Management and spatial resolution effects on yield and water balance at regional scale in crop models

Highlights

• Management choices can influenced significantly model output variables at regional scale.

• Management effect is usually stronger than scaling effect, particularly on regional yield.

• Scaling and management effects varied greatly between crop models and output variables.

• Management and scaling effects are greater when analyzed for individual year.

Abstract

Due to the more frequent use of crop models at regional and national scale, the effects of spatial data input resolution have gained increased attention. However, little is known about the influence of variability in crop management on model outputs. A constant and uniform crop management is often considered over the simulated area and period. This study determines the influence of crop management adapted to climatic conditions and input data resolution on regional-scale outputs of crop models. For this purpose, winter wheat and maize were simulated over 30 years with spatially and temporally uniform management or adaptive management for North Rhine-Westphalia (˜34 083 km²), Germany. Adaptive management to local climatic conditions was used for 1) sowing date, 2) N fertilization dates, 3) N amounts, and 4) crop cycle length. Therefore, the models were applied with four different management sets for each crop. Input data for climate, soil and management were selected at five resolutions, from 1 × 1 km to 100 × 100 km grid size. Overall, 11 crop models were used to predict regional mean crop yield, actual evapotranspiration, and drainage. Adaptive management had little effect (<10% difference) on the 30-year mean of the three output variables for most models and did not depend on soil, climate, and management resolution. Nevertheless, the effect was substantial for certain models, up to 31% on yield, 27% on evapotranspiration, and 12% on drainage compared to the uniform management reference. In general, effects were stronger on yield than on evapotranspiration and drainage, which had little sensitivity to changes in management. Scaling effects were generally lower than management effects on yield and evapotranspiration as opposed to drainage. Despite this trend, sensitivity to management and scaling varied greatly among the models. At the annual scale, effects were stronger in certain years, particularly the management effect on yield. These results imply that depending on the model, the representation of management should be carefully chosen, particularly when simulating yields and for predictions on annual scale.

Introduction

Large-scale assessment studies based on simulations by crop models are frequently used to evaluate the impacts of agriculture. These studies usually focus on predictions of crop production in different contexts, such as climate change, its inter-annual variability, or trends over time (Gaiser et al., 2010; Nendel et al., 2013). Crop models are also used to study carbon sequestration or the greenhouse gas balance at regional or national scale (Gaiser et al., 2009, 2008; Tornquist et al., 2009). Other studies focus on the water balance and its dynamics at the watershed scale. For the latter, crop models are combined with other models (e.g., hydrological) and applied to quantitative water management and irrigation issues (Noory et al., 2011; Robert et al., 2018; Therond et al., 2014).

Crop models are useful tools for large-scale assessment since exhaustive measurements are not feasible or available. However, they were developed to simulate homogeneous fields, each represented by a combination of one soil and one climate. Some of these models were designed to simulate only one season, e.g. one crop and its management, while others are capable of simulating different crops in sequence, mimicking a crop rotation over a longer time period (Kollas et al., 2015). When applied at a larger scale, these models are usually applied in a gridded approach, simulating each grid cell independently, while assuming homogeneity within each grid cell (De Wit et al., 2012; Huang et al., 2015; Mo et al., 2005; van Ittersum et al., 2013). For such approach, it is necessary to provide input data for soil, climate, and management for each simulated unit. Depending on the study and the systems’ heterogeneity, the number of homogeneous units can range from a few to millions. Such data, especially management data, are not easily available at large scales and at high spatial or temporal resolution. Several methods exist to scale-up the data over the whole study area, such as sampling, aggregation from fine to coarser resolution, extrapolation or interpolation of the available data (Ewert et al., 2011). As an alternative, management information can also be simulated for large-scale studies (Hutchings et al., 2012).

Nowadays, it is possible to obtain soil and climate data at a relatively high resolution and at a large or even global scale from databases such as those in the Global Soil Map project (http://globalsoilmap.net/), the European soil portal for soil, the SoilGrids project (soilgrids.org) and the international CORDEX initiative for climate projection (https://www.euro-cordex.net/). On the other hand, the available databases on crop management data are at coarser resolutions such as those reported by Portmann et al. (2010) and Sacks et al. (2010) for crop growing periods or earthstat.org for fertilizer inputs. Usually, the few data available on crop management come from interviews with farmers, local experts, or observation networks. It provides an average date of sowing, harvest, and fertilization for instance or fertilizer input amounts for a given region for different crops and generally concern only one or a few years. Some initiatives such as the observation network of the German weather service DWD documenting key phenological stages as well as sowing and harvest could provide useful data for regional modelling (Kersebaum and Nendel, 2014) but do not cover the wide range of cultivation operations such as nitrogen fertilization for instance. As a result, large-scale studies usually consider management as uniform across the region and fixed over multiple years. However, it is well known that crop management, such as sowing, varies over space and time (Leenhardt and Lemaire, 2002). Additionally, the sowing date significantly impacts crop development and yield (Bonelli et al., 2016), and influences subsequent management actions during season.

To address the scarcity of the data and to adapt the management to the local and annual conditions, some authors suggested using management rules. Such management rules aim at reproducing the behavior of farmers and their crop management strategies (Maton et al., 2005; Nendel, 2009; Senthilkumar et al., 2015). In addition, these rules would help identify better management strategies. For example, suitable climate and soil conditions could be identified to perform cultivation operations (e.g., avoiding soil compaction by triggering an operation when the soil is not too wet or avoiding the risk of frost for spring crops). This adaptive management, based on management decision rules, could have a strong impact on model outputs but is rarely investigated at a large scale. Since the impact of input data aggregation and adaptive management can differ according to the output variables and crop models, these effects should be investigated with respect to a range of different crop models, output variables, and cultivation operations (i.e. sowing, soil tillage, irrigation…).

The objective of this study was to analyze the effect of adaptive management and spatial resolution on regional yields, evapotranspiration, and drainage predicted by a set of crop models. The main issues addressed were (1) whether adaptive management and/or input resolution influence the crop models’ outputs at the regional scale, in which way and how much and (2) whether the scaling effect varies when management changes over time and space.

To meet this goal, we quantified the impact of adaptive management and input resolution on the regional mean of simulated yield, evapotranspiration, and drainage for each individual year as well as for the 30-year average. We further analyzed whether the impact of management or spatial resolution depended on the crop model, output of interest, crop, or cultivation operation. To do so, we introduced adaptive management for sowing dates, fertilization dates, and crop maturity classes based on decision rules and variable amounts of nitrogen fertilization.