morocco_fig_6-25b.jpg
Main Menu
Home
News
Project information
Information System (HIS)
Calendar
DESIRE publicity
Links
Download
Sitemap
Contact us
Search
Upcoming events
Partner login





Lost Password?
Statistics
Visitors: 5364880

DESIRE logo


Locations of visitors to this page
(per August 25, 2010)

 

Home arrow Working blocks arrow Work Block 5: Regional evaluation of remediation strategies
Work Block 5: Regional evaluation of remediation strategies PDF Print E-mail
Written by Joost   
Coordinator: 3 – University of Leeds, UK
Participants: 1, 2, 5, 7, 8, 9, 11, 15, 16, 18, 19, 21, 22, 23, 24, 26, 27, 28

Objectives
a)    Develop a model for the main bio-physical and socio-economic processes interacting within an agroecosystem, building on existing experience in combination with results generated within WBs 1-4;
b)    Calibrate this model using data from field sites (WBs 3 & 4);
c)    Apply the model to the agroecosystems around key field sites to evaluate the likely environmental, environmental and social effects of proposed remediation strategies (identified in WB3);
d)    Evaluate model outputs with stakeholders and short-list remediation strategies for wider application by policy-makers and extensionists (to be disseminated in WB6).

General information
The purpose of WB5 is to evaluate the likely effects of remediation strategies (trialed in hotspots in WB4) at a regional scale, in order to make policy and extension recommendations. This will be done through a combination of integrated modelling and multi-stakeholder dialogue. This will be an area substantially larger than the hotspot trial areas, such as the river basin in which the hotspot is situated.
Extrapolation of the effects of the conservation plans for a larger area than the hotspot area can be directly based on the indicators which are defined in WB2 and given a scientific basis in WB4. They can be extrapolated more or less directly, using the geoinformation in the harmonized information system HIS (WB6), e.g. topographic data, earth observation land use maps and soil information. This will result in an aerial extent of desertification and the effects of the conservation measures and a qualitative appraisal of the severity.
A more quantitative modeling approach will be followed to develop a limited number of future scenarios that show the effects of land management and/or climate change on desertification. To do this, we will combine an agent-based model with both a regional economic model and a natural environment model. The agent-based model uses simplified rule based human-environment interactions, whereby the behavioural rules are partly based on the outputs of the environmental model. This will allow us to simulate landscape dynamics through the interactions between the socio-economic and environmental system (Parker et al, 2003). Regional economic models (based on input-output analysis) provide quantitative information about production and consumption (Idenburg, 1993). The environmental model is based on the hydrological large-scale model PESERA (Pan-European Soil Erosion Asessment), that has an explicit basis in physical hydrology and that can be extended to include vegetation dynamics.

Image

Fig. 4.4. Activities within WB5 and linkages with other WBs


Land-based activities play an important economic role in each of the dryland countries selected for this research, and have a variety of implications for land degradation, including pollution, erosion and loss of vegetation palatability and diversity. Thus an integrated systems approach is required to model land use decisions whilst reflecting changes in society such as demographic trends or changes in consumer preferences. Fig. 4.4 indicates some of the main activities that will take place in WB5, together with their interdependence on other work blocks.

Modelling strategy
The proposed models for integrations of biophysial and socio-economic components falls into two categories; micro-level/local models of biophysical and socio-economic processes and macro-level/regional models of relationships between the economy and the environment.

