©2023 International Society of Arboriculture Arboriculture & Urban Forestry 49(4): July 2023 193 validated each map of current climate suitability across Australia for each species, as well as current suitability across the native range for exotic species (i.e., those not native to Australia). Experts classified each model as acceptable or unacceptable. SDM Method: Climate Suitability at the Postcode Level We used the Postal Areas ASGS Ed 2016 Digital Boundaries from the Australian Statistical Geogra- phy Standard (Australian Bureau of Statistics 2011.) to extract climatic suitability at the postcode level (n = 2648) for each species. To do so, we calculated the median value of the 10 climate suitability maps for each species for each time period, thereby result- ing in 3 maps per species (i.e., for 2030, 2050, 2070). Next, by overlaying the climate suitability maps with postcode geospatial data, we extracted the average suitability score for each postcode. This score (rang- ing from 0 to 1) was then converted to categorical data using values of 2 species-specific thresholds commonly used to convert MaxEnt output from con- tinuous maps to binary maps: (1) fixed cumulative value 5 logistic threshold and (2) fixed cumulative value 10 logistic threshold. When applied to maps, these thresholds result in binary surfaces that include all but 5% and 10%, respectively, of the training sam- ples as presences. By applying these thresholds to the average suitability score for each postcode, we classi- fied postcodes as unsuitable for the species if the postcode’s suitability value lay between 0 and the value of the first threshold; marginal if the postcode’s value was between the first and second threshold, and suitable if it was above the second threshold. Niche Method For some species, there was insufficient data to create SDMs (i.e., fewer than 30 occurrence records) or the SDM did not pass the model validation step (i.e., poor AUC or TSS value or unrealistic map of current cli- matically suitable habitat). This was typical of exotic species with few records in Australia from GBIF/ ALA. For these species, we used a simplified approach whereby we obtained CHELSA data for average maximum temperature of the warmest month and average precipitation of the driest month, for each occurrence. We then calculated the minimum, maximum, 5% and 95% values across these species’ occurrences for the baseline period. This climate niche represents the conditions for which we know the species can likely tolerate. global climate models identified as per Sanderson et al. (2015)(ACCESS10, CESM1BGC, CESM1CAM5, CMCCCM, FIOESM, GISSE2H, INMCM4, IPSLC- M5AMR, MIROC5, and MPIESMMR) for Repre- sentative Concentration Pathway 8.5 (RCP 8.5). The RCP 8.5 trajectory was chosen to provide the worst- case scenario when modelling. CHELSA does not provide data for the year 2030; hence, we generated these for each global climate model through a linear interpolation based on the baseline and 2050 data. As such, the 3 future time periods we used are centred in 2030, 2050, and 2070. All data were reprojected to a Mollweide equal-areas projection (ESRI: 54009)(1 km × 1 km). SDM Method: Modelling Approach Our approach to model fitting and validation proce- dures can be found in Burley et al. (2019). For each species, we modelled the distribution of suitable cli- mates for 2030, 2050, and 2070 using a machine learning approach, MaxEnt (Phillips et al. 2006; Elith et al. 2011), which has been shown to be more effec- tive compared with other models when true absence data is unavailable (Elith et al. 2006). In addition to occurrence records, MaxEnt requires user-defined background locations for which it can compare the climate to that where the species is present. For each species, 50,000 random background locations were extracted from within 500 km of the target species’ occurrence records. Following Burley et al. (2019), linear, product, and quadratic features were used to fit models in addition to default values for other parameters. Models were deemed to be of acceptable quality if values for the average test AUC (area under the receiver operating curve) and TSS (true skill statistic) were ≤ 0.7 and 0.5, respectively. Models were projected onto climate data representing current and future time periods. Additionally, Multivariate Environmental Similarity Surfaces (MESS) maps were generated to identify areas where the model was required to extrapolate into non-analogue climates (Elith et al. 2010; Di Cola et al. 2017). Regions of model extrapolation were then removed from maps of projected climate suit- ability to reduce uncertainty. While AUC and TSS values are useful for identi- fying poor quality models, models with acceptable values of these variables can still produce ecologi- cally unrealistic maps of the distribution of a species’ suitable climate. Hence, 3 plant experts independently
July 2023
| Title Name |
Pages |
Delete |
Url |
| Empty |
Search Text Block
Page #page_num
#doc_title
Hi $receivername|$receiveremail,
$sendername|$senderemail wrote these comments for you:
$message
$sendername|$senderemail would like for you to view the following digital edition.
Please click on the page below to be directed to the digital edition:
$thumbnail$pagenum
$link$pagenum
Your form submission was a success. You will be contacted by Washington Gas with follow-up information regarding your request.
This process might take longer please wait
