Results:
A total of 249,769 records of Reynoutria specimens with locality
information (i.e., latitude and longitude coordinates) were downloaded
from the GBIF database. These included 232,744 records for R.
japonica , 11,608 records for R. sachalinensis , and 4,377 records
for R. × bohemica . After geographic binning, and
grid-based filtering, the European R. japonica dataset included
721 records; the European and North American R. sachalinensisdatasets included 432 and 114 records, respectively; and the European
and North American R. × bohemica included 342 and 92
records, respectively. Model accuracy for the MaxEnt analyses based on
the ‘area under the receiver-operator curve’ (AUC) approach (Merow et
al., 2013), indicated that all models had high AUC scores (EuropeanR. japonica AUC = 0.948, European R. sachalinensis AUC =
0.966, European R. × bohemica AUC = 0.969, North AmericanR. sachalinensis AUC = 0.985, North American R. ×bohemica AUC = 0.986). In total, 10 of the 18 BioClim variables
provided ≥ 10% contribution to the MaxEnt model predictions for at
least one of the datasets (Table 1). In Europe, five variables
contributed ≥10% to the climate suitability envelope for R.
japonica , four variables contributed ≥10% to the climate suitability
envelope for R. x bohemica , and three variables
contributed ≥10% to the climate suitability envelope for R.
sachalinensis . Two of the 18 variables, Bio 6 (min temperature of the
coldest month) and Bio14 (precipitation of the driest month), were
included in models for all three focal species. In North America, two
variables contributed ≥10% to the climate suitability envelope forR. × bohemica , and three variables contributed ≥10% to
the climate suitability envelope for R. sachalinensis . Two of the
18 BioClim variables, Bio 1 (annual mean temperature) and Bio 19
(precipitation of the coldest quarter) were present in models for both
focal species.
Predictions of climate suitability estimated from the European records
indicated that the source locality of the Kyushu strain had no predicted
suitability based on records for all three focal knotweed species, the
source locality of the Murakami strain had no-to-low predicted
suitability based on records of R. japanoica and no predicted
suitability for based on records of R. × bohemica orR. sachalinensis , and the source locality of the Hokkaido strain
had low-to-medium predicted suitability based on records for all three
target knotweed species (Figures 1-3). A presentation of predicted
climate suitability based on European records across a wider portion of
central and east Asia are presented in Supplemental Figures S1-S3.
Predictions of climate suitability estimated from the North American
records indicated that the source locality of the Kyushu strain had
no-to-low predicted suitability based on records for R. xbohemica and no predicted suitability based on records forR. sachalinensis , the source locality of the Murakami strain had
medium-to-high predicted suitability based on records for both R.× bohemica and R. sachalinensis , and the source locality
of the Hokkaido strain had medium-to-high predicted suitability based on
records for R. sachalinensis and low-to-medium predicted
suitability based on records for R. × bohemica (Figures 4
and 5). A presentation of predicted climate suitability based on North
American records across a wider portion of central and east Asia are
presented in Supplemental Figures S4 and S5.