Evaluation of a state-wide drainage monitoring in Mecklenburg-Vorpommern using artificial intelligence methods

One challenge for modern agricultural management systems is to reduce their detrimental effects on water quality of water bodies. With this in mind, monitoring was carried out on behalf of the State Agency for Environment, Nature Conservation and Geology Mecklenburg-Vorpommern in the drainage outlets of 19 arable fields distributed throughout the state. In long-term measurement campaigns, runoff and substance concentration were determined at the drainage outlets. As expected with intensive arable land use, the nitrogen concentrations of most samples were far above the current quality standards and target values for surface waters according to the Surface Water Ordinance. Phosphorus concentrations were also generally very high. In parallel to the measurements, extensive data on agricultural management were collected and then correlated, together with pedological and meteorological data, with the temporal dynamics and spatial patterns of substance concentrations in drainage runoff. After selection and processing, 1037 data sets with 19 measured variables were available for the analyses.

These data were first subjected to a principal component analysis. The first seven principal components were each assigned to specific effects. The values of the principal components were interpreted as quantitative measures of the strength of the expression of these effects in the individual water samples. These were then placed in relation to extensive meteorological, hydrological, soil and management data. Classical correlation analyses revealed a bewildering variety of significant effects. Using random forest models, however, it was possible to map a large part of the observed spatial and temporal variance with just a few explanatory variables in each case.

The temporal dynamics of the nutrient concentrations in the outlet of the drains were mainly determined by hydrological conditions and weather. In contrast, direct short-term effects of individual arable farming measures on nutrient dynamics in the drainages could not be identified. Instead, clear indications of long-term effects of agriculture were found. In particular, the nitrogen and phosphorus balances of the areas played a decisive role. Soil recovery from long-term fertilisation thus does not seem to be achievable either through minor changes in agricultural management or in the short term.

Furthermore, it was shown that the many years of intensive use of the land had a massive impact not only on the nitrogen and phosphorus contents, but also on almost all other substances studied. Nitrogen and phosphorous data alone can only provide limited information on the source and development of soil eutrophication. This is still given too little attention in studies on the substance balance of agriculturally used land.