Agricultural Production Systems Simulator Simulations Research Articles

Evaluating the Agricultural Production Systems sIMulator (APSIM) wheat module for California

Context Computer-based crop simulation models are important tools for agricultural research and management. APSIM (Agricultural Production Systems sIMulator) is commonly used around the world but has not been widely validated in North America. Aims The objective of this work was to evaluate the reliability of APSIM for simulating wheat production in California, with the aim of providing guidance for future field research aimed at model calibration and validation. Methods Environmental and management data from state-wide wheat variety trials of common wheat (Triticum aestivum L.) were used to parameterise the APSIM-Wheat module (ver. 7.10 r4220). Simulated yield and protein data were compared with observed field trial results to test the reliability of APSIM simulations. Key results The most reliable simulation of grain yield had a root-mean-square error of 1040 kg/ha and normalised root-mean-square error of 16% relative to actual field data. Preliminary calibration of the model for Californian wheat varieties did not improve simulation accuracy or precision. Conclusions The accuracy or precision of the simulations was comparable to that of other tests of the APSIM-Wheat module in environments where it has not been previously calibrated but was considered too low to be reliable. The lack of reliability was due to the poor representation of local Californian wheat genotypes by existing APSIM cultivars, as well as possible lack of precision and accuracy of field data. Implications APSIM could be a valuable tool for wheat research and management in California; however, further research is needed to generate suitable field data for model calibration and validation.

Global sensitivity analysis of APSIM-wheat yield predictions to model parameters and inputs

The performance of cropping system models is often limited by uncertainties in model parameters and inputs. This study aims to explore how model yield prediction responds to these uncertainties under various environmental and management conditions. The Sobol’ method was used to investigate the sensitivity of the Agricultural Production Systems SIMulator (APSIM)-Wheat yield prediction to the following factors: air temperature (maximum and minimum), precipitation, initial soil nitrogen content, nitrogen fertilization amount, and soil hydraulic parameters. Eighteen scenarios were defined to consider a combination of three weather conditions (wet, normal, and dry growing season, based on the historical climatology of the Wimmera district of Victoria, Australia), three soil types (sandy, loamy, and clayey soils), and two nitrogen fertilization applications (50 and 100 kg N/ha). Eight thousand APSIM simulations were used to calculate the first order and total sensitivity indices for each scenario. The effects of weather, soil water characteristics and nitrogen availability were estimated by measuring their impacts on sensitivity indices. Our results show that yield was more sensitive either to variables that control water availability (precipitation and soil type) or variables that control nitrogen availability, depending on which was more limiting to wheat growth under the scenario. For example, in case of low nitrogen fertilization scenario, the initial nitrogen content sensitivity ranked first. Variation of this factor contributed 64% to 79% to the variance in simulated yield for clayey soils, nearly four times higher than the second-ranked factor (nitrogen fertilization amount). Soil hydraulic parameters and precipitation were most important when the crop growth was more constrained by water availability than by nitrogen availability. In the case of sandy soils in dry years with high nitrogen fertilization level, soil parameters and precipitation accounted for 83% of the yield variability. Minimum temperature consistently ranked as the least important factor under all scenarios. This work will help researchers better understand the model inputs’ impacts on the simulated yield variability under various soil, management, and weather conditions. The results also support agronomists and practitioners in the optimal use of the APSIM-Wheat model as a management tool.

Controlled traffic farming effects on productivity of grain sorghum, rainfall and fertiliser nitrogen use efficiency

