Future land-use change scenarios for the Black Sea catchment

Future land-use change scenarios for the Black Sea catchment

ENVSCI-1334; No. of Pages 11 environmental science & policy xxx (2014) xxx–xxx Available online at www.sciencedirect.com ScienceDirect journal homep...

2MB Sizes 4 Downloads 50 Views

ENVSCI-1334; No. of Pages 11 environmental science & policy xxx (2014) xxx–xxx

Available online at www.sciencedirect.com

ScienceDirect journal homepage: www.elsevier.com/locate/envsci

Future land-use change scenarios for the Black Sea catchment E. Mancosu a,1,*, A. Gago-Silva b,1, A. Barbosa a,1, A. de Bono b,c, E. Ivanov a, A. Lehmann b, J. Fons d a

European Topic Centre–Spatial Information and Analysis, University of Malaga, edificio CAITI, Campus de Teatinos, 29071, Malaga, Spain b University of Geneva, Institute for Environmental Sciences, Forel Institute, enviroSPACE Lab., Battelle–Building D, 7 route de Drize, CH-1227 Carouge, Switzerland c United Nations Environment Programme, Global Resource Information Database–Geneva, Chaˆtelaine, Switzerland d European Topic Centre–Spatial Information and Analysis, Department of Geography, Autonomous University of Barcelona, 08193, Spain

article info

abstract

Article history:

Plausible future scenarios have been created for the Black Sea catchment, focussing on

Received 15 November 2013

spatially explicit alternatives for land-use changes. Four qualitative storylines (HOT, ALONE,

Received in revised form

COOP and COOL) were first developed, based on interpretation of the respective global

22 February 2014

scenarios (A1, A2, B1 and B2) produced by the Intergovernmental Panel on Climate Change.

Accepted 25 February 2014

Quantitative statistical downscaling techniques were then used to disaggregate the outputs of

Available online xxx

global scenarios at a regional level. The resulting land-use maps were spatially allocated at 1 km resolution in the Metronamica model, using a set of factors related to the identified drivers

Keywords:

of change. The land-use change model was calibrated on historical trends of land-cover

Downscaling

change (MODIS 2001 and 2008) translated into spatial allocation rules, and future land-use

IPCC

projections (IMAGE, 2001) were adopted. Suitability and constraint maps and population

land-use Change

trends were used to regulate the modelling process. The calibrated model was validated by

Metronamica

statistical procedures, visual evaluation and stakeholder involvement in order to ensure its

Scenario development

plausibility and accuracy. This methodology bridged the gap between the global and regional scales. Four simulated future states were produced for the main land-use classes–forest, grassland, cropland and built-up areas, as well as scrublands, crops/natural vegetation and barren land–for 2025 and 2050. The results suggest that the features highlighted in these scenarios are guided by global trends, such as population rise and decreasing agriculture, but with different growth rates and a variety of spatial patterns, with regional variations resulting from local backgrounds and policy objectives. This study aims to provide future land-use data as a potential geographical tool to assist policy makers in addressing environmental emergencies such as water stress and pollution. In particular, the exploration of plausible futures can support future assessments to comply with the EU Water Framework Directive and Integrated Coastal Zone Management policies around the Black Sea. # 2014 Elsevier Ltd. All rights reserved.

* Corresponding author: Tel.: +0034 951952908. E-mail address: [email protected] (E. Mancosu). 1

These authors contributed equally to this article. http://dx.doi.org/10.1016/j.envsci.2014.02.008 1462-9011/# 2014 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Mancosu, E., et al., Future land-use change scenarios for the Black Sea catchment. Environ. Sci. Policy (2014), http://dx.doi.org/10.1016/j.envsci.2014.02.008

ENVSCI-1334; No. of Pages 11

2

environmental science & policy xxx (2014) xxx–xxx

1.

Introduction

With expected global environmental change as the main driver of uncertain futures, scenario exploration has become an essential tool used in regional policy support discussions, with results such as the IPCC scenarios on global climate change (Nakicenovic et al., 2000) and the Millennium Ecosystem Assessment scenarios (Carpenter et al., 2006). These global scenarios have been shown to be plausible by many follow-up regional analyses and new scientific evidence on global warming (Rahmstorf et al., 2007), regional climate patterns and ecosystem change. However, global scenarios have remained of limited applicability at local and regional levels due to the inadequacy of scale (Kumar et al., 2006) and data input limitations. There are inherent difficulties in understanding the impact of global change drivers at local level because of specific bio-physical and socio-political factors acting locally. Therefore, spatially explicit scenario development using wellunderstood local and regional factors is necessary for adequate policy support. Such scenario-building methodologies have been explored within the enviroGRIDS project for the Black Sea catchment (BSc) (Lehmann et al., 2013). Scenario construction and analysis aims at exploring the range of uncertainties related to a future state of a system (Mahmoud et al., 2009) that cannot be well characterized by either probabilistic or deterministic predictions. In environmental science, this method builds on the systematic assimilation of multi-disciplinary quantitative and qualitative data. A first step consists of exploring what can be characterized to some degree by relations and predictions, and then identifying and describing what is left as uncertainty. This provides the added value of identifying unforeseeable points that may change the future state of the system. Environmental scenarios are integrated scenarios, as they need to combine a range of themes and subjects. Scenario themes are typically suggested by the cause and effect relationships between the most critical and most uncertain variables. The different themes are linked in a coherent narrative. Several unique storylines are typically proposed, each of which has its own likelihood of occurring. Environmental scenario construction and analysis has not yet developed into a standardized method of research. Several scenario families have been created in the past, e.g. the IPCC greenhouse gas emission scenarios (Nakicenovic et al., 2000), the Millennium Ecosystem Assessment scenarios (MEA, 2005a), the Global Environment Outlook (UNEP, 2012) and the Great Transition scenarios (Raskin et al., 2012). They all present quite different pictures and working methodologies; however, certain common elements can be identified. Recently a formal framework for scenario definition for environmental decision making was published by an international team (Mahmoud et al., 2009). They outline the following major points of scenarios that aim at characterizing future environmental factors and conditions. These consist of threats to natural ecosystems and socio-ecological systems, and have consequences for landuse. The key issues include:  Water resources – water’s importance for human survival, ecosystem management, economic activities, agriculture,

