Project Topics
Get Complete Project Material File(s) Now! »
Assessing and modelling the provision of multiple ecosystem services
A wide range of methods have been used to quantify the provision of ecosystem services in agriculture. In particular, most rely on models, which differ by their degree of complexity in terms of processes represented and by the drivers they consider. Few assessments rely only on indicators measured in field studies like Raudsepp-Hearne et al. (2010), first because of the lack of appropriate data, and second because models enable sometimes to better approximate ES than available field data: the percentage of forested land is a rather poor proxy of the recreational ES, while models may better approximate it. Modelling is also used to explore scenarios and alternative management options, and to integrate the future changes in drivers (among them climate change, as in Kirchner et al. (2015)). The drivers of ES provision are more or less detailed. Many authors rely on land use and land cover, while other consider more detailed drivers such as agricultural practices, climate scenarios or agricultural public policy scenarios.
Assessments based on land use and land cover
Land cover refers to the physical land type such as forest or open water whereas land use refers to the way people are using the land (cultivated, urban…). Both types of data are easily assessed via remote-sensing and satellite data, and included in GIS, so that they are handy to access and handle. Many assessments have therefore used such data to quantify the provision of ecosystem services. In this kind of assessments, land use is split into classes and the provision of ecosystem services is assessed on this basis. Diverse classifications exist and categories can be more or less precise: Nelson et al. (2009) distinguishes 9 broad land use and land cover classes, while Goldstein et al. (2012) uses 31 classes adapted to the ecoregion considered. Naidoo et al. (2008) use the GLC2000 classification developed by the JRC to assess carbon storage, which comprises 43 land use classes (from water to many detailed vegetation cover types).
A few studies then directly measure or gather data linking one type of land use to the provision of multiple ecosystem services (Raudsepp-Hearne et al., 2010; Fontana et al., 2013). But most authors then engage in modelling to infere ecosystem service levels from the land use and land cover, using various models. Chan et al. (2006) use very simple « benefit functions », combining land use data with other simple statistics. Many other assessments couple land use data with more or less complex simulation models. Bateman et al. (2013) and the UK National Ecosystem Assessment use a large dataset to estimate econometric relationships between ecosystem services and land use, which are not based on the underlying biophysical processes. Most other cases involve models which try to mimic biophysical processes. For example, the modelling suite InVEST gathers a whole collection of separate modelling units of ecological processes based on land cover maps, thus offering estimates of many ES (see box below). Nelson et al. (2009) and Goldstein et al. (2012) rely on it to assess alternative development scenarios of particular two regions. Similarly, Ruijs et al. (2013) simulate ES from land use data with several models: the integrated IMAGE model (MNP, 2006) and the GLOBIO model for biodiversity (Alkemade et al., 2009), among others. Modelling enables to simulate ecosystem service levels, but the issue in these models based on land use/land cover is that they fail to account for other drivers than land use, and they weren’t built to consider dynamics, spatial interactions or feedbacks between ecosystem services (Seppelt et al., 2011). In this sense, Eigenbrod et al. (2010) show that using proxies based on land use doesn’t provide a sound basis for assessment of ES.
Assessing and modelling the provision of multiple ecosystem services
Developed by the natural capital project, InVEST (Integrated Valuation of Environmental Services and Trade-offs) is a suite of models used to map and value ecosystem services. It comprises 18 distinct models of ecosystem services provision across terrestrial, marine and freshwater ecosystems (including for example carbon sequestration, timber harvest, crop pollination, marine water quality, nutrient and sediment retention etc.). The drivers considered in the models are mainly land uses and spatially explicit variables associated to the geography (climate, precipitations, soil characteristics, slope etc.), so that the models can be used together with GIS data, and can easily provide maps of ES. The aim is to provide decision makers with clues about the trade-offs and synergies among ecosystem services (both in biophysical and monetary terms), and to predict the multi-ES impacts of land use scenarios. The 18 models can be run according to several levels of precision. They are also distinct, making it easy for users to use the ones they are interested in, but this setting neglects any direct interaction among the provision of several ecosystem services. Possibilities of linking several modules exist however for marine and freshwater modules. For details, see Sharp et al. (2014).
The Integrated Model (TIM), developped in the UK NEA, follows the same logic, but gives even more emphasis to the economic valuation process. It includes the main ecosytem services of terrestrial ecosystems (agricultural and timber production, water quality, GHG emissions, recreation), and even biodiversity. The methodology differs much from biophysical models such as InVEST. Instead of modelling the biophysical processes which deliver ecosystem services, the models are econometric relations between a service and its drivers, estimated with available data. A detailed description is provided in Bateman et al. (2014).
