Some preliminary results from Climex and Maxent distribution modelling of Drosophila suzukii. Version 2. Changes from Version 1: Oregon locations where D. suzukii is found now georeferenced. Maxent model rerun with better Oregon data (Climex model not rerun because now Oregon data is a better fit to original model) Maxent logistic output now interpreted rather than raw some revised commentary included Detail maps include county boundaries October 15 2009 Martin Damus
Maps from Climex modelling: The Climex model doesn t match climate per se, what it does is model the organism s response to climate. So the output is a map that indicates where the climate is suitable for this organism to complete its life cycle, not where the climate is similar to its native range. This model was built on the native range of D. suzukii in Asia and extrapolated to North America. No adjustments were made to the model to accomodate known N.A. locations (none were deemed necessary). The potential outlier in the central valley of California may be from an irrigated field or from a fruit stand, not a growing location. The Oregon data, now georeferenced, is a better fit than in version 1 (in which the Oregon data simply represented the geographic centre of the county): the previous outlier in Umatilla county has moved over to the marginal zone, out of the unsuitable zone. The green patch in eastern Oregon does overlap with the eastern part of Umatilla county. See the next map.
Detail map of western North America. Note that most of the known locations are in good and better regions, indicating that the model projections are pretty good. I would be extremely interested to know the results of any surveys within the area of Oregon between Umatilla county and counties to the west. Climex model parameters: Moisture Index SM0 SM1 SM2 SM3 0.4 0.7 1.75 2 Temperature Index DV0 DV1 DV2 DV3 9.1 20 28 32 Light Index (not used), Diapause Index (not used) Cold Stress TTCS THCS DTCS DHCS TTCSA THCSA -11-0.0005 0 0 0 0 Heat Stress TTHS THHS DTHS DHHS 36 0.0007 0 0 Dry Stress SMDS HDS 0.3-0.001 Wet Stress (not used), Cold-Dry Stress (not used), Cold-Wet Stress (not used) Hot-Dry Stress TTHD MTHD PHD 30 0.3 0.001 Hot-Wet Stress (not used) Day-degree accumulation above DV0 DV0 DV3 MTS 9.1 32 7 Day-degree accumulation above DVCS DVCS *DV4 MTS 12 100 7 Day-degree accumulation above DVHS DVHS *DV4 MTS 36 100 7 Degree-days per Generation PDD 268
Number of generations predicted by Climex. This is modelled by modifying the degree-days parameter such that the model predicts the reported number of generations in parts of the species native range. It is coarse, due to the paucity of information explicitly detailing local generation numbers over the native range of the fly, and local response by the fly could vary from that modelled in any case. It would be interesting to see if number of generations as modelled translates to severity of damage. It does so far seem that number of generations modelled indicates approximate season of damage fewer generations = later season damage. Though remember that one female can give rise to 22,500 ovipositing daughters in only 2 generations assuming equal sex ratio, so even few generations could mean significant damage.
This is the result of running the known locations through a modelling system called Maxent. It projects climate similarity this time, not biological response to climate. So the matches above indicate how well the climate in the area under each pixel matches the climate where the fly is found. To get the best prediction one is supposed to use all available data, both native and introduced distribution. Translating a continuous distribution (raw Maxent output) to a presenceabsence format requires subjective identification of a threshold. Most authors seem to agree that one of the thresholds here marked broadly similar or very similar are the most accurate, however recently some have suggested that the threshold marked marginally similar would produce a more accurate outline of the total potential climate/geographic space in which the modelled organism may be found. In any case the system has limited power to identify potential new (that is, currently uninhabited) range. Therefore the potential climate/geographic space identified in western N.A. could be considered a more accurate prediction than that in eastern N.A. Climate variables used / available for use: Max. Temperature - Warmest Month *Mean Temp. - Cold Quarter Min. Temp. - Coldest Month *Mean Temperature - Warm Quarter Mean Diurnal Temp. Range *Total Days of Frost *Relative Humidity - Summer *Annual Precipitation *Precipitation - Driest Month Annual Mean Temperature Temperature Annual Range Mean Temperature - Wet Quarter Temperature Isothermality Temperature Seasonality Precipitation - Wet Month Mean Temperature - Dry Quarter Precipitation - Cold Quarter Precipitation - Warm Quarter Precipitation - Wet Quarter Precipitation - Dry Quarter Precipitation - Summer Precipitation - Winter Precipitation Seasonality Soil Moisture Summer Sunshine Wind Speed - Winter Wind Speed - Summer *= were most weighted by the model
I selected those climate variables that I thought would lend maximum information to the climate envelope this fly inhabits. In Asia it is limited by cold in the winter, by humidity during the growing season, and apparently by heat in the south. Using too many variables leads to a model that is too tight-fitting. Let me know if there are variables you think I should have included / excluded in the model, based on your local knowledge of the fly s distribution. Here s a map of only W. North America:
This is the Climex EcoIndex map for Asia. Note that most, but not all, of the Chinese data points are simply the major cities in those provinces in which this fly has been reported. Locations in other countries are latitude-longitude georeferenced specimens. Due to the lack of detail in China, the northern boundary of this fly s range may be incorrect, hence the projections into N. America may indicate too much northern range. Georeferenced data from China would be nice, as would more data from the Indochinese peninsula. Anyone got any?
This is the Maxent output for the native area. Note the overall less suitable climate area predicted by Maxent as compared with Climex. This may be due to too-tight fitting of the model (too many variables selected, not enough fuzzyness built in to the model), and/or to not enough information available to tighten the Climex model (e.g. geographic coordinate information for China, indicators of severity of damage in various areas of Asia, etc.). Finally, I am happy to share the memory-hogging original bitmaps used in this document, as well as all the results of the modelling and the data that went into the modelling. Just let me know. Martin Damus: CFIA Plant Health Risk Assessment: martin.damus@inspection.gc.ca All these maps are to be considered preliminary, based on the best information available to me. Updates will be made as more information becomes available. Use of the information presented in these maps to inform surveying protocols or other plant or pest management scenarios is entirely at your own risk. I do not warrant, guarantee or otherwise imply that these maps indicate exact locations, exact final distribution, or the extent of potential harm Drosophila suzukii may inhabit or cause.