Micro-level models: Linking PESERA and models of land use decision making
The proposed biophysical model is based on the PESERA model that has been used in a number of recent and current EU projects (including DESERTLINKS, tempQsim and DESURVEY).  PESERA is considered to have a sound physical basis, to provide a reasonable compromise between data availability and process details, has been validated by a number of studies, and provides a wide range of useable outputs that help to link it with socio-economic models. On the basis of current development, the model has been undergoing further development to meet a number of needs that have already been identified for DESIRE such as modifying PESERA to capture the role of grazing, fire and wind erosion more effectively. In addition, existing work in a UK-funded Rural Economy and Land Use (RELU) project has prompted a more detailed consideration of the extent to which PESERA can be downscaled to finer grid sizes and incorporate an improved treatment of cell-to-cell transfers within hillslope areas.
For this project, we will couple PESERA with socio-economic model components that capture land use decision making. The socio-economic model will be parameterised from questionnaires using methodologies developed in experimental economics. Specifically, questionaires will elicit preferences for different land use attributes from a range of different land user groups and will enable us to assess how land users may respond to different desertification remediation policies.
This approach will also allow us to characterise different groups of land use actors who may have different land use objectives and behaviours. Simulating both land use behaviours on and biophysical responses for specific landscapes will enable an evaluation of the environmental and socio economic consequence/response to alternative remediation policies. Given the ongoing work in this area, it seems advisable to continue to work within proposed framework, and to optimise compatibility  between the biophysical and socio-economic components, rather than to work with other models that have been less well tested at Europe-wide scales.

Macro-level models: Socio-economic input-output modelling for assessing alternative land management options
The agricultural sectors play an important role within an economy in terms of economic output (value added) of associated sectors and jobs created, indirect effects on sectors such as tourism but also its environmental effects such as water consumption or pollution and soil degradation. Thus an integrated systems approach including biophysical and socio-economic models are required to provide an accounting and modelling scheme for assessing changes in land management within the production-consumption cycle.
Structural economics, with input-output (IO) analysis at its core, is a framework for organizing quantitative information about production and consumption activities into a database and using it to analyse scenarios about the future. Scenarios can be changes in lifestyles (e.g. through demographic changes, increasing rate of urbanisation – see e.g. changes in the economy (e.g. growth or decline of certain sectors – e.g. (Hubacek and Sun 2001), social or economic policies, and changes in availability of environmental resources (Duchin and Hubacek 2003; Duchin and Lange 1994).
The IO framework can be used to model interdependencies in both the money economy and in the physical economy. Thus it can be used in models integrating the economic system and the ecological system (Costanza 1991). The missing link in this is the modelling of the physical economic system (Idenburg 1993). The production system as described by an input-output table shows the flows of goods and products between economic sectors, such as agriculture, and various types of industries and service sectors over a year. Each economic sector is represented by a column of inputs and a row of outputs. Similarly, the consumption side (final demand) of an economy can be also represented by input columns representing different consumption patterns of private households or the government.
In order to combine value and physical data within a consistent methodological framework, we propose extending the IO tables by a set of natural resource parameters that represent consumption patterns of environmental resources in each economic sector. The enlarged IO table, as presented below, provides an accounting scheme for economic activities (zij, Vkj), household and other final consumptions (uis), environmental inputs (Lrj and Lrs), and effects on the environment (dir). The inner parts in the table are in monetary units and the outer parts are in physical units (e.g. in cubic meters).


    Grain, Other-crop, livestock, …    Rural, Urban, …    Total Output    Natural resources
Grain, Other crop,
Livestock    Inter-industrial flows (zij)    Final deliveries (uis)    Goods & Services deliveries  (Xj)    Depreciation & Degradation
(dir)
Capital
Labour    Factor inputs (value-added) (vkj)           
Total Input    Goods, Services, & Factors inputs (Xi)           
Natural resources    Natural resource inputs (Lrj)    Natural resource uses (Lrs)       

Table 4.2. A schematic presentation of an extended input-output table
(Source: modified after Hubacek and Sun 2005).

Economic impact studies use I-O tables to evaluate changes in final demand and technology and their effects on output, employment, and income in a region. Thus scenarios representing demographic changes, changes in lifestyles, consumer choices or technical change can be evaluated by tracing all the ripple (i.e. lifecycle) effects through an economic system.
The combination of the IO analysis combined with the agent-based modelling will be used to develop a database that will then be used to calculate socio-economic indicators such as value added, number of jobs, changes of prices as well as environmental indicators in physical units such as land appropriation and water consumption. The derived indicators are then inputs for the participatory process, the biophysical modelling and information that may be used in the characterisations of the agents in the ABM.

