By Julienne Isaacs
If “technology transfer tool” can be defined as a way to get information into the hands of as many people as possible, weather-based disease forecasting models are the perfect example of how this works in practice.
By Julienne Isaacs
In Quebec, most producers are familiar with Agrométéo Québec (AQ) – or AgWeather Quebec in English – an agricultural management tool that offers real-time weather and climate information tailored to crop and region. It’s meant to help producers make informed spraying decisions.
AQ was developed through a collaboration between Agriculture and Agri-Food Canada (AAFC), the Quebec Department of Agriculture, Fisheries and Food, Solutions Mesonet and Environment Canada. The AQ tool was built to reference more than 40 bioclimatic models from the Computer Centre for Agricultural Pest Forecasting (CIPRA).
Until now, experts in the pest management warning network (Reseau d’avertissements phytosanitaires) have used a “rule of thumb” approach to Fusarium head blight (FHB) risk in the field, says Gaétan Bourgeois, an AAFC research scientist focused on bioclimatology and modelling. No mathematics-based FHB forecasting model is currently being used in Quebec.
“Over the years experts have developed a good sense of weather predictors for risk of infection. They looked at the temperature and forecast,” Bourgeois says. “But it’s a lot of work to gather weather data from each region and many sites and look at the risk index based on that data, so in a world where technology is getting faster and faster and we can grab data, we are aiming to make a map in a way that everything is done automatically. We’ll save a lot of time.”
Bourgeois is co-author on a new study comparing the accuracy of 10 different forecasting models for FHB in wheat. Over two growing seasons, Bourgeois and his colleagues gathered phonological, epidemiological and weather data from four experimental sites in Quebec (two in southern Quebec, one near Quebec City and one near Saint-Jean-sur-Richelieu), representing a variety of the province’s cereal growing regions. The team evaluated a hard red winter wheat cultivar and two spring wheat cultivars, using a variety of seeding dates in order to “create different weather scenarios in time,” Bourgeois says. They sprayed some plots with fungicide in a separate experiment evaluating fungicide efficacy, but the modelling experiment plots were left unsprayed to better assess disease risk.
Models used to predict FHB in Italy, Argentina, Canada and the United States were examined in the study, along with DONcast, a proprietary model used by the company Weather Innovations (WIN).
“Most of the models look at what we call polynomial linear regressions – they’ll look at the number of hours where relative humidity is over a given number over the last five days, at temperature and precipitation, and at flowering,” Bourgeois says. “Some are more dynamic or express the epidemiology of the disease, but essentially it’s all the same thing: they grab the weather, digest it into internal mathematics, and provide the disease forecast.”
Of the models studied, the two American models, De Wolf A and De Wolf B, provided the best disease prediction accuracy – even better than the Canadian models and others, according to Bourgeois. “Now we are in the process of integrating this model into the AQ website,” he explains. By the start of the 2017 growing season, producers will be able to log in on any given day, map the region and look at the FHB risks predicted by the forecasted models.
The AQ system is built to allow researchers to easily replace old models with newer models proven to be more accurate, which means producers will always be accessing the best possible information to help them make spraying decisions. “If we have upgrades or new models that are more efficient, they will be easy to apply,” Bourgeois says.
The search for the best, most accurate disease forecasting models is never over. Since the last study, Bourgeois and his colleagues have discovered more models used to predict FHB risk around the world.
They’ve just received a one-year grant from the provincial government to evaluate these models. The project will begin in April 2017 and run over the summer, for application on the AQ site by the 2018 growing season.
Bourgeois says the AQ site, and the research and development site CIPRA, which hosts bioclimatic models, have been receiving more interest from stakeholders in Quebec. “I’ve seen requests from people wanting to evaluate a model on a specific crop, or specific disease,” he says. “We’re getting a lot of requests of all sorts for many different crops.”
When the team first developed CIPRA, they made the decision not to make spraying recommendations, but rather to offer disease risk relative to weather conditions. CIPRA hosts a crop user guide that includes descriptions of pests, phenology and risks, and suggests times when producers should take action, but no specifics are offered in terms of products or pest control methods.“We let producers make their own decisions,” Bourgeois says.
Which is, after all, the goal of technology transfer.