Data from: Evaluating the Agricultural Production Systems sIMulator (APSIM) wheat module for California
Data files
Apr 17, 2024 version files 10.55 MB
Abstract
Context: Computer-based crop simulation models are important tools for agricultural research and management. The Agricultural Production Systems sIMulator (APSIM) 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, and to identify future research needs, by using pre-existing data from state-wide variety trials.
Methods: Environmental and management data from three seasons of state-wide wheat variety trials, were used to parameterize the APSIM-Wheat module (version 7.10 r4220). Simulated yield and protein data were compared to and actual field data to test the reliability of the 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.
Conclusions: The accuracy of the simulations was comparable to other tests of the APSIM-Wheat module in environments where it has not been previously calibrated, but was too low to be considered reliable. The lack of reliability was due to the poor representation of local Californian wheat genotypes, as well as inaccuracy of management and environmental data.
Implications: APSIM could be a valuable tool for wheat research and management in California, but our work shows the current model is unreliable. Further research is needed to generate field data needed for model calibration.
README: Evaluating the Agricultural Production Systems sIMulator (APSIM) wheat module for California
https://doi.org/10.5061/dryad.qv9s4mwj0
Combined.Field.APSIM.data.csv
This dataset includes the combined yield and grain protein data from state-wide variety trials of wheat conducted by the University of California (UC) Small Grains Program seasons from 2016-17 to 2018-19 and summary statistics comparing APSIM model predictions of the field trials with field data.
Description of the data and file structure
Three files are included in this dataset:
Combined.data.with.predictors.Jan.27.2023.csv
These comprised twelve field locations in the main cereal production regions of California, conducted over three winter seasons from 2016-17 to 2018-19. A total of thirty-four environments (location-by-management-by-year combinations) were sampled. The field locations fall between latitudes 33o to 42o N, a north-south distance of approximately 800 kilometres. Standard deviations among plots for each test location are also reported. The APSIM predictions for grain yield and protein for the environments are reported. All Yields are reported in kg/ha.
Explanation of column headings: LOCATION - Field trial location, YEAR - year in which the trial was harvested, weather - the name of the weather station used as a source of summary climate information, APSIM.cultivar - the name of the APSIM cultivar used to simulate the field trial, APSIM.yield - the yield (kg/ha) predicted for the location by APSIM, Harvest - the date the field study was harvested, APSIM.protein - the grain protein content (%) predicted by APSIM, Mat.DAS - the days to maturity after sowing, Field.cultivar - the numeric code of the wheat cultivar growing in the field trial, Field.yield - the measured mean yield (kg/ha) for the cultivar grown in the field trial, Field.protein - the mean grain protein content (%) measured in the field trial, environment - the combined location name and harvest year, Name_F - the name of the wheat cultivar growing in the field trial, PROTEIN.sd - the standard deviation for measured grain protein content across replicate plots in the field trial, YIELD.sd - the standard deviation for measured grain yield across replicate plots in the field trial.
kgha - the measured mean yield (kg/ha) for the cultivar grown in the field trial, kgha.sd - the standard deviation of yield across replicated plots in the field trial, Num.Years - the number of seasons the variety was grown in the field trials, difference - the difference between the actual and predicted yield at the location, grains_per_gram_stem - Kernel number per stem weight at the beginning of grain filling (g), potential_grain_filling_rate - predicted grain fill filling rate (g/day), max_grain_size - The maximum grain size (g) in the simulation, tt_end_of_juvenile - Thermal time from sowing to end of juvenile (°Cdays), tt_floral_initiation - Thermal time from floral initiation to flowering (°Cdays), tt_flowering - Thermal time for anthesis phase (°Cdays), tt_start_grain_fill - Thermal time from beginning to end of grain filling (°Cdays), tt_end_grain_fill - Thermal time from end of grain fill to maturity (°Cdays), startgf_to_mat - grain filling duration in degree days, °C, vern_sens - Vernalization sensitivity (0 lowest – 5 highest), photop_sens - Sensitivity to photoperiod. (0 lowest – 5 highest), node_sen_rate - The rate of node senescence on the main stem. (°Cdays/node), Initial.water - Plant available soil water at sowing (mm), Rain - Total precipitation during the growing season (mm), Irrigation - Total irrigation applied during the growing season (mm), GDD - Total growing degree days during the growing season (°Cdays), Lat - Latitude of the field site, Soil.type - Approximate soil texture of the field site, Soil.N - Starting soil nitrogen (kg/ha), Sowing.N - Nitrogen applied as fertiliser at sowing (kg/ha), Top.dress.N - Nitrogen applied as fertilisers after sowing (kg/ha).
Supplemental Material Protein 2.csv & Supplemental Material Yield 1.csv
These files are the summary statistics comparing the APSIM predictions with field data. Summary data is available for all unique combinations of APSIM cultivar and field genotype. The nature of the linear relationship between the simulated and field data was primarily assessed using the root mean square error (rmse), and secondarily with the coefficient of determination (R2), slope and intercept of the linear relationship. For comparison to published literature, the root mean square error was also normalized (nrmse).
Explanation of column headings: APSIM Cultivar - the name of the APSIM cultivar used to simulate the crop production, Field variety - the numeric code of the wheat cultivar growing in the field trial, Number of environments - the number of unique location-year combinations in the pair-wise comparison between actual and predicted values, r2 - root mean square error, Adjusted r2 - adjusted root mean square error, slope - slope of the linear fit, intercept - y-axis intercept of the linear fit, rmse - the root mean squared error of the linear fit, mae - mean absolute error of the linear fit, nrmse - the normalised root mean squared error of the linear fit.