Controlled traffic farming (CTF) is a mechanisation system in which all machinery has the same (or modular) working and track width so that field traffic can be confined to the least possible area of permanent traffic lanes. CTF enables productivity of non-compacted crop beds to be optimised for given energy, fertiliser and water (rainfall) inputs. This study investigated the agronomic response and economic return of grain sorghum grown in compacted and non-compacted soils to represent the conditions of non-CTF and CTF systems, respectively. Yield-to-nitrogen (N) responses were derived following application of urea, 3,4-dimethyl pyrazole phosphate-treated urea (DMPP), and urea ammonium nitrate (UAN, 32% N) at rates between 0 and 300 kg ha −1 N. Selected soil properties were measured to guide parametrisation of the Agricultural Production Systems Simulator (APSIM), which was used to assess long-term (55 years) effects of CTF and non-CTF soil conditions on crop productivity, rainfall use efficiency (RUE) and develop rainfall-runoff relationships. Grain yield and yield components (harvest Index, grain thousand-grain weight, number of grains) were significantly higher in CTF compared with non-CTF. On average, the most economic N rates, and corresponding grain yields, were 144 and 3428 kg ha −1 , and 100 and 1796 kg ha −1 for CTF and non-CTF, respectively. When N inputs were optimised, agronomic efficiency calculations showed 18% increase in CTF compared with non-CTF. Nitrogen use efficiency (NUE) was 1.75 times higher in CTF than in non-CTF. Rainfall-use efficiency was about 65% higher in CTF, which concurrently reduced the amount of runoff compared with non-CTF. Average rainfall season (330–450 mm in-crop) grain yield was 30% lower in non-CTF compared with CTF. For subtropical conditions of Australia, long-term APSIM simulations showed that increased productivity and inter-season yield stability can increase gross margin of grain sorghum by AUD74 ha −1 or greater depending on the adopted tillage system and in-crop rainfall. In non-CTF systems, improvements in NUE and RUE are constrained by soil compaction. Enhanced efficiency fertilisers, such as DMPP-treated urea, cannot compensate for other stresses caused by soil compaction and therefore cannot achieve the same NUE and RUE as the CTF system. Adoption of CTF delivers improved resource-use efficiency and profitability in rainfall-limited environments. Long-term (1960–2015) simulation of grain yield (A) and runoff (D) as a function of in-crop rainfall. Modelled data for grain sorghum-fallow cropping on a Red Ferrosol in Toowoomba (Queensland, Australia) in controlled (CTF) and non-controlled traffic farming (non-CTF) systems, respectively. Continuous lines show the best fit to predicted data. Dotted vertical lines show the 30th (left) and 70th (right) percentiles rainfall, respectively. • Controlled traffic farming (CTF) increased long-term grain yields by 30% in average-rainfall seasons. • Fertiliser nitrogen use efficiency (NUE) was 1.75 times higher in CTF compared with non-CTF. • Rainfall use efficiency (RUE) was 65% higher in CTF compared with non-CTF, which significantly reduced runoff. • Avoidance of (random) field traffic is a pre-requisite for jointly improving fertiliser NUE and RUE. • In non-CTF, improvements in NUE and RUE are constrained by compaction. • Enhanced efficiency fertilisers cannot compensate for other stresses caused by compaction.

A protocol to build soil descriptions for APSIM simulations

Soil processes have a major impact on agroecosystems, controlling water and nutrient cycling, regulating plant growth and losses to the wider environment. Process-based agroecosystem simulation models generally encompass detailed descriptions of the soil, including a wide number of parameters that can be daunting to users with a limited soil science background. In this work we review and present an abridged description of the models used to simulate soil processes in the APSIM (Agricultural Production Systems sIMulator) framework. Such a resource is needed because this information is currently spread over multiple publications and some elements have become outdated. We list and briefly describe the parameters, and establish a protocol with guidelines, for building a soil description for APSIM. This protocol will promote consistency, enhancing the quality of the science done employing APSIM, and provide an easier pathway for new users. This compilation should also be of relevance to users of other models that require detailed soil information.•This paper presents a brief description of the models for simulating soil processes in the APSIM model.•The method stablishes guidelines to define the parameters for building a soil description for APSIM.

Assessing the long-term impact of conservation agriculture on wheat-based systems in Tunisia using APSIM simulations under a climate change context.

