Defining cost-effective ways to improve ecosystem services provision in agroecosystems

Abstract

Mitigating climate change through the adoption of environmental-friendly agricultural practices also affects biodiversity and the provision of other non-marketed ecosystem services (hereafter, ES). In this paper, we investigate a method to identify cost-effective strategies to improve the provision of these ES. We model the link between agricultural practices and the provision of ES, to illustrate the general antagonism between agricultural production and the provision of non-marketed ES, as well as synergies among the latter. We run efficiency analyses on the simulated agroecological data to explore the interactions among ES and identify efficient bundles of ES. Improving the provision of non-marketed ES comes at a cost in terms of production. The bundle of ES provided by an alternative management option has an opportunity cost corresponding to the profit loss compared to the most profitable management option. We determine which strategy costs less to improve the provision of non-marketed ES: to adopt a given set of agroecological practices over the whole agricultural area, or to dedicate only a part of the landscape to the provision of the non-marketed ES. This result is helpful to determine if agroenvironmental policies should target large areas with uniform low requirements, or several smaller areas with higher environmental conditions. It can be used to determine cost-effective ways to mitigate climate change through agricultural practices reducing greenhouse gases emissions and increasing carbon storage in soil while maintaining other ES.

Appendices

Appendix A. Agroecological model: mathematical details

A.1 Notations

We denote a particular field/area by the index x and a particular time by the index t.¹⁴ Management options are denoted by k. They correspond to a combination of agricultural practices. At each time, each field/area has a management option denoted by k(x,t).

Control variables correspond to agricultural practices. We consider the following practices:

• The land use U = {G; C}, corresponding to the choice between grassland or cropland

• pesticide intensity, with three levels: FTI = {0; 1; 2} (no pesticides, medium, or high use)

• fertiliser intensity: F = {0; 1; 2; 3; 4} (from no fertiliser to high input)

• presence of non-crop habitat such as grass or flower strips: NCH = {0; 1} (no; yes)

• biomass input such as crop residues or cover crops: BI = {0; 1} (no; yes)

• tillage regime: T = {0; 1} (conventional tillage; conservation/reduced tillage)

A management option k(x,t) is thus a combination within the set U × FTI × F × NCH × BI × T. Grassland options always coincide with no pesticide use, no fertiliser input, no NCH, no biomass input and no tillage. Considering all possible combinations, we obtain 122 management options.

There is a single state variable in the model, the soil organic matter SOM_x(t).

A.2 Equations for ES assessment

Soil organic matter and nitrogen

The evolution of soil organic matter SOM_x(t) is given by

$$ SOM_{x}(t+1) = SOM_{x}(t) - (m_{k(x,t)} + \lambda_{k(x,t)}) SOM_{x}(t) + I_{k(x,t)} $$

(2)

The mineralisation rate m_k(x,t) and the organic matter leaching rate λ_k(x,t) depend on the management option k. So does the input of organic matter I_k(x,t), which is the sum of crop residues and non-harvested part of the plants.

The total available (mineral) nitrogen is

$$ N_{x,t} = \frac{c_3}{c_2} \cdot m_{k(x,t)} SOM_{x}(t) + f_{k(x,t)} + LN_{k(x,t)} $$

(3)

where c₃ and c₂ are conversion parameters to calculate the amount of nitrogen in soil organic matter, f_k(x,t) is the mineral nitrogen from applied fertiliser, and LN_k(x,t) is the mineral nitrogen stemming from livestock (if relevant).

Nitrogen emitted as nitrous oxide is proportional to total mineral nitrogen

$$ N_{Ax,t} = \beta N_{x,t} $$

(4)

with β is the rate of denitrification.

Crops take up part of the nitrogen available for plants, i.e. N_x,t − N_Ax,t. Up to a certain amount N^∗, nitrogen uptake by crops is proportional to the nitrogen available. Above this threshold, the nitrogen uptake slows gradually down as nitrogen available for plants increases.

$$ \left\{\begin{array}{ll} N_{Px,t} = \gamma (N_{x,t} - N_{Ax,t}) & \text{ for } N_{x,t} - N_{Ax,t} < N^{\ast} \\ N_{Px,t} = \gamma N^{\ast} + \frac{\gamma(N_{x,t} - N_{Ax,t}- N^{\ast})}{1 + \epsilon(N_{x,t} - N_{Ax,t} -N^{\ast})} & \text{ for } N_{x,t} - N_{Ax,t}\ge N^{\ast} \end{array} \right. $$

(5)

where γ is the nutrient use efficiency, and N^∗ and 𝜖 parameters determining the shape of this hyperbolic function.