power generation, etc.; the quantity and quality of water are equally important in assessing present and future demands for the resource;  Land-use – issues related to food security, carbon cycling and land-management practices;  Technology – technological changes that affect societal development, economic growth and environmental conservation. The BSc includes parts of 24 countries in different biogeographical and socio-political situations, and encompasses considerable ethnic, socio-economic, cultural, administrative and political diversity. Its climate varies from alpine through continental and arid steppe to humid temperate forest and warm Mediterranean. Rapid coastal development for tourism, industry and transport has accelerated over the last decade. Part of the BSc has also seen a considerable decline in economic activity and population following the collapse of socialist-planned economic systems. In all this diversity of factors and drivers of change, there is one common element – the Black Sea and its coastal influence. The objective of enviroGRIDS is to present large-scale climatic, demographic and land-use scenarios of change in the BSc, starting with key long-term processes of change on land, such as agricultural practices and urban/residential transformations. The modelling process explores examples of such situations using widely available global datasets, followed by presentations and discussions with local experts and refinement of the original results. The aim is to develop qualitative storylines derived from interpretation of the global scenarios and to downscale the globally modelled estimates to simulate disaggregated changes at administrative levels and at 1 km resolution to support evaluation of the impact of land-use change on water resource distribution.

2.

Methods

The development of land-use change scenarios is a widely accepted method for anticipating future trends and supplying tools to enable policy and decision makers to develop sustainable strategies (Nakicenovic et al., 2000). Scenarios are plausible views of the future based on ‘if and then’ assertions – if the specified conditions are met, then future land-use and land cover will be realised in a particular way (Alcamo et al., 2000). The land-use scenarios were implemented in Metronamica (RIKS, 2011a), a Cellular Automaton (CA) modelling system. The first step was to quantify the qualitative storylines by using IPCC descriptions, IMAGE data and partner input. The land-use classes to be used were selected and aggregated and their behaviour modelled to meet the software requirements. Next the key element of the dynamic CA model was defined. Its dynamic behaviour arises from a given land-use cell’s response to its current suitability, zoning and accessibility parameters and the effect of neighbouring cells. The combined influence of these factors generates a transition potential, which is updated in each yearly time step. Finally, the model was calibrated and the suitability and neighbourhood rules

Please cite this article in press as: Mancosu, E., et al., Future land-use change scenarios for the Black Sea catchment. Environ. Sci. Policy (2014), http://dx.doi.org/10.1016/j.envsci.2014.02.008

ENVSCI-1334; No. of Pages 11 environmental science & policy xxx (2014) xxx–xxx

were adjusted to deliver plausible results. The resulting scenarios were created with updated suitability parameters and modified neighbourhood rules according to the scenario storyline.

3.

EnviroGRIDS Storylines

Scenario studies start by quantifying qualitative storylines, which are interpretations of factors related to global scenarios. The enviroGRIDS scenarios are based on those developed by the IPCC (Nakicenovic et al., 2000). They represent four global socio-economic development pathways (Figure 1): the vertical axis has more economically-oriented scenarios at the top and scenarios focussing more on environmental policy at the bottom; the horizontal axis has an emphasis on global policy to the left and regional development policies to the right. Development of the enviroGRIDS scenarios was also partly based on other global studies that have enriched its storylines, such as World Water Vision (Cosgrove and Rijsberman, 2000), the Global Scenario Group (Kemp-Benedict et al., 2002), Four Energy Futures (Bollen et al., 2004) and PRELUDE (EEA, 2007). These scenario studies were then investigated in greater depth (Ivanov and Barbosa, 2010; Barbosa et al., 2011), with qualitative assessment, identification of the main driving forces and presentation of storylines. The enviroGRIDS storylines derived from these studies were then discussed and approved by groups of experts. The four storylines developed for the enviroGRIDS project were given appropriate names in consultation with the stakeholders. The HOT storyline corresponds to the IPCC’s A1FI scenario (emphasis on fossil-fuel – Fossil Intensive), with rapid economic development and free-market policies, in

3

which environmental issues are not the main concern. Strong international cooperation in which environmental concerns are taken seriously (the Kyoto Protocol targets are achieved through such cooperation) is the basis of the COOP scenario, which corresponds to the B1 climate scenarios. In contrast, the ALONE and COOL storylines correspond to the A2 and B2 regionally oriented scenarios, respectively; both preserve local identities, with the possible break-up of the EU and reinforced national military capabilities. In the COOL scenario, economic growth is contained by local environmental policies, with local bodies implementing strategies to promote local sustainable development. In the ALONE scenario, however, regions are very competitive and pressure on the environment is high.

4.

Storyline Quantification

The quantification of land-use scenarios is based on the outputs of the Integrated Model to Assess the Global Environment (IMAGE, 2001) and on projections based on data from Eurostat (Statistical Office of the European Communities) and the UN World Population and Urbanization Prospects (DESA, 2010; DESA, 2011). The European regional projections (forest, grassland, urban and built-up, and cropland) were disaggregated at the level of smaller administrative units (Nomenclature of Units for Territorial Statistics, level 2 – NUTS2) and then used as input to the regional/local land allocation model (Metronamica) for 214 regions in the BSc. The disaggregation method was based on the assumption that all NUTS2 regions have the same growth rate as the Economic Region they belong to. The downscaled indexes had to be adapted because the global data and the regional/local

Figure 1 – Storylines of driving forces for enviroGRIDS scenarios. Please cite this article in press as: Mancosu, E., et al., Future land-use change scenarios for the Black Sea catchment. Environ. Sci. Policy (2014), http://dx.doi.org/10.1016/j.envsci.2014.02.008

ENVSCI-1334; No. of Pages 11

4

environmental science & policy xxx (2014) xxx–xxx

Table 1 – Urban area demand projections according to the enviroGRIDS storylines. Storyline

