Fragmentation revisited 050902
Download as PPS, PDF3 likes654 views
The document discusses landscape fragmentation, emphasizing the importance of landscape metrics in evaluating habitat connectivity and structural changes. It examines various definitions and impacts of fragmentation, along with methodologies for calculating metrics across different spatial scales. Additionally, it highlights the challenges and recommendations for ensuring the quality of input data and selecting appropriate metrics to address management objectives in forest and landscape studies.
Fragmentation revisited 050902
- 1. Fragmentation revisited The use of landscape metrics: Definitions, concepts, calculation and interpretation Niels Chr. Nielsen, SDU-Esbjerg, Denmark nichni@mail.dk Alan Blackburn, Lancaster University alan.blackburn@lancaster.ac.uk
- 3. - Low or diminishing forest cover (Mayaux et al) - Breaking up of habitat (Forman) - Landscape transformation (Kouki & Lofman) - Loss of connectivity (Delbaere & Gulinck) - Presence of isolated patches (Skole & Tucker) - Perforation (Riitters and Coulston) along with other forest patterns: patch, transitional, edge, interior (core) Definitions of Fragmentation However, is it always or necessarily harmful ? Equally important (harmfull) as habitat loss ?
- 4. Increasing fragmentation Habitatloss Ecological impact of fragmentation
- 5. Process and/or pattern Fragmented state Land Cover Fragmentation process (agents, decisions, information) Land Cover Change (rate, locations) Feedback possible – though not necessarily Feedback: Cellular models, GIS with supplementary information LANDSCAPE Remote Sensing, mappingProperty (observed) DRIVES SEEN IN SCALE? Cell size? MMU?
- 6. Choice of spatial metrics ..the ideal shape index should : From Forman (1995) • be easy to calculate, • unambiguously and quantitatively differentiate between different shapes, and finally • permit the shape to be drawn based on knowledge of the index number alone •work over the whole domain of interest, 4 1SqP Edge A* forest −= )*(# PPU sizepixelpixels NP = M forestwindow A*A Edge 10*=
- 7. Moving Windows approach Map 1: Window (user choice): Map 2: Grain = pixel size = 30m Size (extent) = 9 pixels = 270 m Grain = pixel size = 90 m Extent = 30*30 pix = 900*900 m Step = 3 pixels = 90 m Extent = 8*8 pixels = 720*720 m • As implemented with calculation of Fragstats-derived and other spatial metrics for “sub-landscapes” INPUT: “cover type” map(1) OUTPUT: metrics/index value map(2) DeterminesApplied to equals 1 2 3 4 5 Calculate (e.g.) Patch type Richness
- 8. Metrics maps
- 9. Analysis of results Various types of scale-dependence and plots/diagrams quantifying and illustrating it : Tools for selection of most robust and representative metrics of fragmentation 1. Response of metrics to window – or pixel – size (scalograms) 2. Variability and autocorrelation of metric, with changing window size: plots (”variograms”) 3. Relationships between different metrics, with changing window size: scattergrams (single), tables (multiple window sizes) 4. Relationships between same metric derived from different data sources: scattergrams (single), correlograms (multiple window sizes)
- 10. - If single shape index required, use Matheron index - Count or density of background patches (perforating forest), also as alternative to Lacunarity measures - Window sizes around 5 km acceptable for regional monitoring (pixel sizes 100–200 m) - Patch count metrics are highly sensitive to grain/pixel size - If possible normalise, and compensate for window-size effects Recommendations from study of metrics behaviour
- 11. Forest concentration – calculation flow Cover fraction Masking (criteria) 1 scapeCover_land Cover_mask FC win win −= FC20 = 0.086257 FC10 = 0.156578 FC5 = 0.226093
- 12. Forest concentration profile – combined for 50*50 km study area
- 13. Forest concentration profiles Italian regions
- 14. Forest concentration profiles Watersheds in North and Middle Italy
- 15. Management uses – statistical description of fragmentation • Overview of (differences in) landscape structure Recommendataions for.. • Monitoring temporal changes in points or regions • Overcoming/bypassing the MAUP by being multi- scalar and with the region of calculation being user-chosen (F.C. profiles as well as average metrics values) • Meeting the need for indicators of sustainable forest and landscape management? • Targets – threshold values ??
- 16. Management uses – local display M-index Cover fraction PPU Red Green Blue Central Umbria 50*50 km, 25 m pixels Landsat TM June 1996
- 17. Management uses – regional display Northern Italy 700*500 km, 200m pixels Classified WiFS Mosaic 1997 M-index Cover fraction PPU
- 18. Conclusions 1 - The “moving-windows” approach has made it possible to calculate metrics values throughout the study areas and to visualise and statistically analyse regional differences. - Limiting to the use of spatial metrics as indicators is the quality of the input data, i.e. maps or satellite images. Often a higher thematic resolution than what is normally available from LUC data is needed for meaningful comparisons for assessment of forest and nature/habitat diversity. It was however found that binary forest-non-forest maps constitute a sufficient input for analysis of forest fragmentation.
- 19. - Spatial metrics have the potential to function as indicators of landscape structure and diversity. Forest Concentration profiles facilitates comparison of regions. - Which specific metrics to use for a particular environmental assessment will depend on the management objectives for the landscape, forest or nature area of interest. - ToDos: Test, sensitivity analysis. Neutral Landscapes, agent based approaches.. Conclusions 2
Editor's Notes
- #3: Motorway project in central Jutland, Denmark, though sensitive area with lakes and mixed forest Deforestation in Cote d’Ivoire, West Africa (fra http://www.grid.unep.ch/activities/global_change/cote_ivoire.php)
- #6: Semantic network of concepts: Fragmentation typically referring to landscapes, seen from synoptic perspective. Connectivity typically referring to single patch or landscape element, organism persepctive (as in agent-based modelling: random walks etc.) Still a challenge to link the two!!
- #7: Delibarately NOT chosen fractal dimension, following recent literature and own experiences.
- #10: In a previous study, like in most of the examples found in the literature, focus was on changing pixel sizes Ideally for a fragmentation metric, 1. Should be low,or at least predictable, 2 show high var., 3 be the same at all win-sizes, and 4 show good agreement (high correlation)
- #12: From satellite image (WiFS) to forest-non-forest map, with M-W analysis to cover fraction maps, To forest-presence maps (masks), both with different window sizes TO BE ANIMATED!!
- #16: To be modified…
- #18: OBS.Histogram Equlisation of RGB-comb!!!
- #19: From this talk as well as from the project leading to my thesis… Ad point (1), multiple window sizes should not only be used in the search for the ”best” window size. Such a thing may never be identified, having values from different window sizes and presenting them informatively could be useful in comparison and communication of landscape structure. I will not draw any conclusions on the state-of-the-art in current research, just say that new interesting approaches are being tried out at various spatial scales and various levels of implementation