|Main authors:||Ana Iglesias, David Santillán, Luis Garrote and contributions from ISS (China)|
|iSQAPERiS editor:||Jane Brandt|
|Source document:||Iglesias, A. et al. (2018) Report on definition of typical combinations of farming systems and agricultural practices in Europe and China and their effects on soil quality. iSQAPER Project Deliverable 7.1, 87 pp|
The knowledge that humans are impacting the environment at planetary scales has led to reflect on the scientific frameworks that would upscale the interactions and feedback mechanisms empirically observed at local scales (Verburg et al. 2016). The need to anticipate correctly the interactions at the higher level is driven by the scales of social-environmental policies (Vergurg et al., 2016).
Societies and ecosystems interact over many spatial and temporal scales (Cumming et al., 2006) and upscaling refers to the process of reconstituting activities or phenomena at a higher or larger geographical scale (Cumming et al., 2006).
The concept of scale has been extensively reviewed (e.g. Wiens 1989; Levin 1992; Gibson et al. 2000; Turner et al. 2001) and used in subtly different ways in biophysical and social sciences (Gibson et al. 2000). Social and geographical scales are often, but not always, aligned (Cumming et al. 2006). In bio-physical sciences, scale usually refers to the spatial and temporal dimensions of a pattern or process and it is also called “geographic scale”. Geographical scale has two main attributes: resolution of the observations and extent (Turner et al. 2001; Rietkirk et al. 2002). In the social sciences, scale includes the representative nature of social structures from individuals to organizations as well as the social institutions that govern the spatial and temporal extent of resource access rights and management responsibilities (e.g. Barbier 1997; Chidumayo 2002; Ziker 2003; Bodin and Norberg 2005).
Upscaling is often a form of extrapolation to a larger extent or coverage. Most social and environmental variables vary with extent but are not generally proportional to it. These scale-specific variables include the soil health and soil management data considered in iSQAPER. The upscale approach may be addressed by simple statistical transformation, but often the problem is better solved when we have an understanding of underlying social and ecological processes.
Depending on the research question or the environmental policy application, the appropriate resolution of the data varies temporally, spatially, and in the layers of information (e.g., in soil data, the profile). Regardless of the scale used to address a research or public policy question, the temptation is always there to extrapolate from fine-resolution data or to interpolate from coarse resolution studies. In both cases, the relevance of data and analyses conducted on one spatial level to other levels cannot be taken for granted. Spatial heterogeneity on the micro-scale may not be detected using coarse spatial resolution, and conversely, general patterns on the macro-scale may not be detected using fine spatial resolution (Turner et al., 1989; Levin, 1992; Wiens, 1989; Qi and Wu, 1996).
Several general questions need to be considered in geospatial environmental studies, including the following:
- what are the best criteria for selecting the spatial (and temporal) unit of intervention and analysis?
- how do the key measures of risk and management dynamics vary with scale?
- how do we integrate processes occurring at diverse spatial and temporal scales?
All of these questions can only be addressed through solid biophysical, agronomic and socio-economic understanding of the system in time and space.
The upscaling approaches depend on the research question and the spatial extent. The research questions we are addressing in this section of iSQAPERiS are:
- what is the effect of soil management practices on soil ecosystem services? and
- what is the environmental footprint of different climate and management scenarios?
The spatial extent of the analysis is the national or continental level in Europe and China.
Models that represent the scientific knowledge are auxiliary tools that may be used in upscaling, especially when it is necessary to represent socio-ecological processes, which is the case of iSQAPER (soil properties and management practices). Soil properties are represented based on the work carried out in other sections of iSQAPER. In contrast, management practices are socio-economic responses that include behavioural assumptions that are difficult to capture.
Validation of a model is good modelling practice, but is seen as an extremely complex challenge for integrated and complex system models (Parker et al. 2002). Procedures for evaluation and validation are rarely rigorously applied to the global-scale integrated assessment models used to inform major global assessments due to the lack of consistent time series of empirical data.
