Multiple linear regression models to predict seasonal occurrences of two important grapevine pest Lepidoptera in Austria
- 1Institute for Sustainable Plant Production, Department for Plant Health in fruit crops, viticulture & special crop, AGES-Austrian Agency for Health and Food Safety, Vienna, Austria (kerstin.kolkmann@ages.at)
- 2Institute of Meteorology and Climatology, Department of Water, Atmosphere and Environment, BOKU-University of Natural Resources and Life Sciences Vienna, Vienna, Austria
Sustainable crop production will in future face further challenges from insect pests due to climate change. Rising temperatures will enable higher overwintering rates and accelerate the development of thermophilic insects, leading to increased damage risks for regional crop production systems. These changes pose problems for optimum timing of monitoring and control measures, which could be countered with improved and new prediction models.
Within the ACRP-Project RIMPEST[1] new prediction models were developed for the European grapevine moth, Lobesia botrana (Denis & Schiffermüller) and the European grape berry moth, Eupoecilia ambiguella (Hübner) (Lepidoptera: Tortricidae) in Austria including monitoring data from 60 selected monitoring sites and measured weather data from adjacent reference weather stations in the period 1980 to 2023. Stepwise multiple linear regression (MLR) analysis was applied to generate prediction models for the first seasonal occurrence of the different developmental stages (egg, larvae, adult) of the first and second generation of L. botrana and E. ambiguella. As input data for the MLR analysis nine processed weather parameters, different calculation periods and the DOYs (day of year on which the first seasonal occurrence was observed) were used.
The performance evaluation of the six generated MLR models for predicting the different generations and developmental stages of L. botrana resulted in an R2 of 0.51 to 0.92, a RMSE of 2.18 to 3.97 and an average prediction range of 1.90 days too early to 1.40 days too late. For E. ambiguella the validation resulted in an R2 of 0.33 to 0.69, a RMSE of 3.48 to 4.15 and an average prediction range of 2.85 days too early to 1.00 day too early. The MLR models for E. ambiguella first generation egg and larvae were not sufficiently validated as too few datasets were available.
The implementation of the new MLR models for impact assessments under regional climate scenarios can help to determine the potential future risks of L. botrana and E. ambiguella occurrence in Austrian wine-growing regions. The inclusion of future observation data into the analysis, especially from years with extreme weather events, can further improve the prediction accuracy of the MLR models.
[1] ACRP-13th Call Project RIMPEST (KR20AC0K17957) ("The effect of changing climate on potential risks from important insect pests on plant production in Austria and related adaptation options").
https://www.klimafonds.gv.at/report/acrp-13th-call-2020/
https://rimpest.boku.ac.at/
https://www.ages.at/en/research/project-highlights/rimpest
How to cite: Kolkmann, K., Blümel, S., Thaler, S., and Eitzinger, J.: Multiple linear regression models to predict seasonal occurrences of two important grapevine pest Lepidoptera in Austria, EMS Annual Meeting 2024, Barcelona, Spain, 1–6 Sep 2024, EMS2024-311, https://doi.org/10.5194/ems2024-311, 2024.