Urban area projections

EnviroGRIDS scenario

Population growth

EnviroGRIDS scenario characteristics

Density scenario

UN population variant

HOT

low

low

low

ALONE

highest

medium

high

COOP

low

medium

medium

COOL

medium

Increase of urban areas with higher demand in densely populated regions. Urban areas strongly attract agricultural areas around existing settlements Urban areas and demand will increase. Dispersed urban sprawl, with new settlements expected in tourist areas. Inertia in existing urban areas with strong neighbour interaction gradually decreasing with distance – expansion of existing small settlements Urban areas will not increase; however, population density will increase Urban areas will gradually increase and the inertia of current urban areas will continue with small changes

medium

medium

demand were of different orders of magnitude. The final NUTS2 quantities of change were adjusted according to the judgement of regional experts in land-use dynamics from the enviroGRIDS consortium (e.g. for unreal growth or decline rates of agriculture and forest, or the disappearance of shrubland and barren/sparse vegetation in specific regions). The 214 regions were also assigned to three country groups: FSU (former Soviet Union), OECD WEU (OECD members in Western Europe, including Turkey) and REF EEU (economic reform countries in Eastern Europe). Land-use projections for grassland, forest and cropland originated from adapted IMAGE projections. These outputs contain spatially explicit descriptions of global land-use dynamics at 0.5  0.5 arc degree resolution for 17 regions of the world, covering the period from 1970 to 2100. Inputs to the land-use model are the land-use cover from the previous time step, the demand for food, feed, biofuel crops and timber products, and potential vegetation (Alcamo et al., 2000; IMAGE, 2001). Forest is expected to increase by a small percentage in FSU regions in all scenarios while OECD WEU regions show the opposite trend. In the first time step (2008–2025) the REF EEU regions show an increment in all scenarios. In the second time step (2025–2050) the trend is reversed, with a large decrease in A1 and a smaller decrease in B1, while A2 and B2 remain stable. Cropland is expected to decrease in all scenarios during the first time step (2008–2025) with the exception of FSU regions, and to increase slightly in the second (2015–2050), reaching values below the total area in 2008. Only in the A2 scenario is an increase expected in agricultural areas. Grassland tends to remain invariable, although in the OECD WEU regions it will increase by 2025 and then decline by 2050. Due to a limitation of IMAGE projections for urban and built-up areas, which were assumed not to change in scenarios (MNP, 2006), such areas were quantified using data from the UN World Population and Urbanization Prospects (DESA, 2010; DESA, 2011), Eurostat (2010) and national statistical offices. Projections were based on standard UN scenarios, using three standard fertility assumptions: high, medium and low. Population data were downscaled from national to regional (NUTS2) level (de Bono et al., 2011). Future urban areas were estimated by multiplying urban population data by density trends. Density trends are based on historical Corine land cover changes between 2000 and 2006 combined

with future trends in urban densities mainly derived from previous studies (EEA, 2006; Angel et al., 2010a; Angel et al., 2010b). The estimated densities were high, medium and low, with an annual increase of 0.8%, no change and an annual decrease of 1%, respectively. To calculate the urban cell demands for the enviroGRIDS storylines, they were combined with the three assumptions used for population projections (Table 1). Several further adjustments were made, taking account of the regional differences among the three country groups (Figure 2).

5.

Allocation

In Metronamica, the spatial allocation of each land-use cell is determined by different local and regional variables (RIKS, 2011a), creating a model that can deal with the interaction between different dynamics (from cell size to regional demands) for all land-uses simultaneously, for decisionmaking purposes. Cellular Automaton-based models have been implemented in a number of land-use change studies (Verburg et al., 2006b; Verburg et al., 2006a; Barredo et al., 2003; van Delden et al., 2007; Veldkamp and Fresco, 1996) and can be extremely useful for policy support, even though the dynamics between the different land-uses can lead to uncertainties in the model (Verburg et al., 2006b). Metronamica applies interactions between the cells according to their category (function, vacant or feature). The function category includes classes that change as a result of regional and local dynamics; the ones selected in this study were forest, grassland, croplands, and urban and built-up. The vacant category includes classes that are available for further growth of the function category; the ones considered in this study were crop/natural vegetation, shrubland, and barren or sparsely vegetated. The feature category represents non-modelled classes (snow and ice, permanent wetlands and water). Specific land-use allocation rules were created and applied in the model for each scenario, based on the analysis of the observed changes between 2001 and 2008, and on the scenario descriptions and IMAGE 2.2 outputs. Changes observed in IMAGE were divided by economic level of development; consequently the BSc was divided into three economic regions (FSU, OECD WEU and REF EEU). This division provided a better

Please cite this article in press as: Mancosu, E., et al., Future land-use change scenarios for the Black Sea catchment. Environ. Sci. Policy (2014), http://dx.doi.org/10.1016/j.envsci.2014.02.008

ENVSCI-1334; No. of Pages 11 environmental science & policy xxx (2014) xxx–xxx

5

Figure 2 – Urban Population adjustments by group of countries (FSU, OECD WEU and REF EEU).

understanding of the changes within each IPCC scenario. Since there was no direct correlation between the MODIS and the IMAGE 2.2 land-use changes for the observation period, the enviroGRIDS partners played a decisive role in analysing and adjusting each scenario in the BSc. Finally, the land-use demand for each scenario was disaggregated at NUTS2 level. All the data used are listed in Table 2. A suitability map for each land-use type represents the probability of future change, and can be created through logistic regression or multi-criteria methods (Pontius and Schneider, 2001). Land-use suitability is divided into physical and institutional facets (RIKS, 2011a; Van Delden et al., 2007). Physical suitability describes how suited a cell is to a specific land-use type, based on the overlay of different variables, e.g. topographic and climatic variables. Institutional suitability (zoning) refers to the suitability of cells or areas at different scales (local, regional or national) that reflect current and future decision making (RIKS, 2011a, Van Delden et al., 2007), e.g. protection of land for agricultural purposes. The physical suitability of each land-use was first assessed by analysing its frequency distribution on topographic, climatic and biophysical variables. Variables with a statistically significant impact on the occurrence of land uses were selected, for instance where certain land-use types were restricted to a particular range of elevation and climate values. The variables used in this study were elevation, slope, soil quality (Fischer et al., 2008), mean annual temperature and annual precipitation (from WorldClim, Hijmans et al., 2005). A preliminary set of suitability maps was created, using the frequency distributions of the changes observed between 2001 and 2008 and calculating a land suitability criterion for each variable (Barbosa et al., 2011). The initial results obtained with the calibration showed that not all feature changes were being correctly modelled. Therefore, in addition to the average climatic variables, the lowest temperature of the coldest month and the highest temperature of the warmest month combined with the