Assessments based on more detailed drivers
Other assessments try to take into account more detailed drivers, and in particular agri-cultural practices. They are associated with integrated or coupled models, and generally more complex. They better represent agroecological processes and have the advantage to capture feedbacks between ES, as well as dynamic and spatial interactions.
For example, Bekele et al. (2013) rely on SWAT, an agricultural simulation model to assess several ES and include a mix of crop types and agricultural practices as drivers. Following the same strategy, Kragt and Robertson (2014) assess the impacts of two agricultural practices with a rather simple model coupling. Going fruther into complexity, Balbi et al. (2015) integrate several separate models into a unique coding system, and account for drivers such as agricultural practices (irrigation, tillage, and application of both organic and mineral fertilizers) and various environmental conditions. Groot et al. (2012) built their own integrated model to assess ES, which accounts for many variations in farming practices (different crop rotations and crop areas, amounts of green manure and feed crops, number of animals…).
Other authors rely on complex model coupling to assess multiple ES, taking intoaccount very detailed drivers (Schönhart et al., 2011; Kirchner et al., 2015). In both cases, the authors are able to simulate the impact of many management variables, policy scenarios, and external drivers and derive spatially explicit ES indicators for their case studies in Austria. However, their models rely on very extensive calibration and require very large databases to do so. What is more, the separate models have been designed separately for another goal and may be already complex, and coupling them increases this complexity. In this sense, Carpenter et al. (2009) criticise the use of model coupling because it doesn’t fit the scope and content of conceptual frameworks as the MEA. They plead for integrated models designed especially for the research question, that account for non-linear and abrupt changes and quantify trade-offs between ES, such as the one used by Groot et al. (2012).
Quantifying and representing the interactions among multiple ES
Ecosystems are characterized by close interactions and multiple feedbacks between or-ganisms and their environment, as a result, the multiple ES present in agricultural ecosys-tems are interacting with each other. These interactions may be due to a common driver influencing the provision of several ES, or to a direct influence of one ES on the other (Bennett et al., 2009). Most authors of multi-ES assessments study the links among the diverse ES and try to represent them together, using different techniques. They distinguish between two types of relations among ES: synergies (positive relations among ES) and trade-offs (negative relations). The exact definitions and the representations of these interactions differ according to the type of available data and the approach. For example a synergy can correspond to a spatial congruence among several ES (several ES are provided on the same places), or to the co-evolution of two ecosystem services (when one ES increases, the other also increases). And the same holds for a trade-off.
Flower maps
The most basic representation of multiple ES accross multiple scenarios or land uses is to use « spider webs » or « flower » diagramms: levels of the multiple ES are represented as bars or points on multiple axes starting from the center of the flower or the spider web. One diagramm is drawn for each land use or scenario, which are then compared. Without further analyses, interactions among ES are difficult to assess. An example of such diagramms, taken from Raudsepp-Hearne et al. (2010) is shown on Figure 2.1.
Spatial correlations and comparison of maps
Assessments aiming at comparing ES accross several development scenarios or different conservation schemes rely on spatially explicit data of ES provision over a landscape. Even if they could compare ES levels accross different management scenarios, they often study the spatial congruence of ES. Such approaches generally refer to synergies and trade-offs among ES in the meaning of (spatial) correlations or congruence.
Chan et al. (2006); Naidoo et al. (2008); Ruijs et al. (2013) focus on finding the best conservation networks, and proceed in a rather similar way as above, simulating the outcomes in terms of ES levels in each spatial unit for different conservation scenarios. They get results in the form of maps of ES, and by overlapping these maps, they determine « hotspots » where many ES are provided simultaneously. These assessments also calculate correlations between the levels of ES provided accross all spatial units and scenarios. They are thus able to identify some trade-offs and synergies between ES, which are indeed spatial congruences and spatial mismatches between ES. As the correlations don’t control for the specificity of each spatial unit and what drives it to provide the given ES levels, no conclusion can be made on what would happen to other ES if one ES were to be increased. Indeed, two ES like pollination and water quality may be spatially correlated (i.e. provided at the same place), but increasing pollination by sowing more pollinator-friendly flowers may not automatically result in increasing the water quality, because they depend on drivers that are not controlled for in the calculation of the correlation: all spatial units are compared with each other, no matter how heterogeneous they are.
