Assessing and modeling economic and environmental impact of wheat nitrogen management in Belgium

Highlights

• The economic and environmental impact of Nitrogen fertilization was evaluated.

• A complete and generic methodology for tactical N optimization is proposed.

• Climatic conditions occurring between sowing and flag leaf stage greatly impacts N optimization.

• Environment × management interactions have to be considered when optimizing N.

• Environmental consideration is a more limiting factor than expected revenues for N optimization.

Abstract

Future progress in wheat yield will rely on identifying genotypes and management practices better adapted to the fluctuating environment. Nitrogen (N) fertilization is probably the most important practice impacting crop growth. However, the adverse environmental impacts of inappropriate N management (e.g., lixiviation) must be considered in the decision-making process. A formal decisional algorithm was developed to tactically optimize the economic and environmental N fertilization in wheat. Climatic uncertainty analysis was performed using stochastic weather time-series (LARS-WG). Crop growth was simulated using STICS model. Experiments were conducted to support the algorithm recommendations: winter wheat was sown between 2008 and 2014 in a classic loamy soil of the Hesbaye Region, Belgium (temperate climate). Results indicated that, most of the time, the third N fertilization applied at flag-leaf stage by farmers could be reduced. Environmental decision criterion is most of the time the limiting factor in comparison to the revenues expected by farmers.

Introduction

Future improvement in wheat yield will rely on the identification of genotypes and/or management practices that are best adapted to the environment (Chenu et al., 2011). However, the complexity of the genotype-environment-management practice interactions (GEM) requires setting up extensive and costly field experiments. Because resources are limited, in practice, breeders typically select new cultivars that are suited to a specific environment (Semenov and Halford, 2009). For practical reasons, such experiments are usually limited to (i) the geographical area targeted by the breeding programme and (ii) the climatic conditions encountered by the plant during the selection program. Furthermore, (iii) the selection is typically performed under management conditions in which sufficient nutrient levels are supplied to the crop. Incorporating a new trait into a crop takes 10–12 years, and only then it will be known if it has been effective in improving yield in the various environments (Asseng and Turner, 2007).

The environment has two main components that induce variability, respectively soil and weather. Within a given field, differences in texture, structure, and organic matter may induce high variability. These soil characteristics greatly affect the soil moisture content and the available water capacity for plants. They not only drive water stress but also, in turn, impact soil nutrient availability (Basso and Ritchie, 2005). Concerning climatic variables, it has long been demonstrated that both the average values of weather variables and the sequencing of weather events greatly impact the dynamics of crop growth (Semenov and Porter, 1995). Interactive stresses may have a greater impact on the final value of crop characteristics of interest (e.g., grain yield) than individual stresses (Riha et al., 1996). For these reasons, the importance of an accurate characterization of soil and weather inputs data increases as the environment becomes more limiting in terms of plant growth and development (Weiss and Wilhelm, 2006).

Concerning the management of crops, nitrogen (N) fertilization is probably one of the most important practices. The optimum N fertilization is known to vary within the same field and with each growing season as a result of the heterogeneity of soil properties, as well as inter- and intra-annual climatic patterns (Basso et al., 2012b). Furthermore, the decision-making process linked to N management remains complex because even if a spatial map of soil properties exists, the decision regarding the amount of N fertilizer to apply must be made without any prior knowledge of future weather conditions (Basso et al., 2011b). Consequently, experimentally determining how plant characteristics, either individually or in combination, affect crop performance under a wide range of growing conditions is an intractable task (Hoogeboom et al., 2004). In such a context, determining the optimum amount of and the most appropriate timing for N fertilizer is a challenge (Makowski et al., 2001).

Crop modeling approaches are powerful tools to allow a more comprehensive analysis of real-life processes (Sinclair and Seligman, 1996). Crop simulation models, such as STICS (Brisson et al., 2009), SIRIUS (Semenov et al., 2007), and SALUS (Basso et al., 2012a), are computerized representations of crop development, growth, and yield elaboration. They simulate the functions and impacts of the continuum of soil-plant-atmosphere systems (Hoogeboom et al., 2004). They integrate the current understanding of crop growth derived from physiological studies and phenotypic characteristics measured in various environments (Semenov et al., 2009). By dissociating processes that closely interplay in the real world and cannot be always observed directly, crop models have become engineering tools that extend the potentialities of field experimentation (Casadebaig and Debaeke, 2011). By highlighting gaps in our knowledge, they can be used to guide the direction of fundamental research (Semenov et al., 2007). Furthermore, they have demonstrated to be efficient in assisting in analyzing and deconvoluting any combination of complex GEM interactions (Asseng and Turner, 2007, Chenu et al., 2011). For these reasons, crop models have already proven to be well-suited to supporting decision-making and planning in agriculture (Basso et al., 2011a, Ewert et al., 2011). However, to properly address new environmental issues, the purpose of crop models needs to be widened by encapsulating them in modeling platform (Bergez et al., 2014, Brown et al., 2014) or by surrounding them with appropriate analysis algorithms (Dumont et al., 2014a, Talbot et al., 2014).

Crop models can help to improve farmers' decisions by assessing the probability that a certain outcome will occur under specific management practices and the given pedo-climatic conditions of a certain field (Basso et al., 2011a, Basso et al., 2012b, Houlès et al., 2004). Dumont et al., 2013, Dumont et al., 2014a, Dumont et al., 2015a have recently demonstrated how stochastically generated weather can be used to quantify the uncertainty that impacts on yield and N leaching in order to optimize N fertilization. However, until now, this approach had remained limited to strategic management.

The objective of this study is to optimize N management at the intra-annual level by modeling the within-season environment-management interactions. Winter wheat (Triticum aestivum L.) growth was simulated under multiple N strategies and a panel of environments. An environment was here defined by a given soil type and a wide variety of climatic conditions. Stochastically generated climate time series were derived so that the most advantageous and disadvantageous climatic variable combinations could be explored. Such probabilistic climatic scenarios were coupled with historical records made between sowing and the flag-leaf stage. Multi-objective decision criteria were computed to optimize the economic return of the assessed N practices while minimizing the adverse environmental impacts associated with potentially inappropriate N rates.