WP 19: Errors and variation associated with field protocols for the collection and application of macrophyte and hydro-morphological data

Objectives

To describe different types of possible variability and/or errors occurred in standard MTR and RHS assessment to support quality control of collected data set and to minimise the impact of errors on final result.

Methodology/work description

There are several potential sources of variability and/or error in the both Mean Trophic Rank (MTR) and River Habitat Survey (RHS) methodologies that need to be quantified in order to establish the variability associated with the collection and indexation of data.
They can be categorized as:

• the inherent variability of the method
• inter-surveyor variation
• natural background variation

An experimental field-sampling programme will be to quantify the total variation and partition its sources. For this purpose replicate sampling will be undertaken at 24 of the sites sampled in Poland in Workpackage 7 (core stream sampling).

The Workpackage 19 study programme:

• Standardised macrophyte surveys will be undertaken in summer, using MTR (Mean Trophic Rank). Identification of macrophytes will be to species level.
• At the same time, River corridor/habitat surveys will be undertaken using RHS (River Habitat Survey)or the equivalent CEN standard.. River Habitat Survey is field based recording systems, supplemented by additional time variant information acquired from cartographic sources.
• Principal hydrological and chemical data sets will be collected to explain variability sources.
• At each site three separate, fully trained surveyors will each undertaken MTR and RHS (or equivalent) surveys at each site.
• The three surveyors will record their survey data consecutively and independently of each other. They will not confer and the recording sheets will not be viewed again until returned to the Institute. No subsequent changes to the recorded data.
• Individual data and the metrics derived from them will be analysed for variation in the results.
• The sources of variation will be partioned, as indicated below, and their significance for the implementation of the WFD evaluated

The inherent variability of the method:
This variability is likely to be human error in term of wrong estimation, mis-identification or mis-application of the method. Both methodologies are assumed to be repeatable, provided the surveyor is very careful and fully trained in field data collecting procedures.

The inter-surveyor variation:
The authors of both methods put a lot of effort to increase of their reproducibility however the results of field research can be biased by inter-surveyor variation. The three different surveyors that will make surveys in the same places will provide set of data for inter-surveyor variation assessment. Statistical comparison of field research results will be undertaken using a non-parametric randomized block ANOVA (Friedman's test) where the factor 'surveyor' will be tested for significance.

Natural background variation:
This variation can be considered on either a temporal or spatial scale. The background variation can occur during the survey season as result of temporal perturbations in the environment e.g. differences in discharge (high or low water table level), vegetation development etc. Examples on a spatial variation may include variation between sites as a result of differences in either the physical and/or chemical characteristics of sites. Thus the time between surveys should be as short as possible and observation should be carried out exactly in the same sites. For the background variation in temporal scale assessment the selected sites will be surveyed by one surveyor in three different times during survey season. The statistical comparison of distributions of selected characteristics will be undertaken using Smirnov-Kolmogorov and Wilcoxon tests.

The computer program module supporting quality control (QC) of collected set of data will be written using high level programming language. The QC procedures in this program will base on statistical methods used for error/variation of data assessment.

Overview