Several circulation models are forecasting climate changes in the Mediterranean region. Accordingly, it is expected that water scarcity in the region will be higher with drastic shifts of hydrological and erosive watershed responses. In Tunisia, wheat yields have been variable over the years and are lower than the potential yields. In response, the adoption of conservation agriculture (CA), introduced into Tunisia in 1999 to help adaptation to climate change, has resulted in a substantial reduction in agricultural productivity. CA areas increased from 52 ha in 1999 to 14,000 ha in 2015. Using a modelling approach, the present paper evaluates the potential of CA to adapt wheat-based-systems to climate change in Tunisia. The Agricultural Production Systems Simulator (APSIM) model was used to predict the effect of tillage (conventional tillage [CT] vs. zero-tillage [ZT] and soil residue retention [ZT-RR]) on wheat productivity and soil fertility. Two contrasting locations in Tunisia were studied; one semi-arid (Kef) and one sub-humid (Bizerte). Results showed that the sustainable production of durum wheat under climate change conditions in Tunisia is possible through the adoption of CA practices (ZT and ZT-RR) in both sub-humid and semi-arid areas. In fact, mulching (residue retention) is more effective than CT (under semi-arid and sub-humid conditions) in enhancing wheat yield (15%), water use efficiency (18% and 13%) and soil organic carbon accumulation (0.13 t ha-1 year-1 and 0.18 t ha-1 year-1). It is also more effective for soil resilience - preventing water erosion (1.7 t ha-1 year-1 and 4.6 t ha-1 year-1 of soil loss). The present study allowed identification of 260,000 ha as priority areas for CA adoption; this represent one-third of the total cereal area in Tunisia. Appropriate evaluation of the benefits of CA on sustainable agricultural intensification would provide more arguments for effectively supporting CA adoption in Tunisia.

Modelling and evaluating the impacts of climate change on three major crops in south-eastern Australia using regional climate model simulations

The use of regional climate models (RCMs) to localise results from coarse-resolution global climate models has recently attracted more interest in agricultural impact studies. Theoretically, it has advantages over global climate models in terms of realisation of future climate projections. However, there are few studies that have used dynamical downscaling results to assess climate change impacts on Australian cropping systems. In this study, we used post-processing bias-corrected climate data from the NSW/ACT Regional Climate Modelling (NARCliM) project to drive the Agricultural Production Systems SIMulator (APSIM) for modelling and evaluating the response of three major crops (wheat, canola and lupin) to projected climate conditions under various farm management practices in south-eastern Australia. Our results showed that historical crop yields from APSIM simulations forced by RCM output tended to underestimate yields from simulations forced by observations due to large biases in the NARCliM simulations. Therefore, bias correction was used to correct the APSIM outputs before conducting future impact analysis. The bias-corrected results showed that ensemble-mean yields based on 12 RCMs were projected to increase over the study area under the A2 emission scenario. However, the magnitude of yield increase depended on the time periods, crop type and location. Multiple linear regression models showed that the changes of radiation, rainfall and temperature and elevated CO2 concentration could explain 60–78% of crop yield changes. It is interesting to note that residue incorporation and nitrogen application had a large effect on percentage yield increases for wheat due to future climate change, but had limited effects on the response of lupin and canola yields. Our study suggests that sufficient bias-correction method is needed when using NARCliM outputs in crop models. Although uncertainties (e.g. the choice of emission scenarios and RCMs) still exist in our study, the results are of central importance in the development of high-yield adaptive strategies for local farmers and policy makers in south-eastern Australia.

Risks of yield loss due to variation in optimum density for different maize genotypes under variable environmental conditions

AbstractUnderstanding the nature of complex genotype‐by‐environment‐by‐management interactions is crucial to identify risks and opportunities for increasing maize yield and profitability in rainfed production systems. The objectives of this study were to (i) define the conditions where hybrids of different maturity and plant densities are viable options in terms of improving productivity, and (ii) quantify the risk levels associated with different genotype‐by‐management combinations in relation to target environments. Responses to plant density were analysed on field experimentation with different genotypes representing early, medium and late maturity types and 2, 4 and 6 plant/m2 plant densities at three major or potential dryland maize production environments in Queensland, Australia. Agricultural Production Systems sIMulator (APSIM)‐Maize module was employed to simulate yield responses and compute the cumulative probability distribution. APSIM simulations suggested that the risk of expecting a yield level less than 2 t/ha increased up to about 17 and 27% for quick and late maturing types, respectively, when density increased to 10 plants/m2 in marginal environments such as Emerald. In relatively better environments, however the risk increased only up to 10% for late hybrids, and 7% for a quick hybrid at 10 plants/m2. In both high and low potential environments, choice of hybrids and plant densities should be based on seasonal weather forecasts to minimize risks and maximize opportunities for higher yields.

