Eduard Szöcs

Stream monitoring data provides insights into the biological, chemical and physical status of running waters. Additionally, it can be used to identify drivers of chemical or ecological water quality, to inform related management actions, and to forecast future conditions under land use and global change scenarios. Measurements from sites along the same stream may not be statistically independent, and the R package SSN provides a way to describe spatial autocorrelation when modelling relationships between measured variables and potential drivers. However, SSN requires the user to provide the stream network and sampling locations in a certain format. Likewise, other applications require catchment delineation and intersection of different spatial data. We developed the R package openSTARS that provides the functionality to derive stream networks from a digital elevation model, delineate stream catchments and intersect them with land use or other GIS data as potential predictors. Additionally, locations for model predictions can be generated automatically along the stream network. We present an example workflow of all data preparation steps. In a case study using data from water monitoring sites in Southern Germany, the resulting stream network and derived site characteristics matched those constructed using STARS, an ArcGIS custom toolbox. An advantage of openSTARS is that it relies on free and open-source GRASS GIS and R functions, unlike the original STARS toolbox which depends on proprietary ArcGIS. openSTARS also comes without a graphical user interface, to enhance reproducibility and reusability of the workflow, thereby harmonizing and simplifying the data pre-processing prior to statistical modelling. Overall, openSTARS facilitates the use of spatial regression and other applications on stream networks and contributes to reproducible science with applications in hydrology, environmental sciences and ecology.

Webchem: An R Package to Retrieve Chemical Information from the Web

Small streams are important refuges for biodiversity. In agricultural areas, they may be at risk from pesticide pollution. However, most related studies have been limited to a few streams on the regional level, hampering extrapolation to larger scales. We quantified risks as exceedances of regulatory acceptable concentrations (RACs) and used German monitoring data to quantify the drivers thereof and to assess current risks in small streams on a large scale. The data set was comprised of 1 766 104 measurements of 478 pesticides (including metabolites) related to 24 743 samples from 2301 sampling sites. We investigated the influence of agricultural land use, catchment size, as well as precipitation and seasonal dynamics on pesticide risk taking also concentrations below the limit of quantification into account. The exceedances of risk thresholds dropped 3.7-fold at sites with no agriculture. Precipitation increased detection probability by 43%, and concentrations were the highest from April to June. Overall, this indicates that agricultural land use is a major contributor of pesticides in streams. RACs were exceeded in 26% of streams, with the highest exceedances found for neonicotinoid insecticides. We conclude that pesticides from agricultural land use are a major threat to small streams and their biodiversity. To reflect peak concentrations, current pesticide monitoring needs refinement.

Ecotoxicologists often encounter count and proportion data that are rarely normally distributed. To meet the assumptions of the linear model, such data are usually transformed or non-parametric methods are used if the transformed data still violate the assumptions. Generalized linear models (GLMs) allow to directly model such data, without the need for transformation. Here, we compare the performance of two parametric methods, i.e., (1) the linear model (assuming normality of transformed data), (2) GLMs (assuming a Poisson, negative binomial, or binomially distributed response), and (3) non-parametric methods. We simulated typical data mimicking low replicated ecotoxicological experiments of two common data types (counts and proportions from counts). We compared the performance of the different methods in terms of statistical power and Type I error for detecting a general treatment effect and determining the lowest observed effect concentration (LOEC). In addition, we outlined differences on a real-world mesocosm data set. For count data, we found that the quasi-Poisson model yielded the highest power. The negative binomial GLM resulted in increased Type I errors, which could be fixed using the parametric bootstrap. For proportions, binomial GLMs performed better than the linear model, except to determine LOEC at extremely low sample sizes. The compared non-parametric methods had generally lower power. We recommend that counts in one-factorial experiments should be analyzed using quasi-Poisson models and proportions from counts by binomial GLMs. These methods should become standard in ecotoxicology.