Validation of biophysical and socio-economic models
The evaluation and validation of models will occur at two levels: the sample level and the landscape level. Comparisons of the models for different landscapes will help to further our understanding how local environmental, socio-economic and institutional conditions contribute to land degradation.
For both the sample level and the landscape level, land use decisions are the key variable to test, as the cross-over variable that links socio-economic and biophysical models, and as one of the most crucial variables to test the effectiveness of the models. Additional relevant variables include more direct measures of progress towards or away from desertification, for example, soil organic matter content, biodiversity and erosion rates, but data for these variables are not as extensive as for land use, so that they are less useful for validation.

At the sample level
The land use choices will be modelled based on hypothetical choices under different scenarios reflecting the identified drivers of land use change. The success of the models in capturing land use behaviour will depend primarily on, how well defined the different land use options are, and to what extent they are related to individual farmer characteristics and identifiable external socio-economic and environmental conditions. Estimation of land use decision models will give us confidence intervals on parameter values. Furthermore, the use of mixed logit specifications will enable us to explore the extent of heterogeneity in land use preferences across the population of land users. We will use standard maximum likelihood based test statistics to test overall model performance. The models however, are based on hypothetical choices and the validation of the outcomes on the landscape level will be essential for reliability of the policy recommendations.

An assessment of the uncertainty related to the biophysical models will be carried out for the test sites. We intend to approach the uncertainty of model outcomes through Monte-Carlo simulation of the model, drawing parameter values from the distribution of errors where this is defined, and using fuzzy logic methods to constrain distributions of qualitative factors.  Although model complexity is likely to allow only limited exploration of this parameter space, we expect to define uncertainty bands on our forecast outcomes, for both numerical (such as erosion rates) and nominal (such as land use categories) variables. One important aspect of this analysis is to define areas where uncertainty bounds are wide, which may indicate either poor understanding or an inherent uncertainty which may offer greater scope for policy initiatives.

At the landscape level
Results from interviews on farmers’ hypothetical choices will become the rules of the land use agents that inform the land use models. These models will then be validated (and uncertainty bounds established) by using them to predict landscape patterns for areas where we have real land use maps. This will allow us to compare real world versus predicted land use patterns. Additionally, we will also conduct an “internal validation” of the landuse models by confirming that the results reflect stakeholder perceptions of the important factors involved in land use decision-making. The internal validation of the biophysical sub-components will also be used to confirm that the processes represented in the models reflect current knowledge/perceptions of the dynamics of real biophysical mechanisms. 

Comparisons between landscapes
There is likely to be limited scope for global comparisons in land use decision processes, and it is unlikely that socio-economic models will be transferable from one site to another. For example, we expect that it would be impossible to directly transfer results, because land use options and institutional as well as personal decision making processes differ vastly from region to region. However, we expect that the process of soliciting interview responses, building predictive models, integration with biophysical models and validation of models should help refine principles and generic processes such that future landuse decisions modelling can be more efficient in other sites. 