Nelson et al. (2009) and Goldstein et al. (2012) study the ES provided by real land-scapes under different management and development scenarios. They use InVEST model to simulate the ecosystem services provided in different scenarios of land use. To repre-sent the results, they draw maps representing the change for each ES in each scenario and draw them side by side to be compared. These maps offer an easy representation of all consequences of every scenario over the whole landscape. In these two papers, the results are also analysed in terms of joint evolution of the different ES in each scenario, drawing some insights about trade-offs and synergies among ES. These trade-offs andFigure 2.2.: Maps of multiple ES in different land use scenarios in Goldstein et al. (2012) synergies identified depend of course on the landscape and the alternative scenarios, but contrary to the spatial congruences and mismatches, they inform rather well on the existing trade-offs and synergies incurred when choosing one or the other manage-ment scenario, because the alternatives compared are based on the same landscape. To illustrate, Figure 2.2 shows an example from Nelson et al. (2009).
However, both these approaches use simulation models that don’t account for direct feedbacks between ES, and thus are not suited to analyse precisely the relations (synergies and trade-offs between ES).
Use of PPF in the agricultural sustainability literature
PPF is a handy tool to represent trade-offs and synergies between ecosystem services, and possible outcomes of alternative management options. Using production possibility sets and frontiers meets several objectives.
First, the methodology is used to assess how management options score on several objectives: Kragt and Robertson (2014) consider only few management options to com-pare, but they are interested in their impact on several ecosystem services. By considering only few possible management options, their approach doesn’t assess fully the interrela-tions among ecosystem services, but they show that some agricultural practices (as crop residue restitution) is beneficial for many ES. In particular, they are beneficial for both production and regulating ES, which globally stand in a trade-off1.
Second, another objective of drawing a production possibilty set and identifying the frontier is to get an insight of its shape and of the underlying trade-offs and synergies. Using models simulating the outcomes resulting from spatially explicit combinations of land use, Polasky et al. (2008); Bekele et al. (2013); Teillard et al. (2016) assess the interrelations respectively between agricultural production and biodiversity, and among many ecosystem services 2. The mere shape of the PPF serves then as a basis to anal-yse the interactions among ES. Teillard et al. (2016) use the shape of frontier to show that the strength of the identified trade-off among biodiversity and production varies according to the position in the production set, and to the set of management options considered (agricultural intensification, extensification, reallocation of land uses). Po-lasky et al. (2008) conclude that PPF between agricultural production and biodiversity is probably concave, such that a significant increase in biodiversity conservation can be achieved with a very limited loss of agricultural production. The same conclusion is reached by Bekele et al. (2013) for an index aggregating several ES. The same type of argument used in the land-sparing/land-sharing debate (see section 2.3.2). The shape of the PPF is also related to the economic theory, and some authors explore this link. Using theoretical biological functional forms, Brown et al. (2011) show that the production pos-sibility sets can differ from the standard assumptions made in the microeconomic theory, and thus that the prescribed solutions can be ineffective or create contradictory effects. Ruijs et al. (2013) confirm this intuition. The authors estimate a production possibility frontier with 4 ecosystem services from simulated data, and identify that its shape is non-concave. As demonstrated by Brown et al. (2011), if the frontier is not concave, the determination of bundles that maximise ES cannot be made as easily. Groot et al. (2012) explore combinations of many management variables and simulate desirable char-acteristics of the management system (labor balance, agricultural output, environmental variables etc.) They use production possibility frontiers as a multi-objective optimization tool, even if dimensions considered are not all outputs in the strict sense.
Third, the interest of drawing the production possibility frontier is to identify the effi-cient bundles of ecosystem services, and the underlying configurations that lead to these bundles. Besides, if the current situation is comprised among the options represented, it is possible to compare it to the production possibility frontier. Since the frontier repre-sents the efficient situations, it is a way to assess the efficiency of the reference situation. Groot et al. (2012) thus show that substantial improvements are possible simultaneously on every objective (profit, labor balance, nitrogen losses mitigation, organic matter bal-ance), and identify which management options seem to be efficient. Similarly, Bostian and Herlihy (2014) explore in which extent wetland condition or agricultural production could be increased, either separately or together.