Eventually, the remaining nitrogen N_Wx,t is leached to water bodies:

$$ N_{Wx,t} = N_{x,t} - N_{Ax,t} - N_{Px,t} $$

(6)

Greenhouse gases

Greenhouse gases come from 4 sources: emission of nitrous oxide N_Ax,t, changes in soil organic carbon stock ΔSOC_x,t, fossil fuel burning FC_k, and methane emitted by livestock methane_k(x,t).

$$ GHG_{x,t}=g_1 c_4 N_{Ax,t} + FC_{k(x,t)} +{\Delta} SOC_{x,t} + g_2 . \text{methane}_{k(x,t)} $$

(7)

where g₁ is the global warming potential of nitrous oxide and c₄ the conversion parameter of nitrogen into nitrous oxide, and g₂ is the global warming potential of methane.

The change in soil organic carbon is proportional to the change in soil organic matter, with c₃ the carbon content of organic matter:

$$ {\Delta} SOC_{x,t} = SOC(t+1)-SOC(t) = c_3 (SOM_{x}(t+1) - SOM_{x}(t)) \\ $$

(8)

It is an indicator of carbon storage.

Plant growth - production

In cropland, potential yield Y_1x,t depends nitrogen intake N_Px,t and soil quality Q_x

$$ Y_{1x,t} = Q_x (1-\exp^{-n_2 N_{Px,t}}) $$

(9)

with n₂ the marginal effect of nitrogen on yield.

Non-crop habitats reduce cultivated area and thus the potential yield after accounting for the area really cultivated is

$$ Y_{2x,t} = Y_{1x,t}(1-e \cdot NCH_{k(x,t)}) $$

(10)

with e the proportion of the field dedicated to non-crop habitat.

Eventually, damage due to pests reduces yield, the final yield equals.

$$ Y_{3x,t} = Y_{2x,t}(1-D_{k(x,t)}) $$

(11)

Pests and damage D_k(x,t) are supposed to be proportional. Pests feed on crop, so that their carrying capacity depends on the yield, and thus damage is expressed as a fraction of yield. This fraction only depends on the intensity of pesticides.

Crop residues are proportional to crop yield.

$$ Y_{Rx,t} = \rho Y_{3x,t}(1 - BI_{k(x,t)}) $$

(12)

where BI is the binary associated to the crop residue restitution, which equals 1 if crop residues are left on the field.

Water quality

Water quality over the landscape is given by

$$ W_t= \min \{PL_{t} ; NL_{t} ; ML_{t} \} (1-w \sum\limits_x {NCH_{k(x,t)}}) $$

(13)

where PL_t, NL_t, and ML_t are functions expressing pollutant loads in the landscape and w the reduction of pollutants export due to semi-natural elements (in percentage).

Water quality score of the landscape for pesticides:

$$ PL_{t} = \frac{{\sum}_x{FTI_{k(x,t)}} - \underline{PL}}{\overline{PL} - \underline{PL}} $$

(14)

with $\underline {PL}$ and $\overline {PL}$ minimum and maximum levels of pesticide load over the landscape. Here the minimal load is achieved when no farmer uses pesticides and maximal if every farmer uses pesticides.

Water quality score of the landscape for nutrients:

$$ NL_{t} = \frac{{\sum}_x {N_{Wx,t}} - \underline{NL}}{\overline{NL} - \underline{NL}} $$

(15)

Again, $\underline {NL}$ and $\overline {NL}$ describe the minimal and maximal nutrient loads of the landscape. $\overline {NL}$ corresponds to a landscape with high levels of fertilisers, soil organic matter, and conventional tillage.

Water quality score of the landscape for organic matter:

$$ ML_{t} = \frac{{\sum}_x {ML_{x,t}} - \underline{ML}}{\overline{ML} - \underline{ML}} $$

(16)

with ML_x,t = λ_k(x,t)SOM_x(t) the amount of soil organic matter leached on field x. $\underline {ML}$ and $\overline {ML}$ describe the minimal and maximal organic matter loads of the landscape. $\overline {ML}$ corresponds to a landscape with high levels soil organic matter and soil loss.

Pollination

Pollination source score

$$ PS_{x,t} = HF_{k(x,t)} \cdot HN_{k(x,t)} \cdot PM_{k(x,t)} $$

(17)

with HF_k(x,t) the suitability in terms of floral resources, depending on the management k (land-use, pesticide intensity, non-crop habitat), HN_k(x,t) the suitability in terms of nesting which depends only on management option k (land-use, pesticide intensity, non-crop habitat), and PM_k(x,t) a multiplier representing the decreased mortality of pollinators in fields with medium intensity or no pesticides.

A.3 Parameter values

Table 3

Appendix B. Agronomic contexts

The following table summarises the characteristics of the 10 agronomic contexts that we considered, with the exogenous soil quality index Q (representing the potential yield) and the corresponding soil organic matter (SOM), defined as the equilibrium value reached by the dynamic Eq. 2 under the more profitable management option for each agronomic context.

Table 4

Note that, although the decreasing stock of soil organic matter has an opposite effect on yield than the increasing soil quality, the effect of soil quality dominates the impact of soil organic matter, so that the yield increases in a monotonous way across agronomic contexts.

Appendix C. Output of the simulations*

Appendix D. Efficient bundles of ecosystem services in all agronomic contexts

Appendix E. Output of the cost-effectiveness analysis