WP5.1  Coarse scale model development
In order to evaluate the likely effects of remediation strategies (trialled in hotspots, WB4) at a regional scale and make policy and extension recommendations, we will scale up results from field trials and interviews to evaluate regional implications. To do this, we will combine an agent-based model with a regional economic model and a model of the natural environment that is an extension of an existing hydrological model. The model will be adapted to different study areas to reflect indicators and land degradation drivers identified in WBs 1 & 2. This combined model will be used to determine the effects of proposed remediation strategies (identified in WB3) at a regional agro-ecosystem level.
An agent-based model will be used to represent a simplification of complex human-environment interactions, established through the definition of rules, which economic actors or “agents” use to achieve their goals. These rules collectively represent the ‘rational’ behaviour of the agent (Janssen, 2002). Agents interact with their environment and communicate with one another depending on the behavioural rules applied in the model. Integrating physical environmental outputs from the biophysical models with defined rules of agent behaviour, will allow us to simulate landscape dynamics through the interactions between the socio-economic and environmental system (Parker et al., 2003). The decision rules will be quantified based on questionnaires administered at selected study sites during WB3. Answers to a series of hypothetical choice questions will be used to develop decision trees that describe land management decision making processes for different types of agents and their response to the key drivers identified in WB1. Where appropriate data is available the outputs from the agent-based model will be validated against existing land use cover and used to generate a range of spatially explicit scenarios of land use change in response to key drivers identified in WB1. The outputs will be provided at a spatial scale representative of the data available and the relevant management units of the land managers. The agent-based model will also be able to provide information about the likely uptake of different remediation strategies across the landscape under different scenarios. This information will then be fed into a series of coupled biophysical and economic models to evaluate the likely environmental and socio-economic effects of different remediation strategies under different future scenarios. Application to a range of sites allows us to evaluate the extent to which the modelling framework is transferable between specific sites and enable assessment of transfer errors. This is important when considering applying site-specific models at a larger regional scale.  
Regional economic tools such as input-output analysis will be used as a framework to evaluate the likely regional socio-economic effects of different scenarios (based on drivers and indicators identified in WBs 1 & 2) and of different remediation strategies under each of these scenarios. Input-output analysis is well suited to model not only the initial but also second and third round effects, etc. of various activities and remediation measures by capturing all the ripple effects within an economic system. Moreover, input-output analysis can provide a powerful framework for organizing quantitative information about production and consumption activities and to model and analyze scenarios about future change, policies and land use implications. The IO framework can be used to model interdependencies in both the money economy and in the physical economy. Thus it can be used in models integrating the economic system and the ecological system ((Idenburg, 1993, Costanza, 1991; Hubacek and Sun, 2001). Given appropriate data and tables this will allow us to upscale farm-level decisions to aggregate systems effects on the level of the regional or national economy.
The regional environmental effects of proposed remediation strategies will be evaluated through a series of coupled biophysical models. Biophysical model development will be based around an existing hydrological model called PESERA (Kirkby and Neale 1987; Kirkby et al. 2000), developed by partners in EU projects MEDALUS (1991-99) and PESERA (2000-03). This model is being used to predict runoff and erosion across Europe, and has guaranteed technical support for at least 5 years through other EU projects. The model will provide a core bio-physical platform into which additional elements can be built to reflect drivers and indicators identified in WBs 1 & 2 for each site. The particular value of PESERA as a core model is that it is implemented within ARC-GIS.  It has an explicit basis in physical hydrology, based on a 1-D (vertical) partition of precipitation, with validation across Europe at a 1 km grid resolution for sediment yields. It can, however, be adapted to application at a variety of scales (it is already being implemented at resolutions between 1 km and 50 m across Europe). The use of PESERA as the model platform will also enable us to ensure that information is available in a spatially-distributed way with topography and other factors incorporated. It will be possible to show how the same management in one part of catchment will have a different impact/risk associated with it compared to that management occurring in another part of the catchment. This will allow the development of distributed decision-making in a move away from simple blanket policies.
The core of PESERA is a one-dimensional (vertical) hydrological partition of precipitation, which makes use of climate, soils, geology and land use data to estimate evapo-transpiration, snow pack and meltwater, Hortonian and saturation-excess overland flow, shallow sub-surface lateral flow and groundwater recharge. Land use data is be used to inform a growth model for biomass, cover and soil organic matter, taking account of fertilizer, grazing and other agricultural exchanges. The distribution of overland flow runoff events is then used, with soil, land use and relief parameters, to estimate total erosion loss from the land. The model is currently being extended, within the tempQsim and DESURVEY EC projects, to estimate non-point-source losses of total N, P and C, both in solution and adsorbed to sediment. Other developments are being designed to represent wind erosion, grazing intensity and fire risk. The combined model will provide information on rates of runoff and water quality, linked to explicit vegetation growth and/or crop models. It provides an explicit link between future land use and possible changes in land degradation status. 
Finally, information about the likely regional economic and environmental effects of proposed remediation strategies under different scenarios can be fed back into the agent-based model to see how these effects are likely to influence land use decisions. Consequent changes in land use decisions can be fed back into the economic and biophysical parts of the model to generate new environmental and economic outcomes for the region. In this way, the model not only captures the likely effects of different remediation strategies; it can also capture likely responses of land users to remediation strategies, and provides information about likely uptake under different scenarios. By contrasting scenarios created using different drivers from WB1, it is also possible to examine the extent to which changes in policy could affect land use decisions and uptake of different remediation strategies.
This integrated model will be developed to capture key drivers and (biophysical and economic) indicators relevant to all study areas, and adapted to represent key additional drivers and indicators specific to individual study sites. As such, the core model will provide a coarse but widely transferable evaluation of remediation strategy effects that could be applied beyond this project, but that is sufficiently adaptable to provide more nuanced results for specific regions if desired. This approach is currently being developed and applied in ongoing research by University of Leeds partners, funded by the UK Government Research Councils (Dougill et al., 2006). 

