允许对区域进行聚合，从而确定需要单独分析区域。其次，我们开发了多元线性回归模型与最佳模型技术，其中人为土地利用分类是自变量，和在流营养盐浓度的曲线图是因变量。这些回归模型的截距（即，假设线性外推至原点，在没有人类活动的情况下的预期营养物浓度）代表参考营养物浓度。这一分析表明，更大比例的农田和城市土地有很强的积极影响，在流营养盐浓度，无论是在东部堪萨斯和整个美国接壤。最适合各区域的方法可能取决于参考地点和其他数据来源的相对可用性。协方差/参考方法为参考站点数量有限的区域提供了一种潜在的方法。
Establishing reference nutrient conditions for rivers and streams is necessary to assess human impact on aquatic ecosystems and protect water quality and biotic integrity. Several methods have been proposed: ( 1) percentiles from statistical distributions of all rivers and streams in a region or dataset, ( 2) reference stream approaches, and ( 3) modeling river networks from existing reference streams. We propose an additional statistical method to estimate the influence of anthropogenic land uses on lotic nutrient concentrations. First, we quantify regional variation by using analysis of covariance, where total nitrogen or total phosphorus is the dependent variable, region is the categorical predictor, and percentage of anthropogenic land use ( e. g., cropland, urban land) is the covariate. This allows for the aggregation of regions if there is not a significant regional effect, or if there is a significant regional effect, identifies the need to analyze regions separately. Second, we develop multiple linear regression models with best-model techniques in which anthropogenic land-use classifications are the independent variables, and the logarithms of in-stream nutrient concentrations are the dependent variables. The intercept of these regression models (i.e., expected nutrient concentration in the absence of human activities assuming linear extrapolation to the origin) represents the reference nutrient concentrations. This analysis suggests that larger percentages of cropland and urban land have strong positive influences on in-stream nutrient concentrations, both in eastern Kansas and across the conterminous United States. The most appropriate method for regions may depend on the relative availability of reference sites and other data sources. The covariance/reference approach offers a potential method for regions with limited numbers of reference sites.