|Main authors:||Oene Oenema, Meindert Commelin, Piet Groenendijk, John Williams, Susanne Klages, Isobel Wright, Morten Graversgaard, Irina Calciu, António Ferreira, Tommy Dalgaard, Nicolas Surdyk, Marina Pintar, Christophoros Christophoridis, Peter Schipper, Donnacha Doody|
|FAIRWAYiS Editor:||Jane Brandt|
|Source document:||»Oenema, O. et al. 2018. Review of measures to decrease nitrate pollution of drinking water sources. FAIRWAY Project Deliverable 4.1, 125 pp|
A systematic literature search and inventory has been conducted to collect experimental data and information for a quantitative analysis of the effectiveness of measures aimed at decreasing nitrate pollution of ground water and surface waters. All data were stored in a database and analysed statistically using R-programming (https://www.r-project.org/).
Here we present the results of the literature search and data analyses. A total of 84 literature sources have been collected about nitrate pollution. A total of 44 papers contained data about pollution of surface waters, and 80 papers about pollution of groundwater, indicating that 40 papers contained data for both surface waters and groundwater pollution. Table 13 provides an overview of the database. Annex 3 of »Review of measures to decrease nitrate pollution of drinking water resources presents the list of references of the studies that have been examined.
The literature search and data screening took more time than initially expected and the number of studies included in the database is too limited to conduct a full meta-analysis. The 84 studies included in the database have been conducted mostly in the EU-28, but some studies originate from other continents. Most of the studies from EU originated from western Europe. Most studies dealt with nitrogen input control, fertilization type and application method, crop type and cover crops.
A full meta-analysis of an updated database is presented in »Most promising measures and practices.
Effectiveness of measures was derived from the response ratio (RR), which is the nitrate leaching loss from a treatment measure divided by the nitrate leaching loss of the reference treatment (control treatment). The latter is usually current practice or conventional practice. The ratio may vary from 0 to more than 1; a value smaller than 1 indicates that the treatment measure decreases the nitrate leaching loss relative to the reference treatment. A ratio of 1 means no effect. Instead of a relative comparison of nitrate leaching loss, the response ratio was sometimes derived from a comparison of nitrate concentration in waterbodies or from the amounts of soil mineral N in the soil between treatments, depending on the availability of the data in the reviewed publications (»Review methodology).
Table 13 provides an overview of the response ratio RR of some key treatment measures. The overall mean RR ranged from 0.5 to 1.04, indicating a wide range of effectiveness of the measures. Most measures had an RR in the range of 0.5-0.7. Treatments related to fertilizer type and application method, and time of application had a RR close to 0. The same holds for irrigation. This overview suggests that N input control measures, adjusting crop type/crop rotation, growing cover crops, adjusting mulching/tillage and use of nitrification inhibitors are the most effective measures.
Table 13: Summary of the database on measures aimed at decreasing nitrate pollution of groundwater and surface waters
|Measures||Number of studies||Response ratio RR (±sd)||Number of comparisons||Outliers|
|Nitrogen input control||14||0.67 (0.29)||33||1|
|Fertilization type and method||15||1.04 (0.36)||25||1|
|Timing of application||3||0.99 (0.43)||16||0|
|Nitrification inhibitors||2||0.50 (0.16)||10||0|
|Crop types and crop rotations||20||0.56 (0.36)||27||7|
|Cover crops||12||0.61 (0.36)||32||0|
|Mulching/tillage methods||9||0.66 (0.22)||16||1|
(Status 1 October 2018). In total, there were 228 experimental comparisons, but 55 of these were as yet excluded due to treatments and data that could not be interpreted.
*e.g. experiments about energy crops and drainage, among others
Treatment measures greatly differ in their effectiveness. There is also a large variability in effectiveness within a set of treatment measures. A few additional comments have to be made here. Firstly, the number of studies/comparisons differed greatly between treatment measures; some of the treatment measures (e.g. N input control measures, adjusting crop type/crop rotation, growing cover crops) have a much greater experimental basis than others (e.g., use of nitrification inhibitors, time of application). Secondly, the standard deviation of the mean response ratio tended to be very large, indicating large variability in the effectiveness. Third, the mean response ratios have as yet not been corrected for the number of measurements and variance within studies. Fourth, the effectiveness of the treatment measures has not been analysed yet for different environmental and socio-economic conditions.