WP 5.2: Fine scale modeling and pedotranfer functions
Clearly PESERA is not a comprehensive model and should be used together with other models to take into account processes that are scarcely visible at a regional scale but that are the drivers for a series of interrelated events and actions leading to degradation and desertification. One important process of this kind is tillage, which continuously modifies field (and slope) morphology and reduces soil fertility (Van Oost 2006, De Alba et al. 2006). Management practices such as land leveling (Borselli et al. 2006) also strongly modify soil structure and soil depth, as well as transforming slope morphology. Land levelling is widely used for implementing major land use changes (e.g., almost essential for terracing vineyards and fruit tree plantation). Other significant fine-scale processes include gully formation and headcut retreat, and piping, which is usually strongly associated with terracing in arid and semi-arid land (where terraces usually end with a counter-slope to harvest local runoff). In sub-humid environments shallow mass movement are also widespread and constitute one of the major factors of soil degradation (e.g. in Southern Italy). Gullies, pipes and shallow mass movements are usually removed from fields using bulldozers, sometimes using large mouldboards. Hence, the intrinsic degradation provoked by these processes leads to remedial action which can solve the trafficability of the field but further modifies soil characteristics. These fine scale processes must be taken into account on a larger scale for evaluating the rate of desertification and for scenario analysis. Similarly, vegetation interacts and modifies the outcome of any rainfall event, both for different plant species but even on an individual plant level This can be understood and dealt with if a good description of the plant architecture and plant water needs are known. Natural and semi-natural vegetation stands are usually composed of a mixture of many plant species. This is also true also for many new and traditional crops (e.g. agro-forestry systems, see e.g. Diemont et al. 2006). The multispectral Remote Sensing techniques used in WB4 will help to acquire this spatial information. A number of fine scale models exist that use this information to quantify water fluxes. Currently these models are being put together for describing a relatively complex vegetation systems in RECONDES (due to end in January 2007, see project list Annex 1).  Each plant species is described through its degree of soil surface cover, mean height of fall of re-dripped drops, plant water storage (related to LAI), and plant structure.
Soil is invariably at the core of better understanding of degradation processes. We need good rules for estimating relevant soil properties from more basic soil characteristics. This is usually achieved using pedotransfer functions (PTFs) which have been developed to satisfy input requirements of simulation models of water and solute transport in field conditions (Wösten et al., 2001). Since PTFs link soil properties to hydrological characteristics, changes of these properties caused by land levelling, tillage and of renaturalization can be translated to model input parameters. PTFs are mainly based on soil texture, organic carbon content and/or bulk density, where sometimes accuracy is gained by grouping soils accoreing to their texture class. Recently soil structure has been included in PTFs (Ungaro et al., 2005). A number of PTFs reviews are available in the specialised literature (Van Genuchten and Leji, 1992; Tietje and Tapkenhinrichs 1993; Wösten et al., 2001; Pachepsky and Rawls, 2004). Public soil databases have been established for the purpose of PTFs development (UNSODA, Leij et al., 1996; HYPRES, Lilly, 1997; WISE, Batjes, 1996).  The level of reliability of any given PTF is strictly related to the specific composition of the calibration data set which in turn may reflect the geographic origin of the data set and the relevance of the different pedogenic processes in a given area. For these reasons the extrapolation of PTFs should always be preceded by a careful evaluation  of their applicability to specific data sets (Ungaro et al., 2005). In the last decade research provided new estimation tools other than multiple regressions: artificial neural networks (see e.g. Schaap et al, 1998;  Minasny and McBratney 2002), group method of data handling (Rawls et al.,1999; Ungaro et al. 2005), regression tree modelling (McKenzie and Ryan, 1999). Soil texture, organic matter content and bulk density, being related to landscape position and land surface shape, terrain attributes have been related to soil basic and hydraulic properties improving hydraulic conductivity estimation in specific environment (Romano and Chirico, 2004). More recently changes in vegetation and land cover and were found to improve the accuracy of locally calibrated PTFs and this represents a promising new direction in PTFs development.

