Scaling issues in multi-criteria evaluation of combinations of measures for integrated river basin management

In integrated river basin management, measures for reaching the environmental objectives can be evaluated at different scales, and according to multiple criteria of different nature (e.g. ecological, economic, social). Decision makers, including responsible authorities and stakeholders, follow different interests regarding criteria and scales. With a bottom up approach, the multi criteria assessment could produce a different outcome than with a top down approach. The first assigns 10 more power to the local community, which is a common principle of IWRM. On the other hand, the development of an overall catchment strategy could potentially make use of synergetic effects of the measures, which fulfils the cost efficiency requirement at the basin scale but compromises local interests. Within a joint research project for the 5500 km2 Werra river basin in central Germany, measures have been planned to reach 15 environmental objectives of the European Water Framework directive (WFD) regarding ecological continuity and nutrient loads. The main criteria for the evaluation of the measures were costs of implementation, reduction of nutrients, ecological benefit and social acceptance. The multi-criteria evaluation of the catchment strategies showed compensation between positive and negative performance of criteria within the catchment, which in the end reduced the discriminative power of the different strategies. Furthermore, benefit criteria are partially computed for the whole basin only. Both ecological continuity 20 and nutrient load show upstream-downstream effects in opposite direction. The principles of “polluter pays” and “overall cost efficiency” can be followed for the reduction of nutrient losses when financial compensations between upstream and downstream users are made, similar to concepts of emission trading.

2 with a high salt load of the Werra River. Like for many German rivers, the morphological conditions of the river courses and the ecological continuity were affected before implementation of the European Water Framework Directive (WFD).
Agricultural land use dominates in the North-Eastern area of the catchment. In former Eastern Germany many dispersed settlements were not connected to the public sewer system and were often not equipped with decentralized wastewater treatment. While the degree of connection was 98 % in Hessen, it was 48 % in the Thuringian part of the Werra catchment in 5 2001. As a consequence, the nutrient load of the catchment was high compared to the relatively extensive land use and the low population density. The ecological community was degraded in several water bodies, showing a good ecological status according to the AQEM assessment system (Hering et al., 2004) only in upstream regions of the Thuringian Forest.
For the implementation of the WFD, an exemplary river basin management plan (RBMP) was elaborated by an 10 interdisciplinary research team, supported by local water authorities (Dietrich and Schumann, 2006). The RBMP provided several alternative strategies for the catchment, which were prepared for a final decision procedure supported by a multi criteria decision support system (Dietrich et al. 2007). Within this paper we focus on spatial aspects of measures for the improvement of the hydro-morphological conditions and the reduction of nutrient loads from point sources and diffuse sources (for a detailed description see Dietrich and Funke, 2009). 15 One of the challenges in spatial decision analysis is the spatial aggregation of criteria. For an RBMP, measures are located throughout the catchment area. The criteria for the individual measures can be aggregated in space to get an overall multicriteria assessment of alternative combinations for the RBMP. This technique was applied in the widely used MULINO-DSS (Giupponi et al., 2002). Alternatively the multi-criteria analysis (MCA) can be applied for each of the locations separately, 20 and then the outcome of the MCA is aggregated in space. Both pathways of aggregation of criteria and space can lead to different overall results (Herwijnen and Rietveld, 1999). The first path better represents the characteristics of the basin, whereas the second path allows different preference structures for the smaller sub-units, hence better represents the local situation. By aggregating criteria, positive and negative effects can be smoothened, with the consequence of reduced distinctive character of the alternatives. This kind of spatial compensation can be addressed by introducing additional criteria 25 as Nijssen and Schumann (2014) showed for flood risk management. In this study, we present a strategic combination approach, which includes a criterion for social acceptance of the measures in order to represent the stakeholders' preference for the local measures.

Morphological state and nutrient emissions 30
The ecological assessment with AQEM showed significant deviations from the species composition, which could be expected for the types of water bodies in that catchment. The salt load of the lower Werra River was not subject of the investigations even if it was known that it is one of the causes of ecological degradation for the affected water bodies. Apart from this, morphological deficits in most river courses ( Fig. 1) were identified as a major problem to address in river basin management (Dietrich and Schumann, 2006), hence in the implementation of the Water Framework Directive. The morphological deficits include the riparian and river bed structure, but also numerous structures from groundsills to reservoir dams which disturb or prevent fish migration. Also the overall saprobial state (Fig. 1), as well as nitrate and phosphorus 5 concentrations were found to be beyond the levels which support a good ecological state according to the WFD. The quantitative investigation of the nutrient cycle was done with a chain of models, combining an agricultural production model to compute nutrient losses from agricultural areas, a point source emission model for sewage treatment, and a coupled SWAT-RWQM1 model to simulate nutrient turnover and transport at catchment scale. The emissions of nitrogen and phosphorus from point and non-point sources show an uneven distribution over the catchment, closely related with urban 10 land use in the case of point sources (Fig. 2) and agricultural land use in the case of diffuse (non-point) sources (Fig. 3).