Production possibility sets and frontiers applied to agriculture and ecosystem services
Fourth, some authors use the production possibility frontier to derive opportunity costs of providing non-marketed ecosystem services. This requires to approximate the frontier by specifying a functional form and estimating parameters, in order to get an differentiable equation of the frontier. Sauer and Wossink (2013) estimate a production possibility frontier with multiple inputs and multiple outputs, and look at the sign of partial derivatives to know if the agricultural output’s value and the ecosystem service value (the payment for ES) show a trade-pff or a synergy. Bostian and Herlihy (2014) estimate the production possibility frontier of agricultural production and wetland con-dition and apply the duality theory to value improvements of wetland condition from the market value of agricultural production. Similarly, Ruijs et al. (2013) estimate the cost of providing non-marketed ES from the monetary value of lost production calculated with the slope of the estimated PPF. They use spatially explicit data, which enables them to go further and determine where the provision of ES is least costly in terms of agricultural production. All these estimated opportunity costs rely on the economic theory, which assumes that firms maximise their profit and usually run at full efficiency. A fifth objective, related to the estimation of opportunity costs, is to use the slope of the frontier to assess which types of regulation instruments are needed to encour-age the provision of non-marketed ES. Whether with theoretical developments (Smith et al., 2012) or with estimated frontier (Sauer and Wossink, 2013), the idea is rather simple: where increasing non-marketed ecosystem services translates into no loss, or a rather small loss of profit, information campaigns or technical advice may be enough to achieve an increase in ecosystem services. However, if the increase in ecosystem services comes with a sharp decrease in marketed commodities or a significant increase in costs, financial incentives are needed, which compensate for the (opportunity) cost of providing non-marketed ES.
Finally, PPF are a useful framework to create transdisciplinary discussions among ecologists and economists (Smith et al., 2012; Cavender-Bares et al., 2015). They are also able to gather representation of biophysical processes, as well as the representation of social preferences.
Table of contents :
I. State of the art
2. Literature review: assessments and representations of multiple ES
2.1. Assessing and modelling the provision of multiple ecosystem services
2.2. Quantifying and representing the interactions among multiple ES
2.3. Production possibility sets and frontiers applied to agriculture and ecosystem services
2.4. Stylised facts identified in the literature
3. The economics of regulating the provision of ecosystem services
3.1. Ecosystem services as joint public goods
3.2. The regulation of joint public goods
3.3. Some elements from connex literature
3.4. Contributions of this thesis
II. Methods
4. Agroecological and economic model
4.1. Presentation of the model
4.2. Framework and notations of the model
4.3. Pollination
4.4. Soil organic matter and nitrogen
4.5. Greenhouse gases
4.6. Agricultural production
4.7. Water quality
4.8. Profit
5. Simulated data
5.1. Summary of simulated data
5.2. Exogenous drivers
III. Economic incentives to maximise the provision of ES under a budget constraint
6. Maximising ES provision
6.1. Introduction
6.2. Production possibility frontiers and interactions among multiple ecosystem services
6.3. Efficient bundles of ecosystem services
6.4. Cost-efficient bundles of ecosystem services
6.5. The cost of two strategies to provide non-marketed ecosystem services .
6.6. Discussion and conclusion
7. Economic incentives to implement solutions maximising ES provision
7.1. Introduction
7.2. Literature review: incentives for the provision of multiple and interacting ES
7.3. Methods
7.4. Results
7.5. Interactions among ecosystem services and the excess budget
7.6. Discussion and conclusion
IV. What land heterogeneity changes
8. Heterogeneous areas
8.1. Introduction
8.2. Literature review
8.3. Heterogeneity in our framework
8.4. Cost-efficiency analysis on heterogeneous areas
8.5. Results: how heterogeneity changes cost-efficient solutions
8.6. Incentives in heterogeneous agricultural areas
8.7. Discussion and conclusion
V. Discussion and conclusion
9. Discussion
9.1. Current context in terms of incentives
9.2. Efficiency of agroecological solutions
9.3. Accounting for the cost of ES provision
9.4. Incentives
9.5. Coping with heterogeneity
9.6. Concluding remarks
Résumé étendu en français
1. Introduction
2. Chapitre 2: État de l’art
3. Chapitre 3: Régulation de biens publics joints dans la théorie économique
4. Chapitres 4 et 5: Modèle et données simulées
5. Chapitre 6: Maximiser la fourniture de services écosystémiques pour un budget donné
6. Chapitre 7: Incitations économiques
7. Chapitre 8 : Le rôle de l’hétérogénéité
8. Discussion et conclusion
Detailed outputs of simulations and analyses
1. Output of the simulations
2. Efficient bundles of ecosystem services in all agronomic contexts
3. Output of the cost-efficiency analysis (context 8)
Model calibration
1. Summary table
2. Pollination
3. Pests
4. Soil organic matter and nitrogen
5. Greenhouse gases
6. Water quality
7. Agricultural production
8. Profit
Bibliography
