Note: Here we evaluate the effect of different scenarios of agricultural management practices on soil environmental footprint and ecosystem services. We build on the conclusions of »Management practices and soil quality concerning the long term effect of agricultural management practices on soil quality indicators, and the results and conclusions presented are further explored in »Effect of management on soil quality .

1. Bottom-up analysis: based on iSQAPER experimental data

Approach

The upscaling model is based on functional relations that establish the effect of categories of management practices on soil quality indicators for each type of farming system. In this section we present the functional relations adopted for the iSQAPER upscaling model. They are based on analysis of information provided by the case studies and from the LTE sites. The starting point are the conclusions of »Management practices and soil quality on the long term effect of agricultural management practices on soil quality indicators.

The dynamic upscaling model is based on functional relations that establish the effect of categories of management practices on soil quality indicators for each type of farming system. Here we present the methodology adopted to specify the functional relations required for the iSQAPER upscaling model. The specification of the functional relations is addressed in »Effect of management on soil quality.

The proposed methodology is based on the combination of two complementary approaches: a top-down approach, where functional relations are derived from global data, and a bottom-up approach, where functional relations are derived from expert assessment based on experiences compiled on iSQAPER long term experiment (LTE) sites and on case study (CS) sites.

The objective of the work is to specify a set of functional relations linking agricultural management practices to soil quality indicators. The relations are formulated in a qualitative way and describe the long-term tendency that can be expected to be observed in a soil quality indicator after the application of a certain category of management practice during a long period of time. For instance, functional relation F_ij is defined as: ΔSQI_i=f(APM_j), where ΔSQI_i is the expected change in soil quality indicator i and AMP_j is agricultural management practice j.

AMPs maybe characterized by the intensity of their application. Initially, APMs will be described by Boolean variables, which means that functional relations will only distinguish between the application or no application of the corresponding AMP. If there is enough information, AMP intensity may be further characterized in a qualitative domain.

The change in SQIs will be described in a qualitative domain of five values, identified as [-- , - , = , + , ++]. The interpretation of these qualitative categories is the following:

• Positive (++): This category means that the management practice will certainly improve the soil quality indicator, with effects larger than 10%
• Beneficial (+): This category means that the management practice has potential to improve the soil quality indicator, but the effects may depend on additional factors. The improvement will be between 5% and 10%
• Neutral (=): This category represents a neutral impact of the management practice on the soil quality indicator under analysis. It corresponds to a positive or negative effect of less than 5%.
• Unfavourable (-):This category means that the management practice may degrade the soil quality indicator, but the effects may depend on additional factors. The degradation will be between 5% and 10%
• Negative (--): This category means that the management practice will certainly degrade the soil quality indicator, with effects larger than 10%

Functional relations

Bottom up analyses are based on information provided by the case studies and from the LTE sites. The starting point are the conclusions in »Management practices and soil quality on the long term effect of agricultural management practices on soil quality indicators (Tables 14 to 16).

Table 14. Effect of agricultural management practices on crop yield

Table 15. Effect of agricultural management practices on soil organic carbon

Table 16. Effect of agricultural management practices on water holding capacity

2. Top-Down analysis: based on probabilistic estimates from spatial data

In the top down analysis information on functional relations is derived from global data available in the data catalogue. We have explored two types of data-based inference: linear regression and conditional probability. In this section we present preliminary results of these two approaches, which will be further developed, validated and combined with bottom up approaches in »Effect of management on soil quality.

The preliminary analyses have been performed on the global datasets of soil quality indicators and other variables. We have explored the effect of potential causal variables on soil quality indicators. We selected the variables with better quantitative, spatially explicit information in the data catalogue.

Regression analysis

In this section we present regression analyses on available data for available soil quality indices. We selected crop yield as SQI and irrigation as AMP.

Yield Figure 45 shows the results of the linear regression analyses performed on crop yield as a function of irrigated area (in the grid cell) for the seven farming systems in Europe. We present the scatter plot of the global data (grid cells with values of both variables >0), the linear fit (represented by the red line), the linear regression equation and the correlation coefficient. Figure 46 shows the results of the linear regression analyses performed on crop yield as a function of irrigated area (in the grid cell) for the seven farming systems in China. Results shown in Figures 45 and 46 are not very encouraging. The scatter plot does not show a clear relation between both variables for any farming system and correlation coefficients are very low, suggesting that the degree of actual dependence between both variables is very low.