precipitation of the warmest quarter were added to the model to restrict the suitability of specific land-use types. Initial runs of the calibration model showed a lack of control over the allocation of vacant land-use types (shrubland, natural vegetation), due to the lack of variables that could account for the changes occurring in those land-use types. To correct this problem buffer areas two cells wide were added around cells of those classes existing in 2001 and 2008. The assumption was that new areas of shrubland and natural vegetation would preferentially appear near existing ones. The suitability maps were updated after the calibration process with the mean annual temperature and annual precipitation projections for 2025 and 2050 from WorldClim (Ramirez-Villegas and Jarvis, 2010). The land-use change model also took account of factors related to policy directives such as zoning maps, which constrain the inherent dynamics of the system (EEA, 2007). The main purpose was to limit or stimulate particular changes or permanence in land-use by assigning a potential change level to each individual class. In protected areas (Natura2000, EEA; WDPA, UNEP/GRID-Europe) constraints were used to control urban sprawl and to induce potential afforestation. A similar procedure was followed with fire events and flood risk, based on the 2009 Global Assessment Report on Disaster Risk Reduction (UNISDR, 2009). Accessibility parameters were also included in the model in the form of major and secondary roads and the rail network, in vector format. Although the suitability maps were created for specific times in the past (2008) and future (2025 and 2050), the total area defining the spatial allocation of land-use cells (transition potential) was recalculated yearly from the neighbourhood rules, suitability data, zoning parameters and accessibility parameters, thereby creating dynamic transition potential maps for each year. This dynamic behaviour is an outcome of the neighbourhood relationship rules, which define relationships as a function of distance between land-use types. These rules, which are updated every time step as a result of new

Please cite this article in press as: Mancosu, E., et al., Future land-use change scenarios for the Black Sea catchment. Environ. Sci. Policy (2014), http://dx.doi.org/10.1016/j.envsci.2014.02.008

ENVSCI-1334; No. of Pages 11

6

environmental science & policy xxx (2014) xxx–xxx

Table 2 – Summary of datasets used in the allocation process. Theme

Name

Units

Description

Source

Land-Use Zoning

MODIS (MCD12Q1) World Database on Protected Areas (WDPA) Fire events

Square km Square km, several protection levels

LP DAAC UNEP

Square km, 4 levels

10 land-use classes Areas subject to restriction due to their natural importance, classes defined by IUCN and UNESCO Areas at global risk estimated by fire event density 1997–2008

Square km, 10 levels m asl

Areas at global risk of flood hazard, at 4 levels Elevation values

Number of inhabitants

Total number of residents in a NUTS 1, 2, 3 region (AL, LI, AT, BG, CH, CZ, DE, HR, IT, PL, RO, TR, MK, HU, SI, SK) Total number of residents in a region (UA) Total number of residents in a region (MD, ME, BA, RS, RU, GE) Urban and Total number of residents in a country: projections From 2008 to 2050 From 2008 to 2050 Nomenclature of main soil types is based on the terminology of the FAO legend for the Soil Map of the World

Flood risk Physical environment Demographic

Digital Elevation Model, slope GISCO

GAUL

Attractor/ Suitability

Number of inhabitants Number of inhabitants

WUP, WPP

Number of inhabitants

Mean temperature Mean precipitation Soil geographic database of Eurasia (SGDE)

Degrees/Square km mm/Square km Soil Typological Units

Distance from road network (major and secondary roads) and from rail network (rail network and rail stations)

km

cells being spatially allocated, define a cell’s inertia, or how strongly or weakly it is held in its present location, by setting a positive or negative value at distance 0; urban and built-up classes always have higher inertia than other land-use types. At a distance (>1 cell) the attraction or repulsion exerted by neighbouring cells of the same or different class is defined by assigning a positive or negative value, which may be constant or vary with distance. Although most land-use types show attraction over short distances, beyond a radius of eight cells they have no neighbourhood interaction and their change will only depend upon their suitability and zoning parameters. The values assigned for inertia and attraction/repulsion between actively modelled land-use classes were based on the behaviour observed in the calibration process and were then adjusted until a good overall fit was achieved for the entire catchment.

6.

Results

6.1.

Validation

The calibration results were assessed to measure the goodness of fit between the observed and the simulated land-use maps.

Global Assessment Report on Risk Reduction (GAR), UNEP/GRID-Europe

Shuttle Radar Topography Mission (SRTM, 90 m resolution) Eurostat, GISCO, Edition 2006 Regional Centre for Integrated Environmental Monitoring and Ecological Studies of Odessa I.I. Mechnikov National University

FAO, GAUL 2008 admin level1

UN Population Division: World Urbanization Prospects and World Population Prospects WorldClim – Global Climate Data WorldClim – Global Climate Data European Soil Information System (EUSIS), Joint Research Centre

Global Road Data, 2006; ESRI Data & Maps, 2006

During the calibration process two reference land-use maps were used for parameterization of the model, one for the start year (2001) and one for the end year (2008). Ideally, the observed land-use map used for calibration should not be used for validation (Pontius and Schneider, 2001). In this study, since an independent land-use map was not available, the 2008 reference map was used for both calibration and validation, and the problem of overfitting of the model was borne in mind. This problem typically occurs when calibration is able to reproduce the features included in the land-use map but might fail to represent general features when applied to the future (Soares et al., 2013). The Kappa statistic (Cohen, 1960) Klocation (Pontius, 2000) and Fuzzy Kappa (Hagen-Zanke, 2003) were used to quantify the mismatch between the reference and simulated land-use maps. Kappa has been widely used in land-use modelling (Pontius and Schneider, 2001, Van Vliet et al., 2011) to estimate the proportion of agreement between observed and simulated land-use maps. However, Kappa-based indices of agreement have been criticized (Pontius and Millones, 2008). To overcome the rigidity of the cell-by-cell comparison of two maps using Kappa, Hagen-Zanker (2003) proposed the Fuzzy Kappa method, which incorporates a comparison of neighbouring cells when assessing the accuracy of thematic maps, taking

Please cite this article in press as: Mancosu, E., et al., Future land-use change scenarios for the Black Sea catchment. Environ. Sci. Policy (2014), http://dx.doi.org/10.1016/j.envsci.2014.02.008

