Development and implementation of a phosphorus module of the STICS crop model to simulate tropical and temperate agroecosystems functioning

Overview & Synopsis

Soil-crop models are useful tools for understanding the complexity of agrosystems. The modeling of the P cycle in agrosystems remains underdeveloped compared with that of nitrogen (N) or water. In addition, crop models managing P use the same approach. Other aspects, such as P demand and soil-plant P transfer, are often poorly represented, which limits the genericity and robustness of these crop models. The central hypothesis of my work is that it is possible to represent the P cycle in agrosystems using a more mechanistic, robust and generic approach. In order to achieve this, it was first necessary to implement a module simulating the P cycle in the STICS soil-crop model, which has already demonstrated its robustness for simulating nitrogen and water, and which incorporates relevant formalisms to extend its validity domain to P. Secondly, the dependence of the P cycle on that of N raised the question of the agronomic importance of N-P colimitation and its implementation in soil-crop models. In the first part, I coupled the STICS model with a model simulating soil P supply, crop P demand, P uptake by roots, P partitioning in the plant and the consequences of P deficiency on crop growth. The model was evaluated on a dataset derived from four long-term phosphate fertilization trials for maize at four sites in France. The evaluation showed that the model performed well in simulating the final accumulation of P (R² = 0.82; Relative Root Mean Squared Error (rRMSE) = 16%) and biomass (R² = 0.62; rRMSE = 10%) in the crop, as well as P in grain (R² = 0.7; rRMSE = 20%) and grain yield (R² = 0.56; rRMSE = 11%). The model is also able to reproduce the dynamics of P accumulation and crop response to P limitation between different sites and treatments within the same site. The proposed model can be used to analyze and optimize crop response to P availability in temperate zones over a single growing season. Its extension to longer time scales or to tropical zones requires the integration of new mechanisms. In the second part, I conducted a literature review on N-P colimitation in agrosystems. I established a conceptual framework for classifying crop response to colimitation. Various fertilization trials from published works were analyzed using this conceptual framework. Of the 32 trials analyzed, 84% followed the multiple limitation hypothesis (MLH), which implies an adapted crop response to colimitation, while only 16% followed the law of minimum (LM), which is still used in agronomy. The NxP interaction was found to be significant in 63% of cases. Mechanisms that could explain crop response to N-P colimitation were identified. These experimentally observed mechanisms were compared with an inventory of soil-crop models managing N-P colimitation, the majority of which use LM. Avenues for improving colimitation management by crop models were proposed. In a third and final part, I applied the previously mentioned conceptual framework to the study of sugarcane response to colimitation for 3 elements: N-P-K (potassium). Crop response was characterized using nutrition indexes based on the establishment of corresponding dilution curves. Linear mixed models were developed to process data from four multi-year sugarcane fertilization trials in La Réunion. Colimitation situations were found to be frequent on the different sites (98% of situations were colimited by at least two elements). MLH was also able to explain the variability of sugarcane response better than LM, confirming the results of the literature review. In line with the scientific literature, the statistical models developed showed that the N, P and K nutritional status of sugarcane at harvest explains 38% of the variation in sugarcane yield at harvest. The integration of NxP interaction mechanisms would enable better management of N-P colimitation in the model, and a better agronomic understanding of these situations.