|
OPTIMIZATION FRAMEWORK:
RIPARIAN BUFFERS |
Motivation
Figure 1: Distance between patches of riparian trees in the Zwalm catchment
MAIN ARTICLE: Witing et al (2022). Riparian reforestation on the landscape scale – Navigating trade-offs among agricultural production, ecosystem functioning and biodiversity. Journal of Applied Ecology. DOI: 10.1111/1365-2664.14176
Riparian forests along streams and rivers provide many functions that are essential for terrestrial and aquatic ecosystems. Unfortunately, stream-riparian networks are subject to increasing demands and multiple human pressures, affecting their connectivity, driving habitat and diversity losses, threatening ecosystem services, and causing stakeholder conflicts.
Strategies to mitigate such conflicts require designing riparian landscapes in a way that they can simultaneously meet multiple competing demands. Thus, the results of CROSSLINK case studies have been used to develop an optimization framework for stream-riparian green and blue infrastructure capable of balancing different environmental and socio-economic objectives. The focus of this optimization framework was on the identification of spatial configurations that minimize trade-offs and support the multifunctionality of the case study areas.
Riparian forests along streams and rivers provide many functions that are essential for terrestrial and aquatic ecosystems. Unfortunately, stream-riparian networks are subject to increasing demands and multiple human pressures, affecting their connectivity, driving habitat and diversity losses, threatening ecosystem services, and causing stakeholder conflicts.
Strategies to mitigate such conflicts require designing riparian landscapes in a way that they can simultaneously meet multiple competing demands. Thus, the results of CROSSLINK case studies have been used to develop an optimization framework for stream-riparian green and blue infrastructure capable of balancing different environmental and socio-economic objectives. The focus of this optimization framework was on the identification of spatial configurations that minimize trade-offs and support the multifunctionality of the case study areas.
Methods & tools
The development of a preliminary framework for the optimization of landscape-level riparian green-blue-infrastructure included several methodological steps and tools. All steps that are presented in the following sections have been applied to the CROSSLINK case study of the Zwalm River in Flanders (Belgium).
Target indicators and spatial model variables
Figure 2: Examples for the subdivision of a riparian corridor for determining land-use properties of a specific local riparian segment
The results of analyses that were carried out in the CROSSLINK case studies lead to the definition of environmental and socio-economic objectives. These objectives served as target indicators for the optimization. They are specific for each case study basin and include biodiversity, functional indicators (species traits), supporting processes (e.g., litter decomposition, algal productivity) as well as socio-economic trade-offs (e.g. loss of arable land).
A variety of case study-specific physiographic and land-use parameters has been determined by conducting a comprehensive GIS analysis. The suitability of the parameters as explanatory variables for the optimization models, has been tested. The spatial parameters derived can be grouped into 3 categories: (1) local properties related to a specific river segment upstream of a sampling site (see Figure 2), (2) catchment properties related to the riparian catchment and total catchment upstream of a sampling site, (3) connectivity properties including a set of distance measures.
A variety of case study-specific physiographic and land-use parameters has been determined by conducting a comprehensive GIS analysis. The suitability of the parameters as explanatory variables for the optimization models, has been tested. The spatial parameters derived can be grouped into 3 categories: (1) local properties related to a specific river segment upstream of a sampling site (see Figure 2), (2) catchment properties related to the riparian catchment and total catchment upstream of a sampling site, (3) connectivity properties including a set of distance measures.
Model development & reforestation experiments
Linked biophysical-statistical models have been developed to quantify the influence of forested riparian buffers, land use and other human activities on the identified target indicators (each representing an objective). The data set of the CROSSLINK field campaigns has been used to identify the relationships between target indicators and spatial model parameters. A variety of site types are represented in the data set such as undisturbed reference sites, buffered (forested) and unbuffered as well as downstream sites. Since it was required to apply the models to the entire riparian area of the case studies, the riparian corridors have been divided into riparian segments. The segment-based approach is in accordance to the sampling design of the field studies and reduces computation time of the optimization tool. Each segment is characterized by a set of spatial variables for its three areas of potential influence (local, riparian, catchment).
For each riparian segment, seven spatially explicit reforestation experiments have been developed that gradually increase existing forest patches, thus representing different intensities of riparian reforestation (see constrained baseline scenario on the top of Figure 2). Here, the forest area was constrained by other land-use categories (e.g. urban) in the riparian corridor. Within the optimization framework these segment-scale scenarios were recombined to thousands of new catchment scale solutions.