Overall, the design of conceptual models and the structure of modelling frameworks should be used as a tool to structure our current understanding of the system, rather than as a way to develop theory on socio-ecological systems.
The overwhelming number of possible feedbacks in complex systems can cause our models to become overly complex (Voinov and Shugart, 2013). Feedbacks make models extremely sensitive to error propagation in which small deviations in initial parameters can lead to large system-wide changes, especially in the case where the feedback is reinforcing itself.
In general terms, if we are to address policy-relevant issues in our approaches, we will need to provide a higher spatial and temporal resolution in our models accounting for the scales at which policy making operates.
Current large scale assessment models are not often taken very seriously by people in the region because they generate information that is simply not useful at the local level.
A few ways have been proposed to better incorporate these multi-scale issues in large-scale models. Most of these, reviewed by Ewert et al. (2011) for agricultural systems, are based on the linking of models operating at different scales in a top-down manner in which local dynamics are simulated in response to higher-scale model dynamics (e.g. Raworth 2012). Bottom- up interactions and feedbacks can conceptually be implemented in such coupled model systems but are only infrequently operationalized due to the complex and iterative interactions between models that would become necessary. Alternative approaches of capturing cross-scale dynamics by a more explicit representation of the scalar dynamics in a single approach have been given much less attention (Ewert et al.; 2011; van Wijk 2014). Some have warned that cross-scale dynamics are probably highly a-symmetric: where the importance of effects going up-scale (from land user up to global trade ﬂows and climate change) are likely to be relatively weak, the feedbacks from the global processes down to local land users are very strong (e.g. price changes, regulations, subsidies, etc.) (Giller et al, 2008). However, while we agree on the a-symmetry of these cross-scale dynamics these are strongly depending on the process characteristics and societal context.
An alternative approach is the upscaling of local dynamics through the identification of aggregate response patterns that are based on the scaling of local responses. Instead of representing the behaviour of individuals, in this approach the agency (aggregate behaviour) of communities is captured while still retaining the differential characteristics of these communities based on their composition and socio-cultural context (Dobbie et al. 2015). Upscaling may also be achieved through nesting detailed models at individual level within a more aggregate model to derive aggregate responses.
Upscaling methods that include modelling can be divided into four major classes (Bierkens et al. 2000):
- averaging observations or model outputs,
- finding representative parameters or input variables,
- averaging model equations, and
- model simplification.
The different classes are based upon five criteria (Bierkens et al., 2000):
- whether a model is involved,
- whether the model is linear in its input variables and parameters,
- whether the model can be employed at many locations or time steps,
- whether the form of the model is the same at the two scales involved, and
- whether the larger scale model can be analytically derived from the smaller scale model.
Geostatistical methods estimate variability as a realization from a stochastic function, where the weights depend upon both the sample configuration and a model of spatial-temporal structure estimated from the data (e.g. block kriging).
Approaches for finding non-exhaustive representative information include deterministic functions and stochastic methods (Bierkens et al.; 2000). Deterministic functions provide full coverage with a method of interpolation.
Stochastic methods, with statistics estimated from known information, describe unknown variations with conditional realizations from a stochastic function, ultimately yielding a single probability distribution.
Methods for averaging model equations (i.e., temporal/volume and ensemble averaging) and model simplification (i.e., lumped conceptual and meta modelling) are discussed by Bierkens et al. (2000).
Clustering methods are useful to detect the agricultural farming systems with different degree of impacts of the agricultural management practices on soil quality. Clustering methods use spatial statistics as exploratory tools that allow the detection and identification of clusters without a pre-determined hypothesis about cluster location (Besag and Newell 1991; Lawson 2001). Here we use a probabilistic approach to detect the significant effect of agricultural management practices on soil quality indicators. This approach may provide three measures of clustering:
- the nearest neighbor distance (i.e. the distance from the tested measurement);
- the maximum clustering distance (the distance where clustering is maximized); and
- the significant clustering distance (i.e. the distance at which clustering is statistically significant) (Getis and Franklin, 1987).
