1. Ecological niche modelling using cluster analysis to determine suitable environments of Giant Hogweed occurrence. Simon Fonji

2. Presentation Outline Background Objective Research questions Methods Results Conclusion

3. Giant Hogweed General Physiology: Why is it such a successful Invader in Latvia? - Mature height of 6 to 21 feet makes it taller than any other herbaceous plant or grass in Latvia - Leaves up to 1.5 meters across on lower 1/3 of stalk readily shade out competitors in the understory - Compound umbel (flower) can reach 2 meters in diameter and produce up to 100,000 seeds with over 90% viability rate - Huge taproot serves as nitrogen storage device and provides tremendous regenerative potential

4. Problems: Human Health Risk Extreme Phototoxic Dermatitis

5. Native Region: The Caucasus Includes: SW Russia Georgia NE Turkey Armenia Azerbaijan Northern Iran

6. Introduction to Latvia In the 1960’s, Giant Hogweed was planted in Latvia in places like Madona and Cesis as a silage crop to feed cattle

8. Objective use cluster analysis to group the most common classes of conditions where Giant Hogweed is present use the results from cluster analysis to create a Giant Hogweed suitability map Determine priority areas for conservation planning

9. Research Questions Can cluster analysis be used to accurately predict where Giant Hogweed will likely grow?

10. Data Two Landsat TM images obtained in the summers of 1992 and 2007. PPGIS data and other GPS data of Giant Hogweed Latvian demographic data from 1992 & 2007 Digital topographic layers and municipal boundaries.

11. Methodology Use presence-only data to group cluster of conditions of environmental variables Hierarchical-cluster analysis was used to determine the number of clusters K-cluster analysis was used to group the most common classes of conditions where Giant Hogweed is present. Variables used in cluster analysis are: slope, elevation, distance to roads/rivers/urban centers, land cover, land cover change, population density.

12. Hierarchical-cluster and K-cluster analyses Results Stage Cluster Combined Coefficient s Stage Cluster First Appears Next Stage Cluster 1Cluster 2Cluster 1Cluster 2 3671361935.399351364371 368103775.4423630369 103035.776368353372 37011915.830360365371 11366.018370367372 1106.471371369373 11257.108372366374 12409.326373362375 123910.5543740376 132711.80037500 Standardized Cluster Means 1234 Zscore(agricultural change)-.24007.90767-.07773-.16594 Zscore(elevation).08065-.05849-.33569-.07463 Zscore(dist_road_km).03159-.42826.33472-.25259 Zscore(dist_urbancenter_km).21010-.43524-.45707-.55315 Zscore(dist_water_km).04965-.11011-.03520-.19479 Zscore(forest change)-.23607-.313131.91316-.23138 Zscore(agriculture land cover).48235-.67534-1.93205.12605 Zscore(forest land cover)-.32168-.298582.35100-.12066 Zscore(urban land cover)-.349581.78076-.44222.01226 Zscore(nochange).47631-1.24443-1.09161.27748 Zscore(population density)-.15205.08813.61235.49152 Zscore(slope)-.23951-.20978-.182953.12492 Zscore(urban change)-.348241.84792-.36147-.05113 ANOVA ClusterError FSig. Mean Squaredf Mean Squaredf Zscore(agricchange)21.4773.77137327.855.000 Zscore(dem)2.1253.9713732.189.089 Zscore(dist_road_km)5.6503.9143736.178.000 Zscore(dist_urbancent er_km) 12.8393.89437314.358.000 Zscore(dist_water_km 0).79431.055373.753.521 Zscore(forestchange)54.6063.526373103.717.000 Zscore(lcagric)77.2583.366373211.123.000 Zscore(lcforest)82.3593.288373285.990.000 Zscore(lcurban)76.2283.441373172.950.000 Zscore(nochange)66.0993.461373143.285.000 Zscore(pop_den2)9.05431.0313738.785.000 Zscore(slope)90.7643.330373274.725.000 Zscore(urbanchange)80.1523.484373165.553.000 Table 5, F-test and significance of variables in the model Hierarchical cluster results K-cluster results

13. Four clusters were produced from K-cluster analysis characterized by specific environmental conditions Examples of conditions: –Cluster 1: 1.Elevation ≥64 and ≤183 meters 2.Slopes ≥0 and ≤0.89 degree 3.In agricultural areas 4.Near low population centers –Cluster 4: 1.Elevation ≥46.8 and ≤172.5 meters 2.Slopes ≥3.8 and ≤12.4 degree 3.Near urban centers 4.In high population centers These conditions were used to build a Giant Hogweed suitability map

14. Combine cluster For each cluster find range of variable values where the majority of Giant Hogweed pixels fall map shows all pixels whose environmental variables resemble any of the four clusters, and are therefore likely candidates for future Giant Hogweed colonization.

15. Habitat suitability map For each cluster, this shows the area where all 13 of the utilized variables fall between the 10 th and 90 th percentile. 27% of the Giant Hogweed testing points fall within this zone. 85% of the Giant Hogweed testing points fall with the zone where at least 12 of the 13 utilized variables fall between the 10 th and 90 th percentiles.

16. Results Cluster 1 occurs in mostly agricultural areas where the population density is low. These areas are mostly rural areas where agricultural abandonment is common. Cluster 2: occurs mostly in urban areas close to roads and urban centers. This is probably near big cities where urban expansion is taking place and where people have gardens or small farms in their compounds. Cluster 3: occurs near urban centers where land cover type and land cover change to forest. These are probably remote areas near cities that are far from roads and along forest edges. Cluster 4 occurs mostly in urban areas near roads where the population density is high. These are most likely small farm areas and gardens in big cities. Cluster 4 is very similar to cluster 2.

17. Conclusion Cluster analysis can be used as an effective tool to select sites with favourable conditions for Giant Hogweed occurrence based on environmental factors LULCC, demographic and geographic factors influence the spread of Giant Hogweed in Latvia. Giant Hogweed suitability map to predict potential spread – useful for managing and controlling Giant Hogweed in Latvia Applicable to other regions with high Giant Hogweed occurrence: – Larger Baltic region and Russia – Central and Western Europe – Northern USA and Canada

18. Thank you for listening.