For each riparian segment, seven spatially explicit reforestation experiments have been developed that gradually increase existing forest patches, thus representing different intensities of riparian reforestation (see constrained baseline scenario on the top of Figure 2). Here, the forest area was constrained by other land-use categories (e.g. urban) in the riparian corridor. Within the optimization framework these segment-scale scenarios were recombined to thousands of new catchment scale solutions.
Figure 3: Simplified overview of the workflow with the embedded CROSSLINK optimization framework
Integration into CoMOLA
The models, reforestation experiments and data were then integrated into a multi-objective optimization environment (see workflow figure) to identify synergies and trade-offs between the target indicators of a case study. The optimization was carried out using the innovative Python environment CoMOLA (Constrained Multi-objective Optimization of Land-use Allocation; Strauch et al., 2019), developed at the Helmholtz Center for Environmental Research – UFZ. In our project, CoMOLA utilizes the Non-dominated Sorting Genetic Algorithm-II (NSGA-II) to optimize (riparian) land-use maps for all target indicators simultaneously. Therefore numerous (tens of thousands) simulations of different potential spatial land use configurations are generated to explore the ‘potential solution space’ of a study site and to identify optimal solutions along a Pareto frontier.
The allocation and combination of the segment specific reforestation experiments have been optimized and the whole set of pareto-optimal solutions was analyzed to identify (1) functional trade-offs among the objectives, (2) priority regions for riparian reforestation and (3) the required reforestation intensity.
Link to CoMOLA on GitHub: github.com/michstrauch/CoMOLA
The allocation and combination of the segment specific reforestation experiments have been optimized and the whole set of pareto-optimal solutions was analyzed to identify (1) functional trade-offs among the objectives, (2) priority regions for riparian reforestation and (3) the required reforestation intensity.
Link to CoMOLA on GitHub: github.com/michstrauch/CoMOLA
Application to the riparian areas of the Zwalm river
Figure 4: Location and land-use of the Zwalm case study catchment in Flanders (Belgium).
The CROSSLINK optimization framework was applied to the riparian area of the Zwalm catchment. The riparian corridor of the Zwalm has been divided into 489 riparian segments with a length of 300 m. Forest allocation was optimized to increase EPT (Ephemeroptera, Plecoptera, Trichoptera) taxa richness, functional diversity of diatoms and terrestrial cotton strips decomposition, while minimizing losses in agricultural production potential. Important spatial model variables that have been identified include, inter alia, the share of forest, grassland and arable land in the local riparian segments, the area share and average distance between forest blocks in the upstream riparian corridor and a broad set of physiographic parameters.
EPT taxa richness, Diatom diversity and cotton strips decomposition all benefited from the expansion of riparian forests, although the optimal forest allocation differed. While optimization results showed only small functional trade-offs among the three environmental objectives, strong trade-offs could be observed between environmental objectives and potential losses in agricultural production value. Compared to a simple scenario approach, the spatial optimization of reforestation measures provided solutions where all four target indicators could be improved. The results showed that intensive reforestation in the headwater regions of the Zwalm River was often combined with a slight reduction of the forest patch size in the downstream parts of the catchment. Thus the optimization results for the Zwalm suggest to connect the existing forest patches of the upstream headwaters and expand them further downstream.
EPT taxa richness, Diatom diversity and cotton strips decomposition all benefited from the expansion of riparian forests, although the optimal forest allocation differed. While optimization results showed only small functional trade-offs among the three environmental objectives, strong trade-offs could be observed between environmental objectives and potential losses in agricultural production value. Compared to a simple scenario approach, the spatial optimization of reforestation measures provided solutions where all four target indicators could be improved. The results showed that intensive reforestation in the headwater regions of the Zwalm River was often combined with a slight reduction of the forest patch size in the downstream parts of the catchment. Thus the optimization results for the Zwalm suggest to connect the existing forest patches of the upstream headwaters and expand them further downstream.
Important notes
All information presented in this section are preliminary. The methods described are in continuous development and shall present the main principles and the progress made. This also applies to the preliminary results and their interpretation. In case you have any questions please contact Felix Witing and Prof. Martin Volk from the Helmholtz Center for Environmental Research – UFZ.