ENVSCI-1334; No. of Pages 11 environmental science & policy xxx (2014) xxx–xxx

location fuzziness into account for near cell-by-cell agreement. Kappa, Klocation and Fuzzy Kappa were implemented with the Map Comparison Kit (MCK) (RIKS, 2011b), available from the Research Institute for Knowledge Systems Website (RIKS, 2011a). During calibration, the model parameters were repeatedly adjusted until the goodness of fit was optimized. Values for Kappa range between 1, if agreement between the maps is perfect, and 0 if the observed proportion correctness is no more than would be expected by chance (Pontius, 2000). According to the Kappa classification proposed by Landis and Koch (1977), the model results (Kappa = 0.72, and Klocation = 0.73) indicate reasonable overall agreement between the observed and simulated maps, not only in quantity but also in the ability to spatially allocate land-use in the correct position. Kappa was also estimated for each landuse category. The urban and built-up classes had the highest Kappa values (0.96, and 0.88, respectively) between the simulated and real maps. The Kappa values obtained for forest and cropland showed an acceptable level of agreement (between 0.80 and 0.60). Grassland was the class with the lowest Kappa value (0.63); this may be due to the aggregation of classes made and the use of weaker driving forces for grassland. Natural vegetation and shrubland showed moderate agreement between the real and simulated maps (0.48 and 0.58, respectively). These classes displayed unstable behaviour as they fluctuated annually; therefore they were difficult to model and predict. Fuzzy Kappa results represent degrees of similarity between maps, with values from 0 to 1: the greater the overall similarity, the higher the value. The Fuzzy Kappa value obtained, calculated as the mean for all cells in the map, was 0.68, which indicates an acceptable level of overall agreement between the calibration result and reality.

7.

Future Scenarios

Once the calibrated model had been satisfactorily validated, land-use maps covering the entire BSc at 1 km2 resolution were produced for four explicit scenarios at two time steps (2025 and 2050). The model mapped the effects of regional and local driving forces according to the storyline descriptions. The HOT (A1) scenario expects the highest economic growth, with low population increase and high environmental pressure. An expansion of urban and built-up areas to the detriment of other land-use classes directly reflects this. This increase is mostly located in the REF EEU and FSU regions, whereas, WEU countries show less expansion than in other scenarios. The growth of new urban areas is mainly concentrated around existing cities while new agglomerations appear in rural areas, replacing agricultural land. While in the first phase (until 2025) there is a widespread increase in forest growth in all regions, in the second step (until 2050) considerable deforestation affects all parts of the BSc. Characterized by lower levels of trade and a regionally oriented economic policy, the ALONE (A2) scenario shows strong competition between agriculture and urban areas. Indeed a high rate of urban growth can be seen throughout the BSc; the OECD WEU regions have a higher growth rate than in

7

the HOT (A1) scenario. The global macro-economic situation and the high protection level for European agriculture increase the demand for land for agriculture throughout the BSc after 2025, just as urban sprawl took precedence over other classes before this date. In the first phase the forest class shows a positive trend in the REF EEU regions, while forest area shrinks in the other regions. The COOP (B1) scenario features high economic growth and low population increase. This is reflected by moderate growth of urban area, with a higher rate in Western countries due to a stricter spatial policy. In general urban areas expand in a very compact fashion. Strong emphasis is placed on the implementation of global mitigation and adaptation policies to cut greenhouse gas emissions and reduce the impact of climate change. Therefore, the afforestation process is strongly supported and consequently agricultural areas tend to decline, mainly in areas less physically suitable for agriculture. Some variation in this behaviour occurs in the OECD WEU regions, where the changes are less severe. The combination of land abandonment and the focus on safeguarding the environment means that the main feature of the COOP scenario is its very high percentage of forest area and natural spaces. Unlike the others, the COOL (B2) scenario may be described as ‘business as usual’. It is characterized by gradual changes and less extreme development (Nakicenovic et al., 2000). There is moderate economic and population growth, and consequently urbanization is very low and variations in agricultural and forest area are negligible. The changes are concentrated more in the eastern part of the BSc, especially in the FSU regions, where conversion from cropland to urban is higher, and where grassland and forest show significant growth. The differences between the scenarios were compared in terms of the main land-use changes to provide an overview of landscape evolution. As with global scenarios, all the enviroGRIDS scenarios show a constant expansion of urban and built-up areas, albeit with contractions in some regions. This trend is directly correlated with the assumption of increased prosperity and strong economic growth, spurring the demand for new urban areas (e.g. for second homes or the expansion of tourism facilities). Even in areas under low population pressure, a variety of factors still drive urban sprawl. These are rooted in the desire to build new suburban environments outside the inner city (EEA, 2006). In the resulting maps the environmentally-oriented scenarios (B) show expansion of urban areas – by 11% and 9% for COOL and COOP, respectively – but much less than the 27% observed in the economically-oriented scenarios (A). Urban growth influences changes in the other classes in the final HOT and ALONE maps; as these scenarios are economically-oriented, forest and natural areas shrink by between 6% (HOT) and 10% (ALONE). In COOP and COOL, in which the Kyoto Protocol targets have been met, forest conservation and expansion are expected, although urban growth is still proportionally the largest change. Forests benefit from increased conversion from land abandonment (11% in COOP) and greater protection of natural areas (in both COOP and COOL). In the ALONE scenario it is assumed that the emergence of new agricultural areas exerts strong pressure on forest and natural areas. Note that this study only considers conversions between the aggregated classes used; other conversions such as changes

Please cite this article in press as: Mancosu, E., et al., Future land-use change scenarios for the Black Sea catchment. Environ. Sci. Policy (2014), http://dx.doi.org/10.1016/j.envsci.2014.02.008

ENVSCI-1334; No. of Pages 11

8

environmental science & policy xxx (2014) xxx–xxx

Figure 3 – Comparison of pattern changes between scenarios and the base year (2008) in the Mures river sub-catchment between Romania and Hungary.

from one type of crop to another or in the type of built-up area (e.g. industrial to residential) were not taken into account and do not appear in our statistics.

8.

Discussion

The enviroGRIDS scenarios provide four diverse visions of plausible futures characterized by regionalized land-use patterns. Analysis of a sub-catchment in Central Europe (Figure 3) shows how differently change occurs in the four cases in both amount and spatial pattern from baseline landuse (2008) towards the future scenarios (2050). In the economically-oriented scenarios, HOT and ALONE, urban and built-up areas grow more than other land-use classes. Conversely, as the COOP and COOL scenarios are more focussed on environmental protection, they generate a pattern strongly marked by forest and natural vegetation.

