When I was a graduate student, I found that diamondback moth could not survive in the cold winter in northeast China (at Gongzhuling in 1988) based on my field survey in late winter, low temperature survival experiment and overwintering trials in the field. I further found that the immigrations of diamondback moth were synchronized with a well-known migratory species, the oriental armyworm. I inferred the moths migrated into Northeast China with the help of the southwest air flow, since the highest wind speeds were most frequently from the southwest and occurred at the same day with the peak of catches of the diamondback moth in pheromone traps.

Twenty years later, in 2008, my friend Dr. Li-Hua LV from Guangdong Academy of Agricultural Sciences (we studied in Prof. Rui-Lu Chen’s laboratory for our Msc.) invited me to join a project on pesticide-resistance monitoring and sustainable control of diamondback moth. He knows that I had experience to study ecology of diamondback moth. I asked myself whether the overwintering status and long-distance migration of diamondback moth were related to pesticide resistance? I knew the pesticide resistance in south of Yangzi River in China was very serious problem, but hardly heard the same problem in north part of China.

Anyway, I was interested in the geographic distribution of overwintering survivals especially the overwintering boundary of this moth in China and even in the whole world. We have conducted a low-temperature survival experiment in laboratory that simulated winter temperatures of 10 selected sites across a latitudinal gradient in China. To verify the cold survivals from the laboratory, we conducted field experiments for winter survival in multiple years at various geographic sites in China and preliminarily concluded that diamondback moth can stably overwinter in the south of the Yangtze River in China, sporadically overwinter in Henan Province, and cannot overwinter in the north China.

In 2015, we invited Dr. Volker Rudolf from Rice University to visit our lab. We discussed the potential collaboration to work on this paper. To predict the global overwintering survival, together with Volker, we identified which metrics and models best-describe winter survival across a climatic gradient based on the laboratory results and finally found exponential model based on low temperature degree-days (LTDD) was the best. To validate winter survival models for field conditions, we compared model predictions to observed survivals in field experiments. We further extrapolated our winter survival predictions to a global scale under present and future climate scenarios. By analysing experimentally-parameterised and field-tested models, we show that climate change over the past 50 years increased the overwintering range worldwide.

Now come to the key issue, if the level of pesticide resistance is associated to the distribution of overwintering success. We classified the regions into three types: 1) permanent overwintering region (survival≥5%), sporadic overwintering region (survival 1%~5%), non-overwintering region (survival<1%). We performed a comprehensive literature survey to collect data on pesticide resistance of the diamondback moth worldwide. After all selection with our criteria, we finally gathered 1,806 entries for pesticide resistance of field populations of the diamondback moth. We extracted data from each selected publication. We conducted a meta-analysis to test if pesticide resistance levels vary across different types of overwintering regions. We found that the pesticide resistance in permanent overwintering areas was significantly higher than that in non-wintering areas.

Volker suggested that besides overwintering success, other factors such as the temperature during growing season may also impact and even confound the role of overwintering success. We then systematically considered the role of other factors such as the effective temperature degree-days (ETDD) during growing season (by changing the annual number of generations), types of pesticides with different action mechanisms (by changing speed of resistance evolution), sampling location and year (may related with the intensity of pesticide application). To isolate the effects of overwintering types (permanent, marginal, transient) on resistance from other factors, we fit a linear mixed model with resistance ratio as the response variable, overwintering-type and pesticide variety as fixed factors, ETDD as a covariate, and combination of sampling location and year as a random factor. We evaluated the significance of each fixed factor, covariate, and their interactions. We still found that overwintering type is a very important factor to determine the pesticide resistance after considering these factors.

Pesticide resistance varies markedly from region to region due to different histories of insecticide application on the farm level, which could obscure the differences in statistics between different regions. However, stakeholders of pest control are especially concerned about high levels of pesticide resistance in their regions. Thus, we built quantile regression models driven by LTDD to predict the global distribution of resistance levels, especially we focused on the significant 0.85 quantile model for the distribution of the top 15% resistance levels in China and the world in 2012~2016. The meta-analysis of global datasets reveals that pesticide resistance levels are linked to the species’ overwintering range. By facilitating local persistence all year round, climate change can promote and expand pesticide resistance of this destructive species globally.

• Here I want to say this work is the results of close collaboration among all contributors. Dr. Wei Zhang mined the potential information from bodies of literatures, conducted meta-analysis for pesticide resistance and overwintering success and processed the manuscript submission with extraordinary patients. Ms. Yu Peng spent longtime to downloaded, processed, and GIS-mapped the big dataset for global distributions of low temperature, overwintering-survival, and pesticide-resistance for many times. Volker drafted the introduction, drew the nice picture of the diamondback moth, and read the manuscript for many times and gave many helpful comments. We calculated every step carefully. With Volker's encouragement, our manuscript was firstly submitted to Nature, and the editor assigned for review. The main challenge was the data analysis of our field experiments in south China, where the diamondback moth can continue the life cycle, we treated the low-temperature survival as > 90%. The reviewers suspected that if this assumption is removed, our conclusion may not be tenable, resulting in a rejection of the manuscript. We then removed this part of data and the conclusion is still valid. The associate editor Dr. Walter S. Andriuzzi of Nature Communications suggested that the weighted effect size and publication bias should be considered in our meta-analysis. During revising period, we reconducted the meta-analysis according to editor’s suggestion. We incorporated host availability of Brassicaceae plants to refine the pest distributions when one reviewer commented to consider the humidity in the global distribution. Finally, we received the acceptance email from Nature Communications.