The fine scale resolution data, when available, can provide detailed information on the processes responsible for the effect of soil management practices, allowing testing and validation of spatial data and to improve the continental scale estimates. However, data for the spatial analysis and management decisions often take place on a much coarser resolution, and more general mechanisms may not be inferred from such fine resolution data (e.g. greening policies).
While models are mostly used as tools for researchers aimed at exploring system functioning, co-design and co-production of research has become important in global change research (Cornell et al. 2013).
Co-production approaches are used in decision support systems in which the algorithms are updated with stakeholder input during the process (Eikelboom and Janssen, 2013; Vonket al. 2005). Co-design is the central aim of »Soil management scenarios.
Here we try to provide answers to three questions: What are the appropriate levels of abstraction and representation given the questions we seek to address? What tools are available to use in iSQAPER?
The typical upscaling–downscaling exercise involves the following four steps (Bloschl 2005):
- analyzing the local data and scrutinizing the literature to decide on the model type,
- estimating the parameters from the data,
- verifying the upscaling–downscaling model against an independent data set, and
- performing the actual upscaling–downscaling step.
Bloschl (2005) discusses upscaling–downscaling for six important cases:
- upscaling point rainfall to catchments,
- temporal disaggregation of rainfall,
- statistical downscaling of the output of global circulation models,
- flood frequency as a function of catchment scale,
- upscaling and downscaling soil moisture, and
- subsurface media characterization and generation.
Overviews of both upscaling and downscaling methods are provided in the following articles.
Costantini and L'Abate (2016) reviewed the upscaling approaches that are available that represent aspects of soil health and soil management. Approaches differ in scope, purpose and structure. Most approaches are designed in response to either a science question or a management question, and address a specific spatial and temporal scale.
Here we intent to use the upscaling results to support management and policy decisions, and the questions posed by different stakeholders.
We will use scenarios to explore the possible outcomes of uncertain (societal) developments. Scenario simulations are important in raising policy issues and creating societal awareness of possible future challenges. Scenarios are used to capture some of the assumed range in uncertainty of major drivers of societal change such as population, economic development and policy.
For policy design, upscaling and scenario analysis can play a role in designing possible solutions, e.g. the optimal distribution and location of soil management choices. In this sense, the approach is goal-oriented and may use optimization techniques to design solutions accounting for present and future boundary conditions set by the socio-ecological system (Seppelt et al., 2013). Although such approaches can account for the constraints associated with the implementation of the prescribed ‘optimal’ management, they do not provide insights in the pathway to achieving these outcomes.
By making simulations the approach can support target-setting by analysing the trade-offs resulting from alternative ‘optimal’ management strategies.
Alternatively, upscaling can be used to investigate the effectiveness and unintended consequences of proposed policy measures through ex-ante assessment (Helming et al., 2011).
Appropriate methods must reconcile different spatial and temporal scales. Appropriate approaches require insights into the questions posed by a range of stakeholders, and address the concerns of policy makers and society as a whole.
How do current upscaling approaches handle these issues and address the questions of soil environmental footprint? The selected set of soil quality indicators can be integrated in various ways to combine the relevant dimensions of environmental footprint. In this section of iSQAPERiS, we direct our attention to typical combinations of farming systems and agricultural practices. In particular, a cluster analysis of the five farming systems indicators is employed to investigate the structure of the data space. Here, crop dimensions remain transparent, as they are not merged into one value which is a usual procedure in conventional studies. In contrast, the cluster analysis keeps the individual dimensions discernible. The cluster method, however, does not automatically generate an environmental footprint indicator ranking. This needs an additional qualitative interpretation of the different clusters. The qualitative interpretation is feasible because it has to be performed only for a limited number of resulting representative indicator combinations.
Note: For full references to papers quoted in this article see