The ALONE and COOL scenarios preserve agricultural areas in 2050. In the first case, this is due to a policy that safeguards agriculture, which implies an increase in demand for agricultural areas, and in the second case, it is linked to the limited changes derived from global trends. The resulting scenarios for the entire BSc include three distinct groups of regions (Figure 4). The FSU group occupies half of the BSc, while the REF EEU group is slightly larger than the OECD WEU group. The observed expansion of urban area – from 1.7% of total area in 2008 to 1.9% in COOP and COOL scenarios and to 2.5% in HOT and ALONE – is greatest in the REF EEU regions and least in the FSU regions. On the other hand, the REF EUU and FSU regions both show the highest percentage of agricultural land abandonment, especially in the HOT and COOP scenarios where agriculture falls from 45% to 37% of total area. Deforestation, characteristic of the HOT and ALONE scenarios, mostly affects the REF EEU and FSU regions due to their weaker protection plans. In the OECD WEU

Please cite this article in press as: Mancosu, E., et al., Future land-use change scenarios for the Black Sea catchment. Environ. Sci. Policy (2014), http://dx.doi.org/10.1016/j.envsci.2014.02.008

ENVSCI-1334; No. of Pages 11 environmental science & policy xxx (2014) xxx–xxx

9

scenarios is not recommended, due to the fact that they were derived mostly from global storylines. However, other forms of implementation, such as landscape change impact, are strongly encouraged since this research incorporates local zoning measures such as natural protected area plans (Natura 2000, WDPA) and natural hazards (flood risk and fire events). Land-use changes are considered to be the main drivers of change in the dynamics of ecosystem functions and in landscape patterns across Europe (EU, 2013). The major reduction in forest and natural vegetation areas with the expansion of arable land, as occurs in the economicallyoriented scenarios, could have unfavourable effects on the structure and functionality of ecosystem services (MEA, 2005b). The strategic importance of coordinated planning for the BSc has already been shown in other similar studies (Erickson, 1995). The main reason for preparing these scenarios is that they can be used in the Soil and Water Assessment Tool (Gassman et al., 2007) to evaluate future water resources through impact assessments (Rouholahnejad et al., 2013). This will enhance the importance of this study and enable it to address related problems such as water scarcity and drought, flood risk, and wetland management, which have to be assessed under the European Union’s Water Framework Directive (WFD) and Integrated Coastal Zone Management (ICZM) policies.

9.

Figure 4 – Land-use comparison between the resulting scenarios and the base year (2008) in percentages for the entire BSc and for the three groups of regions.

regions in the economically-oriented scenarios, grassland and natural vegetation show the largest decrease in area. The challenge of defining land-use scenarios through downscaling techniques, as underlined in other similar studies (Rounsevell and Metzger, 2010; Verburg et al., 2006b; Verburg and Overmars, 2009), entails conceptual and technical difficulties. These difficulties are related to qualitative interpretation and also to the level of detail and the precision or availability of spatial information. The model’s limited capacity of automated reasoning to simulate natural phenomena (important steps in the model require user input, e.g. neighbourhood rules and physical suitability) and computing limitations (on the number of actively modelled land-use classes) still represent a challenge in land-use modelling and simulation. However, scenarios are a useful tool for stakeholders to explore the implications of alternative forms of land development. They raise awareness and encourage creative and imaginative thinking about more distant futures. Although future changes will depend on many factors, these scenarios can help in planning more robust and flexible policy measures to deal with the future (Rounsevell and Metzger, 2010). Implementation of a local impact plan based on these

Conclusion

The general aim of the enviroGRIDS project is to provide the basis for more sustainable development in the BSc. In this study the interactions between observed land-use classes were analysed and projected to provide four different plausible scenarios. The method used involved deriving narratives for the different storylines, and quantifying them through a downscaling process from global to local level. The land-use demands were implemented in Metronamica, a valid multi-site allocation model, which has already been applied in several projects (e.g. LUMOCAP, DG Research; MOLAND, JRC; PRELUDE, EEA; and ET2050, ESPON). The calibration of the model was the most time-consuming phase, requiring close attention to ensure consistent and useful input, together with a control check by regional experts in land-use dynamics. Historical data and regional differences, physical and political, were used to build a coherent model. Throughout the project’s lifespan the enviroGRIDS partners have provided suggestions and key information for the quantification of the storylines and the calibration phase. Given the extent of the study area, however, broader stakeholder involvement could have provided more agreement on driving forces and refined the resulting change trends. The successful statistical validation of the model and its approval by regional experts guarantee the quality of the explicit land-use scenarios produced. This is the first time this methodology has been applied in the BSc and it therefore provides an opportunity to promote scenariobased policy discussions on the evolution of the landscape, for instance with a view of identifying priority areas for biodiversity or ecosystem conservation, assessing water resource vulnerability, or addressing many other policyrelevant environmental issues.

Please cite this article in press as: Mancosu, E., et al., Future land-use change scenarios for the Black Sea catchment. Environ. Sci. Policy (2014), http://dx.doi.org/10.1016/j.envsci.2014.02.008

ENVSCI-1334; No. of Pages 11

10

environmental science & policy xxx (2014) xxx–xxx

Conflict of interest The authors declare that there are no conflicts of interest.

Acknowledgements This research was funded by the European Commission through the Seventh Framework Programme (FP7) enviroGRIDS project (Grant Agreement No 227640).

