Transport of phosphorus (P) from agricultural fields to water bodies deteriorates water quality and causes eutrophication. To reduce P losses and optimize P use efficiency by crops, better knowledge is needed of P turnover in soil and the efficiency of best management practices (BMPs). In this review, we examined these issues using results from 10 Swedish long-term soil fertility trials and various studies on subsurface losses of P. The fertility trials are more than 50 years old and consist of two cropping systems with farmyard manure and mineral fertilizer. One major finding was that replacement of P removed by crops with fertilizer P was not sufficient to maintain soil P concentrations, determined with acid ammonium lactate extraction. The BMPs for reducing P leaching losses reviewed here included catch crops, constructed wetlands, structure liming of clay soils, and various manure application strategies. None of the eight catch crops tested reduced P leaching significantly, whereas total P loads were reduced by 36% by wetland installation, by 39 to 55% by structure liming (tested at two sites), and by 50% by incorporation of pig slurry into a clay soil instead of surface application. Trend analysis of P monitoring data since the 1980s for a number of small Swedish catchments in which various BMPs have been implemented showed no clear pattern, and both upward and downward trends were observed. However, other factors, such as weather conditions and soil type, have profound effects on P losses, which can mask the effects of BMPs.

Phosphorus losses from arable land need to be reduced to prevent eutrophication of surrounding waters. Owing to the high spatial variability of P losses, cost-effective countermeasures need to target parts of the catchment that are most susceptible to P losses. Field surveys identified critical source areas for overland flow and erosion amounting to only 0.4–2.6 % of total arable land in four different catchments in southern Sweden. Distributed modelling using high-resolution digital elevation data identified 72–96 % of these observed erosion and overland flow features. The modelling results were also successfully used to predict occurrence of overland flow and rill and gully erosion in a catchment in central Sweden. Such exact high-resolution modelling allows for accurate placement of planned countermeasures. However, current legislative and environmental subsidy programmes need to change their approach from income-loss compensation to rewarding high cost effectiveness of implemented countermeasures.

Suspended solids (SS) are important carriers of pollutants such as phosphorus (P) in streams, but the sampling frequency in monitoring programs is usually insufficiently frequent to capture episodic SS and total P (TP) peaks. The suitability of turbidity and conductivity as a surrogate for SS and TP was studied using 108 monitoring stations located in catchments of different sizes, land uses, and pollution levels. The use of high-frequency turbidity measurements to estimate SS and TP loads was compared with the use of two sampling methods (grab, flow-proportional sampling) in a case study. When all samples were considered, turbidity was a good predictor of SS ( r 2 = 0.76) and TP ( r 2 = 0.75). For single sites, there was a large range in how well turbidity could predict the two variables. The site-specific turbidity-SS relationship was significant at 87% of sites (mean r 2 = 0.72). The site turbidity and conductivity-TP relationship was significant at 78% of sites (mean r 2 = 0.62). A stronger turbidity-SS relationship was found in catchments with a higher percentage of agricultural land. The turbidity and conductivity-TP relationship was stronger when the TP concentration was high. In the case study, TP loads were smallest when estimated with grab sampling, which missed several discharge peaks. Loads estimated with high-frequency turbidity measurements were 19–51% smaller than with flow-proportional sampling, probably due to differences in sampling points. High-frequency turbidity measurements can be a viable alternative to conventional sampling methods in studies on concentration dynamics and load estimates.

Soil-crop models are used to simulate ecological processes from the field to the regional scale. Main inputs are soil and climate data in order to simulate model response variables such as crop yield. We investigate the effect of changing the resolution of input data on simulated crop yields at a regional scale using up to ten dynamic crop models. For these models we compared the effects of spatial input data aggregation for wheat and maize yield of two regions with contrasting climate conditions (1) Tuscany (Italy, Mediterranean climate) and (2) North Rhine Westphalia (NRW, Germany, temperate climate). Soil and climate data of 1 km resolution were aggregated to resolutions of 10, 25, 50, and 100 km by selecting the dominant soil class (and corresponding soil properties) and by arithmetic averaging, respectively.Differences in yield simulated at coarser resolutions from the yields simulated at 1 km resolution were calculated to quantify the effect of the aggregation of the input data (soil and climate data) on simulation results.The mean yield difference (bias) at the regional level was positive due to the upscaling of productive dominant soil(s) to coarser resolution. In both regions and for both crops, aggregation effects (i.e. errors in simulation of crop yields at coarser spatial resolution) due to the combined aggregation of soil and climate input data increased with decreasing resolution, whereby the aggregation error for Tuscany was larger than for North Rhine Westphalia (NRW). The average absolute percentage yield differences between grid cell yields at the coarsest resolution (100 km) compared to the finest resolution (1 km) were by about 20 to 30% for Tuscany and less than 15 and 20% for NRW for winter wheat and silage maize, respectively.In the Mediterranean area, the prediction errors of the simulated yields could reach up to 60 % when looking at individual crop model simulations. Additionally, aggregating soil data caused larger aggregation errors in both regions than aggregating climate data.Those results suggest that a higher spatial resolution of climate and especially of soil data input are necessary in Mediterranean areas than in temperate humid regions of central Europe in order to predict reliable regional yield estimations with crop models.For generalization of these outcomes, further investigations in other sub-humid or semi-arid regions will be necessary.

No abstract