WP5.3 Model integration, calibration and application
The coarse scale model and fine scale models developed in 5.1 and 5.2 will be integrated and calibrated using existing data (identified in WB1), pedo-transfer functions from 5.3, information from interviews conducted in WB3 and data collected in field trials (WB4). The fine scale models of WP 5.2 will be used to produce maps of shallow mass movements, gullies, pipes, tillage and land leveling modification of landscape at the appropriate (local) scale for representing them. Their occurrence will then be compared with landscape characteristics available at coarser scales. This will identify the coarser possible scale at which the occurrence of the said processes can still be estimated and algorithms will be developed for making the estimation possible using coarse scale data (evaluation of the deterioration of information). Vegetation effects will be similarly upscaled. Algorithms and/or models will be linked to PESERA both via modification of input data (e.g. effect of land leveling on soil characteristics and slope morphology, effect of vegetation cover on runoff generation) and/or using PESERA simulation results for triggering gullies/pipes and shallow mass movements.
Detailed calibrations of the fine scale models will be possible within the field areas, and will be used to provide a sub-grid model, providing physically based factorial corrections for application at the coarse scale, both within and outside the specific test areas.
The coarse scale model will be applied to the regions around each field site to evaluate the likely environmental, environmental and social effects of proposed remediation strategies.  The fine scale models and pedo-transfer functions will, at this scale, be used to provide factorial corrections to incorporate the effects of detailed management strategies, soil conditions, and vegetation structural features.

WP5.4 Evaluate model outputs with stakeholders
Model outputs and field trial results will be fed back to stakeholders in each site through focus group meetings with policy-makers and extensionists, to evaluate and short-list strategies. Building on the relationships developed through work with stakeholders in previous WBs, these focus groups will be facilitated by in-country staff employed on the project. In this way, we ensure sensitivity to local context, and avoid language and cultural barriers to effective communication (see section B.4.4). Model outputs will be developed into storyboards illustrated by GIS and/or digitally manipulated photographs to communicate them effectively to a range of stakeholders. The evaluation process will be structured using a Multi-Criteria Evaluation decision aid tool (Yager, 1993). Using this approach, stakeholders will identify (and potentially weight) criteria against which they think strategies should be evaluated. These criteria are then used to systematically evaluate each remediation option on the basis of evidence from trials and model outputs. Outcomes from these focus groups and the methodological approach and models that were used will be disseminated to extensionists, policy-makers and researchers in WB6.

Deliverables
5.1.1    A model that integrates the main bio-physical and socio-economic processes interacting within an agroecosystem;
5.1.2    Models and algorithms for pipes, gullies, shallow mass movement integrated with the PESERA model;
5.2.1    Models and algorithms for estimating tillage and land leveling effects;
5.2.2    Models for describing vegetation effects on hydrology and erosion;
5.2.3    Algorithms for estimating soil complex characteristics including the effect of changes
5.3.1    Model outputs for the regions around each field site to identify the likely environmental, environmental and social effects of proposed remediation strategies;
5.4.1    A list of recommended remediation strategies within each region for policy-makers and extenstionists;
5.4.2    A methodological approach and modelling tool that can be used to evaluate remediation options beyond this project.
 
< Prev   Next >


 

FP6 logodisclaimer