Integrating statistics and geodata to identify indicators for landscape biodiversity
Fragmentation of landscapes is an important factor in biodiversity loss across the world. It is of utmost importance to develop measurements and thresholds to be able to monitor the effects of landscape and habitat fragmentation on biodiversity. Here we will use innovative methods and biogeography theory to develop landscape indicators for landscape fragmentation effects on biodiversity using past and present statistics and geodata. We will focus on different grassland habitats, as biodiversity hotspots, in rural landscape across all biogeographical regions of Sweden. We expect that land-use pressures (urbanization, intensification, abandonment) and fragmentation rate is different in different regions of Sweden, which will also affect biodiversity today and in the future. A combination of historical and present open-source landscape geodata and regional statistics will able us to identify timing and magnitude of fragmentation, land use and habitat quality. By investigating fragmentation and habitat quality change over time from different region we can identify thresholds when fragmentation starts to become negative for biodiversity. We will use historical maps (1800s), regional statistics from late 1800s and modern satellite imageries to map landscape fragmentation at larger landscape scales. Based on historical map analysis we will select smaller landscapes with many or few habitat fragments left. To develop indicators for biodiversity we also need to measure plant biodiversity (gene, species, community, function) in 40 landscapes across Sweden. We will also analyse how green infrastructure in rural landscapes can improve landscape biodiversity. Finally we will make suggestions for how landscape fragmentation should monitored. The project will report on how changes in land use of semi-natural and cultural environments affect the ecological context of different species, their habitats, structures and functions in rural landscape in a Nordic context.
Conclusions
The most suitable historical data to use in landscape analysis largely depends on the spatial accuracy and spatial and temporal resolution required. Parish level agricultural statistics provide data annually but because parishes vary widely in size it can be diffi cult to identify relevant changes affecting grasslands in large parishes dominated by forest. Historical maps have a high spatial resolution, but a scattered distribution both nationally and through time. Historical maps usually cover a relatively small area, with outland rarely mapped. Historical data should be interpreted with an under standing of the limited information that can be extracted from the underlying data. More historical maps should be digitized in geographic information systems to facili tate further analysis. The economic map from the mid1900s can be a good addition for analysing change over time, but with the caveats that maps differ in ages across the country. It is also difficult to reliably separate bare bedrock, clearcuts and grassland. Using historical satellite images was not satisfactory with the difficulty to obtain both cloudfree and snowfree images over all regions during the vegetation period and that it is not possible to use the same training areas across the country. The method of using machine learning and Sentinel2 L2A to monitor seminatural grasslands seems very promising and can be developed further.
Onehundredfifty years ago, there was an average of 42% grassland (open/semiopen land, not arable) within the 48 agricultural landscapes. Today, 2% original grassland remains in Norrbotten and Gävleborg, 6% in Skåne and 10% in Södermanland. Many agricultural landscapes have no grassland left at all, even though all 48 agricultural landscapes are still considered agricultural landscapes.The result shows that the proportion of natural grassland has decreased significantly more than what has been previously shown.
The decisive factor for today’s plant species richness is that there are pastures left in the landscape. Seminatural grasslands with long, continuous management are important, and the bigger grassland the better. Small grassland remnants had a low proportion of species, compared to previous studies. However, road verges in the northern regions are broad due to snow clearing, and have a relatively high proportion of grassland specialists. In agricultural landscapes with few or no pastures, small habitats can be important for many plants, but they cannot compensate for grazed grasslands, especially seminatural grasslands.
Analysis of the grassland specialist Harebell showed that population size plays a major role for its genetic diversity. Larger populations in the landscape means greater levels of genetic variation. Population size was higher nationally in more varied landscapes and landscapes with more pasture and seminatural grassland present. At the regional level, the results are not as clear. The lack of grasslands in many of the landscapes makes statistically robust analyses difficult.
Several of the 48 agricultural landscapes included in the analysis have had seminatural grasslands, according to the GIS layer in the TUVA database, but field visits determined that these were unmanaged and abandoned for several years. To be able to do landscape analyses and monitor grassland status, it is important that both the database and its downloadable GIS files are kept up to date.
The results show the importance of including everyday landscapes and lands capes from several of Sweden’s different biogeographical regions. Most previous landscape ecological studies have been based on agricultural landscapes with a relatively high proportion of habitats and high biodiversity from southern Sweden. By analysing landscape changes from a large number of agricultural landscapes from several biogeographic regions in Sweden, this study provides a higher generality and points even more clearly to the importance of increasing efforts to preserve and restore grasslands nationally before their biological values are lost.
Conclusions
The most suitable historical data to use in landscape analysis largely depends on the spatial accuracy and spatial and temporal resolution required. Parish level agricultural statistics provide data annually but because parishes vary widely in size it can be diffi cult to identify relevant changes affecting grasslands in large parishes dominated by forest. Historical maps have a high spatial resolution, but a scattered distribution both nationally and through time. Historical maps usually cover a relatively small area, with outland rarely mapped. Historical data should be interpreted with an under standing of the limited information that can be extracted from the underlying data. More historical maps should be digitized in geographic information systems to facili tate further analysis. The economic map from the mid1900s can be a good addition for analysing change over time, but with the caveats that maps differ in ages across the country. It is also difficult to reliably separate bare bedrock, clearcuts and grassland. Using historical satellite images was not satisfactory with the difficulty to obtain both cloudfree and snowfree images over all regions during the vegetation period and that it is not possible to use the same training areas across the country. The method of using machine learning and Sentinel2 L2A to monitor seminatural grasslands seems very promising and can be developed further.
Onehundredfifty years ago, there was an average of 42% grassland (open/semiopen land, not arable) within the 48 agricultural landscapes. Today, 2% original grassland remains in Norrbotten and Gävleborg, 6% in Skåne and 10% in Södermanland. Many agricultural landscapes have no grassland left at all, even though all 48 agricultural landscapes are still considered agricultural landscapes.The result shows that the proportion of natural grassland has decreased significantly more than what has been previously shown.
The decisive factor for today’s plant species richness is that there are pastures left in the landscape. Seminatural grasslands with long, continuous management are important, and the bigger grassland the better. Small grassland remnants had a low proportion of species, compared to previous studies. However, road verges in the northern regions are broad due to snow clearing, and have a relatively high proportion of grassland specialists. In agricultural landscapes with few or no pastures, small habitats can be important for many plants, but they cannot compensate for grazed grasslands, especially seminatural grasslands.
Analysis of the grassland specialist Harebell showed that population size plays a major role for its genetic diversity. Larger populations in the landscape means greater levels of genetic variation. Population size was higher nationally in more varied landscapes and landscapes with more pasture and seminatural grassland present. At the regional level, the results are not as clear. The lack of grasslands in many of the landscapes makes statistically robust analyses difficult.
Several of the 48 agricultural landscapes included in the analysis have had seminatural grasslands, according to the GIS layer in the TUVA database, but field visits determined that these were unmanaged and abandoned for several years. To be able to do landscape analyses and monitor grassland status, it is important that both the database and its downloadable GIS files are kept up to date.
The results show the importance of including everyday landscapes and lands capes from several of Sweden’s different biogeographical regions. Most previous landscape ecological studies have been based on agricultural landscapes with a relatively high proportion of habitats and high biodiversity from southern Sweden. By analysing landscape changes from a large number of agricultural landscapes from several biogeographic regions in Sweden, this study provides a higher generality and points even more clearly to the importance of increasing efforts to preserve and restore grasslands nationally before their biological values are lost.