references

Alcamo, J., Leemans, R. & Kreileman, E. 2000. Global Change Scenarios of the 21st Century: Results from the IMAGE 2.1 Model, Pergamon. Angel, S., Parent, J., Civco, D., 2010a. The Fragmentation of Urban Footprints: Global Evidence of Sprawl, 1990-2000. Lincoln Institute of Land Policy, Cambridge, USA. Angel, S., Parent, J., Civco, D., Blei, A., 2010b. The Persistent Decline in Urban Densities: Global and Historical Evidence of ‘Sprawl’. Lincoln Institute of Land Policy, Cambridge, USA. Barbosa, A., Mancosu, E., Gago da Silva, A., de Bono, A., Lehmann, A., Mierla, M., Hanganu, J. & Littkopf, A. 2011. Proposed land use scenario analysis, model input parameters and allocation rules [Online]. EnviroGRIDS FP7 European project. Available: http://envirogrids.net/ index.php?option=com_content&view=category& layout=blog&id=4&Itemid=17 [Accessed October 2013]. Barredo, J.I., Kasanko, M., Mccormick, N., Lavalle, C., 2003. Modelling dynamic spatial processes: simulation of urban future scenarios through cellular automata. Landscape and Urban Planning 64, 145–160. Bollen, J.C., Manders, A.J.G., Mulder, M., 2004. Four Futures for Energy Markets and Climate Change. CPB Netherlands Bureau for Economic Policy Analysis, National Institute for Public Health and the Environment, The Hague. Carpenter, S.R., Bennett, E.M. & Peterson, G.D. 2006. Editorial: Special Feature on Scenarios for Ecosystem Services. Ecology and Society, 11. Cohen, J., 1960. A coefficient of agreement for nominal scales. Educational and Psychological Measurement 20, 37–46. Cosgrove, W.J. & Rijsberman, F.R. 2000. World Water Vision: Making Water Everybody’s Business. London. de Bono, A., Dao, H., Silva, A., Barbosa, A. & Mancosu, E. 2011. Proposed demographic scenario analysis and overview of driving forces and justification, model input parameters and allocation rules [Online]. EnviroGRIDS FP7 European project. Available: http://envirogrids.net/index.php?option=com_ jdownloads&Itemid=13&view=view.download&catid= 12&cid=119 [Accessed October 2013]. Desa, U. 2010. The United Nations Development Agenda: Development for All. New York. Desa, U. 2011. The United Nations Development Agenda: Development for All. EEA, 2006. Urban sprawl, the ignored challenge [Online]. European Environment Agency and Office for Official Publications of the European Communities. Available: http:// www.eea.europa.eu/publications/eea_report_2006_10 [Accessed October 2013]. EEA, 2007. Land-use scenarios for Europe: Qualitative and quantitative analysis on a European scale (PRELUDE) [Online]. European Environment Agency. Available: http://

www.eea.europa.eu/publications/technical_report_2007_9 [Accessed October 2013]. Erickson, D.L., 1995. Rural land use and land cover change, Implications for local planning in the. River Raisin watershed. Land Use Policy 12, 223–236. EU, 2013. Mapping and Assessment of Ecosystems and their Services [Online]. Available: http://ec.europa.eu/ environment/nature/knowledge/ecosystem_assessment/ pdf/MAESWorkingPaper2013.pdf [Accessed October 2013]. Fischer, G., Nachtergaele, F., Prieler, S., Velthuizen, V.A.N., Verelst, H.T., Wiberg, L.D., 2008. Global Agro-ecological Zones Assessment for Agriculture. Food and Agriculture Organization (FAO), Laxenburg, Austria and FAO, Rome, Italy. Gassman, P.W., Reyes, M.R., Green, C.H., Arnold, J.G., 2007. The soil and water assessment tool: Historical development, applications, and future research directions. Transactions of the Asabe 50, 1211–1250. Hagen-Zanker, A., 2003. Fuzzy set approach to assessing similarity of categorical maps. International Journal of Geographical Information Science 17, 235–249. Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., Jarvis, A., 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25, 1965–1978. IMAGE, 2001. T he IMAGE 2.2 implementation of the SRES scenarios: A comprehensive analysis of emissions, climate change and impacts in the 21st century. Ivanov, E. & Barbosa, A. 2010. Existing scenarios and data compilation on integrated scenarios using demographic, climatic, land cover from global and Black Sea Basin studies [Online]. EnviroGRIDS FP7 European project. Available: http://www.envirogrids.net/index.php?option=com_ jdownloads&Itemid=13&view=view.download&catid= 12&cid=100 [Accessed October 2013]. Kemp-Benedict, E., Heaps, C. & Raskin, P. 2002. Global Scenario. Group Futures: Technical Notes. PoleStar Series. Stockholm. Kumar, K.R., Sahai, A.K., Kumar, K.K., Patwardhan, S.K., Mishra, P.K., Revadekar, J.V., Kamala, K., Pant, G.B., 2006. Highresolution climate change scenarios for India for the 21st century. Current Science 90, 334–345. Landis, J., Koch, G., 1977. MEASUREMENT OF OBSERVER AGREEMENT FOR CATEGORICAL DATA. Biometrics 33, 159–174. Lehmann, A., Giuliani, G., Mancosu, E., Abbaspour, K.C., So¨zen, S., Gorgan, D., Beel, A. & Ray, N. 2013. Filling the gap between Earth observation and policy making in the Black Sea catchment with enviroGRIDS. Environment Science and Policy. Mahmoud, M., Liu, Y., Hartmann, H., Stewart, S., Wagener, T., Semmens, D., Stewart, R., Gupta, H., Dominguez, D., Dominguez, F., Hulse, D., Letcher, R., Rashleigh, B., Smith, C., Street, R., Ticehurst, J., Twery, M., Van Delden, H., Waldick, R., White, D., Winter, L., 2009. A formal framework for scenario development in support of environmental decisionmaking. Environmental Modelling & Software 24, 798–808. MEA, 2005a. Ecosystems and Human Well-Being: Our Human Planet: Summary for Decision Makers. MEA, 2005b. Ecosystems and Human Well-being: Biodiversity Synthesis. MNP, 2006. Integrated modelling of global environmental change. An overview of IMAGE 2.4. Nakicenovic, N., ALCAMO, J., DAVIS, G., DE VRIES, B., FENHANN, ¨ BLER, A., JUNG, T.Y., KRAM, J., GAFFIN, S., GREGORY, K., GRU T., LA ROVERE, E., MICHAELIS, L., MORI, S., MORITA, T., PEPPER, W., PITCHER, H., PRICE, L., RIAHI, K., ROEHRL, A., ROGNER, H.H., SANKOVSKI, A., SCHLESINGER, M., SHUKLA, P., SMITH, S., SWART, R., VAN ROOYEN, S., VICTOR, N. &

Please cite this article in press as: Mancosu, E., et al., Future land-use change scenarios for the Black Sea catchment. Environ. Sci. Policy (2014), http://dx.doi.org/10.1016/j.envsci.2014.02.008