Nitrogen input control measures appear effective; the mean response ratio was 0.67 (Table 13). It suggests that reducing N input decreases nitrate leaching. The relationship between N input and nitrate leaching loss is expected to be curve-linear, i.e., nitrate leaching loss is relatively low at low N input, but increase progressively when N input increases to a level where the crop is less and less able to take up the applied N (diminishing returns).
Our results clearly indicate that the response ratio RR increases the stronger the N input decreases (Figure 21). The results indicate that there is large variation in RR, which has not been examined in depth yet. A main source of variation between studies is the difference in environmental conditions, e.g., soil type, crop type, rainfall surplus). Another source of variation will be the reference treatment; if the reference treatment is a situation with excess N input, the response of a reduction in N input will large on average. However, if the reference treatment is a situation where N input is at or below the optimal level of N, seen from the perspective of crop growth, the response of a reduction in N input may be relatively small, as discussed before.
Changing the type of fertilizer used, from mineral to organic, or from ammonium-based to nitrate-based, or from nitrate-based to urea-based fertilizers appears to have no robust effect on the nitrate leaching loss (Figure 22). Also the method of application (e.g., broadcasting versus band application versus injection) appears to have no robust effect (included in these data). The mean response ratio was 0.99, with the 95% confidence interval (±2s) ranging from 0.44 to 1.53.
The time of N application appeared to have no robust effect on nitrate leaching loss either (Figure 23). The mean response ratio was 1.04 and the 95% confidence interval (±2s) was from 0.44 to 1.53. There is a huge variation in response ratio, which is likely related to the differences in the set-up of the various studies and in the treatments examined. Response ratios 1 likely relate to different variants of autumn application (early autumn versus late autumn). Evidently, the response are large (two ways), indicating that timing is important, but the different treatments have to be sorted in a more logical manner before more definite conclusions can be made.
Nitrification inhibitors added fertilizers and/or animal manure delay the microbial oxidation of ammonium (NH4+) into nitrate (NO3-), and thereby may decrease the risk of nitrate leaching, depending also on the presence of crop that can take up the ammonium from the soil. Nitrification inhibitors may also decrease the emission of the intermediate nitrous oxide (N2O), which is a powerful greenhouse gas. So far, only 2 studies have been included in the database, with results from 10 experiments. The mean response ratio was 0.5, and the 95% confidence interval (±2s) was from 0.44 0.14 to 0.87. These results suggest that using nitrification inhibitors gives a robust decrease in nitrate leaching loss (Figure 24). However, the number of studies and comparisons is too low to derive such conclusion now.
Figure 25 summarizes the response ratios of changing crop types and/or crop rotations on nitrate leaching loss. The mean response ratio was 0.56 and the 95% confidence interval (±2s) was from 0.15 to 0.98. These results suggest that changing crop types and/or crop rotations give a robust decrease in nitrate leaching loss (Figure 25). Further analyses are needed to unravel the effects of co-variables on the mean response ratio. The response ratio of introducing a change in crop type and/or crop rotations may also depend on climate and soil types. Also, the measured response ratio of a treatment may be related to the duration of the experiment. These possible effects need to be examined further.
Cover crops grown after the harvest of the main crop may mop up residual mineral nitrogen in the soil, but also increase evapotranspiration, and add organic matter to the soil when the cover crop is ploughed down in the top soil. Figure 26 shows that the results collected in the database so far. The mean response ratio was 0.61 and the 95% confidence interval (±2s) ranged from 0.27 to 0.94, suggesting a robust decrease in nitrate leaching loss. Some treatments showed an increase in nitrate leaching loss (positive RR), which may be related to N fertilization of the cover crop, the growth of leguminous cover crops, and/or to the effects of soil cultivation associated with the growth of the cover crop. Evidently, this needs to be examined further.
Minimum tillage and mulching measures influence the infiltration capacity of the soil and the potential for overland flow and runoff, and thereby the nitrate leaching loss (Figure 27). The overall mean response ratio was 0.66 and the 95% confidence interval (±2s) ranged from 0.08 up to 1.23. The mean response ratio for mulching was 0.57 and that for minimum tillage was 0.74, while the 95% confidence intervals (±2s) ranged from 0 to 1.22 for mulching and from 0.17 to 1.31 for tillage.
Results of changes in irrigation on nitrate leaching loss were highly variable, with a 95% confidence interval (±2s) ranging from 0.41 to 2.34 (Figure 28).
Note: For full references to papers quoted in this article see