Figure 45. Europe
Figure 46. China

Cereal (1st row left), Rice (1st row right), Maize (2nd row, left), Soybean (2nd row, right), Vegetables 3rd row, left), Pasture (3rd row, right) and Permanent crops (4th row)

Soil organic carbon Figure 47 shows the results of the linear regression analyses performed on Soil Organic Carbon(SOC) as a function of irrigated area (in the grid cell) in Europe and China. They correspond to the total farming area, because no data were available for individual farming systems. As in the previous figures, we present the scatter plot of the global data (grid cells with values of both variables >0), the linear fit (represented by the red line), the linear regression equation and the correlation coefficient. In this case, although the correlation coefficient is still very low, there seems to be a relation between SOC and irrigated area. It appears that as irrigated area is greater, SOC is drastically reduced. This relation is not linear, and this explains that the correlation coefficient is so low. A possible non-linear functional relation will be explored in »Effect of management on soil quality.

Figure 47

Water holding capacity Figure 48 shows the results of the linear regression analyses performed on Water Holding Capacity (WHC) as a function of irrigated area (in the grid cell) in Europe and China. No data were available as a function of farming systems. We present the scatter plot of the global data (grid cells with values of both variables >0), the linear fit (represented by the red line), the linear regression equation and the correlation coefficient. In this case, there seems to be a relation between WHC and irrigated area. WHC appears to grow as irrigated area grows. This relation is not linear, as shown by the low values of the correlation coefficient, and will be explored in »Effect of management on soil quality.

Figure 48

Conditional probability analysis

Conditional probability analysis is an alternative approach to linear regression analysis. To explore the relation between two variables, we analyse the joint probability distribution function (PDF). We compare the marginal PDF of the SQI with the PDFs of the SQI conditioned to different values of the AMP variable. If the two variables are not related, the different PDFs are similar. If the variables are related, the PDF of the SQI will change as the conditioning values change.

Yield Figure 49 shows the results of the conditional probability analyses performed on crop yield as a function of irrigated area (in the grid cell) for the seven farming systems in Europe. We present the global PDF of crop yield (in blue) and the PDFs of crop yield conditioned to no irrigated area (in black) and to irrigated area greater than zero (in red).  

Figure 50 shows the results of the conditional probability analyses performed on crop yield as a function of irrigated area (in the grid cell) for the seven farming systems in China.

Figure 49. Europe
Figure 50. China

Cereal (1st row left), Rice (1st row right), Maize (2nd row, left), Soybean (2nd row, right), Vegetables 3rd row, left), Pasture (3rd row, right) and Permanent crops (4th row)

The effect of irrigation is more apparent in this analysis than in the regression analysis. By comparing the PDFs with and without irrigation we can estimate the expected effect of irrigation on crop yield. We work under the assumption that all other factors influencing crop yield (climate, soil type, other management practices, and other scenarios to be co-defined with stakeholders) have a similar effect in irrigated and non-irrigated areas. Although this assumption is certainly questionable, the results suggest that irrigation has a global positive effect on crop yields for most farming systems in Europe. The relationship is more clear for Maize and Soybean, and less apparent in Cereal and Permanent Crops. In the cases of Rice and Pastures the effect is also significant, but the discontinuities in the PDFs suggest that data may not be enough to draw a sound conclusion. The effect of irrigation in China is weaker than in Europe. Results suggest a positive effect on Rice and vegetables and a neutral or negative effect in the other farming systems, although the PDFs are very close in most cases.

Soil organic carbon Figure 51 shows the results of the conditional probability analyses performed on Soil Organic Carbon (SOC) as a function of irrigated area (in the grid cell) in Europe and China. They correspond to the total farming area, because no data were available for individual farming systems. As in the previous figures, we present the global PDF of crop yield (in blue) and the PDFs of crop yield conditioned to no irrigated area (in black) and to irrigated area greater than zero (in red).Results for both regions suggest that irrigation reduces soil organic carbon, or, at least, that it is generally applied to soils with less organic carbon content. However, the data available produce an irregular PDF, particularly in China, and this may introduce some uncertainty on the conclusions.

Figure 51 

Water holding capacity Figure 52 shows the results of the conditional probability analyses performed on Water Holding Capacity (WHC) as a function of irrigated area (in the grid cell) in Europe and China. Results show a positive effect of irrigation on WHC, particularly in Europe. This effect may be due to the fact that there is less probability to invest in irrigation of a soil has low WHC. This relation will be further explored in »Effect of management on soil quality.

Figure 52

Note: For full references to papers quoted in this article see

» References