ENVSCI-1334; No. of Pages 11 environmental science & policy xxx (2014) xxx–xxx

DADI, Z. 2000. Special Report on emissions scenarios. In: NAKICENOVIC, N.A. S., R. (ed.). Cambridge. Pontius, R. & Millones, M. 2008. Problems and solutions for Kappa-based indices of agreement. In Studying, modelling and sense making of planet earth. Mytilene, Lesvos, Greece. Pontius, R.G., 2000. Quantification error versus location error in comparison of categorical maps. Photogrammetric Engineering and Remote Sensing 66, 1011–1016. Pontius, R., Schneider, L., 2001. Land-cover change model validation by an ROC method for the Ipswich watershed, Massachusetts. USA. Agriculture Ecosystems & Environment 85, 239–248. Rahmstorf, S., Cazenave, A., Church, J.A., Hansen, J.E., Keeling, R.F., Parker, D.E., Somerville, R.C.J., 2007. Recent climate observations compared to projections. Science 316 709-709. Ramirez-Villegas, J. & Jarvis, A. 2010. Downscaling Global Circulation Model Outputs: The Delta Method Decision and Policy Analysis Working Paper No.1. Raskin, P., Banuri, T., Gallopı´n, G., Gutman, P., Hammond, A., Kates, R. & Swart, R. 2012. Great Transition.The Promise and Lure of the Times Ahead. Stockholm Environment Institute. RIKS, 2011a. Metronamica–Model descriptions. Maastricht. Research Institute for Knowledge Systems. RIKS, 2011b. MCK Reader: Methods of Map Comparison Kit. Maastricht. Research Institute for Knowledge Systems. Rouholahnejad, E., Abbaspour, K.C., Bacu, V. & Lehmann, A. 2013. A high resolution spatiotemporal distribution of water resources in the Black Sea Basin. Journal of water resources research. Rounsevell, M.D.A., Metzger, M.J., 2010. Developing qualitative scenario storylines for environmental change assessment. Wiley Interdisciplinary Reviews-Climate Change 1, 606–619. Soares, B., Rodrigues, H., Follador, M., 2013. A hybrid analyticalheuristic method for calibrating land-use change models. Environmental Modelling & Software 43, 80–87. UNEP, 2012. GEO-5 Assessment Full Report. United Nations Environment Programme. UNISDR, 2009. Global Assessment Report on Disaster Risk Reduction. United Nations International Strategy for Disaster Reduction (UNISDR), Secretariat. Van Delden, H., Luja, P., Engelen, G., 2007. Integration of multiscale dynamic spatial models of socio-economic and physical processes for river basin management. Environmental Modelling & Software 22, 223–238. Van Vliet, J., Bregt, A.K., Hagen-Zanker, A., 2011. Revisiting Kappa to account for change in the accuracy assessment of land-use change models. Ecological Modelling 222, 1367–1375. Veldkamp, A., Fresco, L.O., 1996. CLUE-CR: An integrated multiscale model to simulate land use change scenarios in Costa Rica. Ecological Modelling 91, 231–248. Verburg, P.H., Kok, K., Pontius, R.G., Veldkamp, A., 2006a. Modeling Land-Use and Land-Cover Change. In: Lambin, E., Geist, H. (Eds.), Land-Use and Land-Cover Change.. Springer Berlin Heidelberg. Verburg, P.H., Overmars, K.P., 2009. Combining top-down and bottom-up dynamics in land use modeling: exploring the future of abandoned farmlands in Europe with the DynaCLUE model. Landscape Ecology 24, 1167–1181. Verburg, P.H., Schulp, C.J.E., Witte, N., Veldkamp, A., 2006b. Downscaling of land use change scenarios to assess the dynamics of European landscapes. Agriculture Ecosystems & Environment 114, 39–56.

11

Emanuele Mancosu holds a BSc degree in Natural Science and an MSc in GIS technology. He is working for the ETCSIA, European Topic Centre on Spatial Information and Analysis, at the University of Malaga from 2011 covering GIS tasks, responsible for data management and supporting environmental research through EEA activities and FP7 EU projects as enviroGRIDS. The area of work is based on land cover monitoring, assessment, modelling and analysis at EU scale. Previously he worked for the ETCLUSI (2008–2010) holding similar roles. Contact point: European Topic Centre – Spatial Information and Analysis, University of Malaga, edificio CAITI, Campus de Teatinos, 29071, Malaga, Spain. Emanuele.ma[email protected] Phone: 0034 951 952 908. Ana Gago-Silva has received a pre-Bologna Licenciatura degree in Geomatic Engineering from the University of Lisbon and an MSc in Geographic Information Systems from University of Ulster. She is a PhD candidate in the enviroSPACE laboratory of the University of Geneva. Her research focus is on downscaling of land-use climatic and demographic datasets. She contributed to the Work Package III from the EU-FP7 enviroGRIDS project. Ana Barbosa holds a BSc degree in Geography and Planning and an MSc degree in Geographic Information System. She has 6 years of experience in applying land use modeling tools. Working at the European Topic Centre (2007–2011), she collaborated in the enviroGRIDS project, namely in the development of land use scenarios, modelling and providing technical support to the GIS activities. Since 2011, she is working as a scientific/technical officer at Joint Resource Centre – European Commission, where she has been participating in European land use modelling activities. Research Associate at University of Nottingham (UK), specialized in Ecology, GIS and remote sensing. Emil joined the Centre for Environmental Management (CEM) in April 2011 after nearly 10 years of extensive international experiences in education and work. At CEM his main subject of work is the application of the ecosystem accounting model for the countries of the Mediterranean and the Black Sea, including North Africa, Middle-east and South Europe (a work package part of PEGASO: People for Ecosystem-based Governance in Assessing Sustainable Development of Ocean and Coast (EU FP7 Collaborative Project, 2010– 2014, www.pegasoproject.eu). Dr. Andrea de Bono. MSc and PhD in Geology, specialized in Geomatics. He joined UNEP/GRID – Geneva in 2001. Currently he is responsible for data flow in the UNEP Environmental Data Explorer platform, and coordinator for the Global Exposure Database for the Global Assessment Report on Disaster Risk Reduction (UNISDR). He was involved in the integration of scenarios of demographic, climatic and land-use changes within the enviroGRIDS project. Dr Anthony Lehmann is the enviroGRIDS project initiator and coordinator. He holds a Masters Degree and a PhD in Aquatic Biology from the University of Geneva, and a Postgraduate Master in Statistics from the University of Neuchaˆtel. He specialized during his career in combining GIS analyses with statistical models. At the University of Geneva he is in charge of the enviroSPACE laboratory exploring Spatial Predictions and Analyses in Complex Environments.

Please cite this article in press as: Mancosu, E., et al., Future land-use change scenarios for the Black Sea catchment. Environ. Sci. Policy (2014), http://dx.doi.org/10.1016/j.envsci.2014.02.008