Datasets used in the prioritisation
The biodiversity data included information on 31 features (represented as raster layers) that described distributions of habitats (20 layers) or species or species groups (11 layers). The features were produced from multiple sources including national, regional, and municipal authorities and environmental NGOs. A feature was included in the prioritisation if it met the following criteria: (i) it included ecologically relevant information (e.g. distribution of a species or quality of a habitat), (ii) it covered the entire study area, (iii) it was of good quality and up to date, (iv) we were able to access metadata on the production chain of the data, (v) its resolution/scale was detailed enough for the analysis, and (vi) with other data, it constituted diverse data that would broadly represent biodiversity for the purpose of planning. See Supplementary material (Tables S1-S2) for the full list of biodiversity layers and their data providers.
Expert elicitation
Expert elicitation was required in several phases of the analysis process. The Regional Council of Uusimaa formed an expert group of around 20 representatives of environmental experts of major municipalities, SYKE: the Finnish Environment Institute, other relevant authorities and NGOs to give their opinion for different steps of the analysis. The group met with the planners and the researchers every second month. During these meeting, we first trained the group to understand the principles of SCP and use of Zonation as a tool. Later, the expert group participated through facilitated group discussions on selecting 1) the features included in the analyses, 2) weighting of species or habitats and 3) values for connectivity decay distances for different features. After the analyses were carried out, the expert group 4) provided feedback on the visualisations and 5) evaluated the outcome and suggested changes to the analysis structure and amendments to the source data.
Data pre-processing
Zonation analysis necessitates that all input data are in raster format, with the same extent and pixel size. As the biodiversity data were originally diverse and heterogeneous, including point, polygon, and raster type data, with varying values describing presence/absence or abundance of species, or the biodiversity value of habitats, each input data layer was pre-processed separately. We converted all data to raster format with a spatial resolution of 100 metres and having the same projection and extent using ArcGIS 10 software (ESRI, Redlands, CA, USA). First, we buffered point data, mainly observations of rare/endangered species, with a species-specific radius defined by the experts (see Supplementary S7). Buffered areas represent the habitats of the species in a suitable manner, especially when using condition layer (see more detailed description below) that “cuts off” or lowers the value of known unsuitable habitat areas within the buffers around the species observation sites.
We determined raster cell values in four ways, depending on the type of the original data. If the data were based on field inventories (e.g. great cormorant Phalacrocorax carbo), we used abundance-based continuous values in the raster data. If the values were based on continuous indices (e.g. forest layers that were calculated as a function of stem volume and tree age) we used them as they were. If the data were observation-type presence-only data (e.g. otter Lutra lutra), we used binary values 1 and 0. If the data included some kind of earlier classification (e.g. valuable esker habitats that had been classified as nationally, regionally, or locally important), we used hierarchical categories determined by expert decision. In addition to the cell values, Zonation considers the distribution of each input layer. Cell of a rare species weigh proportionally more compared to cells of widely occurring species.
Zonation analyses
We produced a set of prioritisations with Zonation, which allowed planners to assess the importance of the same areas from different ecological perspectives. The 'basic' analysis included all the species and habitat layers as input features. Layers were assigned with individual weights that were defined by expert elicitation. Habitat layers were given an aggregate weight of 200 (a weight that was divided for different habitat layers based on their relevance for conservation), and species layers an aggregate weight of 100. See Supplementary Table S1-S2 for full list of the weights used.
We then developed multiple versions of the prioritisation analyses. We made them with and without considering connectivity between species distribution and habitat patches. For those analyses accounting for connectivity, the connectivity distance values were defined separately for individual habitat types and species with the expert group (see “Expert elicitation”). When connectivity is accounted for, higher priorities are given to areas that are well connected (spatially aggregated), even if the local habitat quality in some grid cells would not be as high as in some other areas (Lehtomäki et al. 2009).
We applied so called hierarchical prioritisation analysis (Mikkonen and Moilanen 2013) to examine how well the existing protected areas contribute to biodiversity protection. Additionally, we used the hierarchical analysis to identify the most efficient expansions of a protected area network.