The effect of bias correction and climate model resolution on wheat simulations forced with a regional climate model ensemble

ABSTRACTRegional climate model (RCM) simulations are often used with agricultural models to assess the impact of climate change on agriculture. This study assesses the suitability of climate data sets from an RCM ensemble for forcing simulations of wheat yields for New South Wales, Australia, performed using the Agricultural Production Systems sIMulator (APSIM). Differences in yields between APSIM simulations forced with RCM output for the 1990–2009 period and APSIM simulations forced with observations are examined, as are simulated changes in yields between 1990–2009 and 2060–2079. The RCM ensemble downscales four global climate models (GCMs) using three different RCMs, firstly to a horizontal resolution of approximately 50 km, and then to approximately 10 km. All of the RCM simulations, 50 and 10 km, are able to simulate the spatial pattern of yields across the study region. However, some simulations have biases in yields. The largest of these are due to biases in rainfall during the growing season inherited from the GCMs. If the RCM output is bias‐corrected, the largest positive yield biases are reduced because the underlying biases in growing season rainfall are reduced. Simulated future changes in yields are affected because there is a non‐linear relationship between simulated yields and growing season rainfall. Downscaling to 10 km, rather than 50 km, is only beneficial when bias correction is used. The bias‐correction technique used does not eliminate all biases in growing season rainfall and increases some of the smaller biases. Nonetheless, because of the potentially large effect of the rainfall biases on simulated future yield changes, this study supports the use of bias correction in assessments of the impacts of climate change that use RCM output to force APSIM. Indeed, our findings suggest that bias correction may be necessary to obtain reliable future changes in other outputs of other biophysical models that respond non‐linearly to climate inputs.

Water use efficiency and crop water balance of rainfed wheat in a semi-arid environment: sensitivity of future changes to projected climate changes and soil type

Wheat production is expected to be affected by climate change through changing components of the crop water balance such as rainfall, evapotranspiration (ET), runoff and drainage. We used the Agricultural Production Systems Simulator (APSIM)-wheat model to simulate the potential impact of climate change on field water balance, ET and water use efficiency (WUE) under the SRES A2 emissions scenario. We ran APSIM with daily climate data statistically downscaled from 18 Global Circulation Models (GCMs). Twelve soil types of varying plant available water holding capacity (PAWC) at six sites across semi-arid southeastern Australia were considered. Biases in the GCM-simulated climate data were bias-corrected against observations for the 1961–1999 baseline period. However, biases in the APSIM output data relative to APSIM simulations forced with climate observations remained. A secondary bias correction was therefore performed on the APSIM outputs. Bias-corrected APSIM outputs for a future period (2021–2040) were compared with APSIM outputs generated using observations for the baseline period to obtain future changes. The results show that effective rainfall was decreased over all sites due to decreased growing season rainfall. ET was decreased through reduced soil evaporation and crop transpiration. There were no significant changes in runoff at any site. The variation in deep drainage between sites was much greater than for runoff, ranging from less than a few millimetres at the drier sites to over 100 mm at the wetter. However, in general, the averaged drainage over different soil types were not significantly different between the baseline (1961–1999) and future period of 2021–2040 (P > 0.05). For the wetter sites, the variations in the future changes in drainage and runoff between the 18 GCMs were larger than those of the drier sites. At the dry sites, the variation in drainage decreased as PAWC increased. Overall, water use efficiency based on transpiration (WUE_T) and ET (WUE_ET) increased by 1.1 to 1.6 and 0.7 to 1.3 kg ha−1 mm−1, respectively, over the baseline historical climate. Significant relationships between changes in wheat yield and PAWC were only seen at three sites. At the dry sites, the impact of a future climate under a soil of high PAWC was less than that under one of low PAWC. Conversely, the opposite response was seen at two wetter sites, highlighting the importance of PAWC and rainfall in determining the interactive response of crops to primary components of the water balance.

