To what extent can ecoclimatic indicators assist crop performance predictions in oilseed rape upon repeated heat stresses?

Modelling is an obligate approach to predict crop yield under a wide range of environmental conditions. The present work aims to improve predictions of crop performance under repeated high temperature events by (i) demonstrating that ecoclimatic indicators (i.e. calculated for a given phenological stage) allow a finer description of the plant critical stages, and (ii) highlighting the non-additive effects of these successive stressful events. For this purpose, statistical models based on large datasets in oilseed rape were developed to look for correlations between ecoclimatic indicators and the plant performance-related variables (i.e. seed yield, lipid and protein content), as follows: (i) the plant cycle was divided into four intervals after flowering, according to the physi-ological stages of development in oilseed rape; (ii) the number of warm days (i.e. above 25 ???C and 30 ???C, respectively) in each interval for 26 combinations of location x year in France was scored; (iii) several statistical models that differed from the combination of ecoclimatic indicators were evaluated; and (iv) the best fit model for each plant performance-related variable was selected following a stepwise approach based on the Akaike Information Criterion. The results highlighted that contrasting final crop performances were tightly related to the timing, frequency and intensity of high temperature events after flowering. In addition, specific combinations of these ecoclimatic indicators were much more predictive of the crop performance-related variables than a single cumulative indicator which reflects the sum of all stresses in the same period. These results support our prior assumption that the outcome of several successive stressful events is not equal to the sum of each individual effect. The proposed approach is a proof of concept of the need to consider stress memory (i.e. the capacity of plants to store and retrieve information acquired during an initial exposure to stress) in predictive crop models, so as to better estimate the effects of repeated stresses and their consequences on crop yield and quality of harvested products.