Furthermore, we ran variants with and without a condition layer that can be used to modify the species or habitat data with other data sources that give additional information about habitat quality (Moilanen et al. 2011b). The basic use of the condition layer is to reduce habitat quality in locations that are known to be impacted by human activities. We used Corine Land Cover (2006) data as the basis for our landscape condition layer, complemented by some local data sources that described e.g. areas that had been built up after the production of the Corine or the species or habitat data sets (Supplementary material, Table S4). Highest condition values (1.0 = untransformed) were given to natural areas such as forests (in Corine) and lowest (0.001–0.1 = heavily degraded) to heavily-modified areas such as mineral extraction sites and industrial areas.
All different versions, or variants in Zonation language, were made with and without water areas and aquatic species, and solely for the region of the Helsinki-Uusimaa or the region plus a 15 km buffer around the region to make sure the results were not influenced by any edge effects.
Visualising and comparing the prioritisation results
Zonation outputs two main data products: a priority ranking map and performance curves of that prioritisation for every input feature. On the priority map, the pixels of the entire study area are ranked based on their importance for all input features. The rank ranges linearly from 0.0 (pixels with lowest value) to 1.0 (pixels with highest value). The performance curves report how large a proportion of each input feature (from the initial distribution of species and habitats) is included in a certain priority fraction of the study area (Di Minin et al. 2014; Moilanen et al. 2014).
We tested various visualisations for these two output types with the expert group and chose a visualisation through which the rank maps and the performance curves are coloured with the same green to sand colour palette. The performance curve background was coloured with the map colours to facilitate the comparison of landscape fractions in each priority bin intuitively (adopted from Pouzols et al. 2014). The map presentations were always shown with performance curves, accompanied with information on the data sets included in the analysis and a checkbox listing of factors that had been considered in the respective analysis. This standard layout was used to report the result of each analysis version. In addition to individual version variations, we produced difference maps for comparing different analysis versions.
Identifying important biodiversity areas for the plan
Continuous priority rasters produced by Zonation can inform the planners and stakeholders, but the actual plan must be made with distinct symbols (polygons, lines, points). To mark ecologically important areas in the new plan, the planners implemented a new planning zone called LUO (an abbreviation of the Finnish word for nature, luonto) to guide more detailed planning. For this, we identified areas in the top 10% priority ranks of Helsinki-Uusimaa that also included cells that belonged to the highest 5% priority fraction (see Supplementary S8). Considering both the ecological values and the scope of the regional plan, the planners selected those areas that were over 50 ha in size for further investigation and generalized the areas consisting of raster cells to smooth lined polygons using ArcGIS 10.0. To produce quantitative metrics of the biodiversity found in the LUO areas, a landscape identification analysis (LSM) was done in Zonation (Moilanen et al. 2005, 2014). Site descriptions included mean priority rank of the LUO area’s cells and list of noteworthy biodiversity features from the LSM analyses. In addition, a feature density index was developed to compare the aggregated biodiversity values across sites of different sizes. The feature density index for the site j is calculated as
where SDSj is the sum of feature distribution proportions of the site j (received from the LSM analysis), Aj is the area of the site j, SDSt is the sum of feature distribution proportions in the entire study landscape (in our case, Uusimaa), and At is the area of the study landscape. In other words, the feature density index compares the aggregation of biodiversity features in a site to the average distribution of biodiversity features in the study landscape. Finally, a descriptive “information card” was made for each LUO area. These included a map and basic information on the characteristics of the site, the biodiversity value and the feature density. Some LUO areas were also checked on-site by local municipalities to verify their importance.
Impact assessment of the strategic Uusimaa 2050 plan
The renewal of the regional plan entity started in 2017. The ecological impacts of the newly proposed plan draft (called the Uusimaa 2050 plan; Regional Counil of Uusimaa, 2018) were assessed with Zonation using the previous results as the starting point, particularly the one presented in Fig. 1 (analysis including connectivity and land use effects, but no protected areas). The plan draft included polygon symbols for future developments with high biodiversity impacts including residential, industrial, and commercial zones. Low biodiversity impact land use symbols included protected areas (including new areas to be implemented by the state) and recreational areas, as well as forestry areas that may have a varying impact on biodiversity depending on the management actions. Line-type symbols were used for roads, railways, and point-type symbols for small commercial centres. For the impact assessment, all data were converted to polygons and then rasterised for the use in Zonation. Linear features were buffered to be polygons with the width of the actual symbols in the Uusimaa 2050 plan (300-600m). Point-type small commercial zones were transformed into polygons with a 300m buffer radius, as suggested by the regional planners. We used the same Zonation post-processing methods and feature density index as in the LUO examination, to compare the planned land uses to the current biodiversity priorities.