The use of uncorrected regional climate model output to force impact models: a case study for wheat simulations

Computationally-expensive regional climate models (RCM) are increasingly being used to generate local climate data for climate change impact studies. These studies usually pro- cess RCM output to remove errors in the simulated climate. However, this paper investigates the suitability of raw output from simulations of a single RCM as input to a biophysical impact model. Our study analyses errors in wheat yields simulated for New South Wales (NSW), Australia by the Agricultural Production Systems Simulator (APSIM) model forced with output from 2 RCM simu- lations with horizontal resolutions of approximately 50 and 10 km over NSW and with output from the global climate model (GCM) simulation that they downscale. Overall, across the NSW wheat belt, the ~50 km simulation has a better simulation of mean yields for the 1990�2010 period than the GCM simulation, and the ~10 km simulation has a better simulation than the ~50 km simula- tion. The average mean yield from APSIM simulations forced with observations is 3.5 t ha �1 . The average magnitudes of errors in mean yields for the GCM, ~50 km and ~10 km simulations are 1.2, 1.0 and 0.5 t ha �1 respectively. We suggest that the improvement in the simulation of mean yields with increasing climate model resolution is largely due to an improvement in the simulation of mean rainfall totals for the growing season. However, for a given value of mean growing season total rainfall, all 3 climate model simulations have a climate that is more conducive to high yields than the observed climate. This difference must be due to errors in other aspects of the simulated climates.

Comparison of APSIM and DNDC simulations of nitrogen transformations and N2O emissions

Various models have been developed to better understand nitrogen (N) cycling in soils, which is governed by a complex interaction of physical, chemical and biological factors. Two process-based models, the Agricultural Production Systems sIMulator (APSIM) and DeNitrification DeComposition (DNDC), were used to simulate nitrification, denitrification and nitrous oxide (N2O) emissions from soils following N input from either fertiliser or excreta deposition. The effect of environmental conditions on N transformations as simulated by the two different models was compared. Temperature had a larger effect in APSIM on nitrification, whereas in DNDC, water content produced a larger response. In contrast, simulated denitrification showed a larger response to temperature and also organic carbon content in DNDC. And while denitrification in DNDC is triggered by rainfall ≥5mm/h, in APSIM, the driving factor is soil water content, with a trigger point at water content at field capacity. The two models also showed different responses to N load, with nearly linearly increasing N2O emission rates with N load simulated by DNDC, and a lower rate by APSIM. Increasing rainfall intensity decreased APSIM-simulated N2O emissions but increased those simulated by DNDC.

Describing N leaching from urine patches deposited at different times of the year with a transfer function

A transfer function (TF) was developed to assist with the estimation of nitrogen (N) leaching from urine-affected areas in grazed pastures. The proposed TF uses a simple function to describe the likely breakthrough curve for urine-N deposited in different months and in various climates and soils in New Zealand. The TF was designed to be integrated into the OVERSEER® Nutrient budgets model to increase the sensitivity of N leaching to the month of urine deposition, but could also be used in any other model that estimates the water balance and plant N uptake on a monthly basis. The inputs required for the TF are typically readily accessible (e.g. soil texture data) and thus do not add any significant complications when added to OVERSEER. The TF retains OVERSEER as the arbitrator of the main items of N-balance in the farm system, but adds functionality by giving a better temporal discrimination of leaching from the farm system. The procedure for parameterising the TF from a comprehensive set of APSIM (Agricultural Production Systems Simulator) simulations is described. Validation of the leaching estimated by the TF was achieved through a combination of testing against an independent set of APSIM simulations and testing against experimental data. The testing of the TF showed very promising performance. The TF explained 75% of the variability of N leaching simulated by an independent APSIM dataset and agreed well with the experimental data.