Development of alternative environmental measures
The objective of river basin management according to the WFD is to reach a good ecological state of all water bodies by 2015, with some possible exemptions e.g. for heavily modified water bodies or due to long lasting sanitation or disproportionate costs. The WFD gives a framework for the development and 6-yearly update of river basin management 15 plans (RBMP). The RBMP collects all measures, which were decided by the respective bodies. Within the Werra project, an exemplary RBMP was developed to address the environmental issues of the catchment that were introduced in 2.1. Different from the formal and final WFD RBMP, in this paper we provide alternative solutions for the selection phase of the decision process, which means that we present not a single solution but alternative measures, which follow the same overall objective.
The following types of measures were considered and then designed for the water bodies in order to fulfil the objectives of 20 the WFD:

Multi-criteria assessment of measures 5
All measures were evaluated with the following methods and criteria (Dietrich & Schumann, 2006): The WFD formulates aspiration levels for ecological criteria. If all combinations of measures for the RBMP can reach the objectives, there is no degree of freedom, which justifies an ecological decision criterion. Nevertheless, overfulfillment of the ecological status (very good status instead of good) provides an additional value and could be formulated as criterion. Furthermore, making use of WFD exemptions reduces the ecological value of the measures, which again 10 justifies a decision criterion. In this study, all measures were planned to reach the aspiration levels only, and exemptions were negotiated separately. Thus, we do not investigate purely ecological criteria for the spatial aggregation issue.
Ecological consequences of the implementation of measures were included in the ecological benefit analysis, which is human centred and expressed in monetary units.

Combination of measures to catchment scale strategies
The final result of the project's planning are several alternative combinations of measures for the RBMP, which can be used as a decision matrix for multi-criteria evaluation and computation of a ranking based on preferences for the different criteria.
This final matrix is computed for the entire Werra catchment. The aggregation of criteria from locations (single measures) via water bodies and their contributing catchments up to the catchment scale was complex and hat to be treated differently 5 for the different criteria. For that reason, the pathway of aggregating criteria first was not possible. We decided to build combinations of measures according to different principles of strategic planning and policy making. Thus we called the final alternatives "strategies".
The aggregation of the criteria introduced in 2.3 faced the following issues: Ecology: Morphological riverbed improvement was mostly assessed local for single measures (creating or improving habitat 10 structures), but there can be additional ecological effects at larger scale by habitat connectivity. The ecological continuity is very important for long distance travelling fishes. Therefore measures are most effective from downstream to upstream, whereas single measures in the middle of the catchment have reduced value when downstream connectivity is not given.

Ecological benefit:
The TEV calculation includes components, which could not be obtained at the scale of single measures or water bodies, in particular by applying the benefit transfer from other studies (Table 1). This includes a super-additive 15 benefit for developing the whole basin into a good ecological status.
Costs were attributed to single measures and aggregated by summation. The complexity of the problem does not allow a spatial multi-criteria aggregation at smaller scales than the overall catchment. Otherwise, much more detailed studies had to be performed regarding the ecological benefit and the social 30 criteria. Furthermore, a decomposition of the upstream -downstream effects of nutrient reduction had to be done. As a consequence, we performed a coordinated catchment strategy development. This follows the following principles:

Reduction of nutrient loads
7th International Water Resources Management Conference of ICWRS, 18-20 May 2016, Bochum, Germany, IWRM2016-73-1 ST1: first reduce point sources, then diffuse sources -the idea is a lower cost and better predictability of the consequences of measures at point sources; ST2: first reduce diffuse sources, then point sources -the idea is to make use of combined beneficial effects from reducing diffuse sources by hydro-morphological structures like riparian buffers; ST3: polluter oriented distribution of measures -the idea is to strictly follow the "polluter pays" principle; 5 ST4: most cost efficient allocation of measures -the idea is an economic optimization of the overall RBMP.
All the four basic strategies were computed and all criterion values were calculated with the respective methods. For the ecological benefit, the willingness to pay for biodiversity was calculated with a declining value. The measures for ecological continuity prefer the removal of structures where possible. Table 2 shows the results of the overall assessment.

10
The polluter oriented strategy ST3 does not only show the highest costs, but also the highest conflict potential because farmers expressed negative about the planned measures ( Table 2). The optimized strategy ST4 is marginally cheaper than ST1, but shows better ecological benefit due to high valued riparian buffers. But, ecologists estimated that 13 instead of 10 resp. 11 water bodies need extended monitoring due to a marginal fulfilment of the ecological objectives, which (under uncertainty) can lead to the need for additional measures. 15

Conclusions
The results of the simulation and aggregation of criteria highlight problems in following the "polluter pays" principle and the WFD requirement of overall "cost efficiency of the program of measures" for the RBMP at the same time. A decomposition of larger scale measures and the redistribution of costs for measures with basin wide effects could be done by concepts like emission trading for nutrients. Then, the cost recovery happens at the polluters, but the spatial aggregation effects of nutrient 20 reduction can be utilized in the best way.