ADDING VALUE TO FIELD-BASED AGRONOMIC RESEARCH THROUGH CLIMATE RISK ASSESSMENT: A CASE STUDY OF MAIZE PRODUCTION IN KITALE, KENYA

SUMMARYIn sub-Saharan Africa (SSA), rainfed agriculture is the dominant source of food production. Over the past 50 years much agronomic crop research has been undertaken, and the results of such work are used in formulating recommendations for farmers. However, since rainfall is highly variable across seasons the outcomes of such research will depend upon the rainfall characteristics of the seasons during which the work was undertaken. A major constraint that is faced by such research is the length of time for which studies could be continued, typically ranging between three and five years. This begs the question as to what extent the research was able to ‘sample’ the natural longer-term season-to-season rainfall variability. Without knowledge of the full implications of weather variability on the performance of innovations being recommended, farmers cannot be properly advised about the possible weather-induced risks that they may face over time. To overcome this constraint, crop growth simulation models such as the Agricultural Production Systems Simulator (APSIM) can be used as an integral part of field-based agronomic studies. When driven by long-term daily weather data (30+ years), such models can provide weather-induced risk estimates for a wide range of crop, soil and water management innovations for the major rainfed crops of SSA. Where access to long-term weather data is not possible, weather generators such as MarkSim can be used. This study demonstrates the value of such tools in climate risk analyses and assesses the value of the outputs in the context of a high potential maize production area in Kenya. MarkSim generated weather data is shown to provide a satisfactory approximation of recorded weather data at hand, and the output of 50 years of APSIM simulations demonstrate maize yield responses to plant population, weed control and nitrogen (N) fertilizer use that correspond well with results reported in the literature. Weather-induced risk is shown to have important effects on the rates of return ($ per $ invested) to N-fertilizer use which, across seasons and rates of N-application, ranged from 1.1 to 6.2. Similarly, rates of return to weed control and to planting at contrasting populations were also affected by seasonal variations in weather, but were always so high as to not constitute a risk for small-scale farmers. An analysis investigating the relative importance of temperature, radiation and water availability in contributing to weather-induced risk at different maize growth stages corresponded well with crop physiological studies reported in the literature.

Use of the APSIM model in long term simulation to support decision making regarding nitrogen management for pearl millet in the Sahel

Soil fertility and climate risks are hampering crop production in the Sahelian region. Because experiments with only a few fertility management options on a limited number of sites and years cannot fully capture the complex and highly non-linear soil–climate–crop interactions, crop growth simulation models may suitably complement experimental research to support decision making regarding soil fertility and water management. By means of a long term (23 years) scenario analysis using the Agricultural Production Systems Simulator (APSIM) model, this study investigates millet response to N in view of establishing N recommendations better adapted to subsistence small-holder millet farming in the Sahel. Prior to this, the APSIM model was tested on a rainfed randomized complete block experiment carried out during the 1994 and 1995 cropping seasons, having contrasting rainfall conditions. The experiment combined, at three levels each, the application of cattle manure (300, 900 and 2700 kg ha−1), millet residue (300, 900 and 2700 kg ha−1) and mineral fertilizer (unfertilized control, 15 kg N ha−1 + 4.4 kg P ha−1 and 45 kg N ha−1 + 13.1 kg P ha−1) at ICRISAT Sahelian Center, Niger. The model suitably predicted plant available water PAW and the simulated water and nitrogen stress were in agreement with measurement (water) and expectation (N) regarding the fertilizer and rainfall conditions of the experiment. APSIM simulations were in satisfactory agreement with the observed crop growth except for the highest crop residue application rates (>900 kg ha−1). For biomass and grain yield, the model performance was relatively good in 1994 but biomass yields were slightly overpredicted in 1995. The model was able to adequately reproduce the average trend of millet grain yield response to N inputs from manure and fertilizer, and to predict the overall observed higher grain yield in 1995 compared to 1994, despite the better rainfall in 1994. The 23-year, long term scenario analysis combining different application rates of cattle manure, millet residue and mineral fertilizer, showed that moderate N application (15 kg N ha−1) improves both the long term average and the minimum yearly guaranteed yield without increasing inter-annual variability compared to no N input. Although it does imply a lower average yield than at 30 kg N ha−1, the application of 15 kg N ha−1 appears more appropriate for small-holder, subsistence farmers than the usual 30 kg N ha−1 recommendation as it guarantees higher minimum yield in worst years, thereby reducing their vulnerability.