Page Banner

United States Department of Agriculture

Agricultural Research Service

Title: Evaluating GPFARM for No-Tillage Dryland Experimental Sites in Eastern Colorado

Authors
item Ascough, James
item McMaster, Gregory
item Andales, Allan
item Hansen, Neil - COLORADO STATE UNIVERSITY
item Ahuja, Lajpat
item Sherrod, Lucretia

Submitted to: Transactions of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: June 13, 2007
Publication Date: October 23, 2007
Repository URL: http://asae.frymulti.com/request.asp?search=1&JID=3&AID=23953&CID=t2007&v=50&i=5&T=1
Citation: Ascough Ii, J.C., Mcmaster, G.S., Andales, A.A., Hansen, N.C., Ahuja, L.R., Sherrod, L.A. 2007. Evaluating GPFARM for No-Tillage Dryland Experimental Sites in Eastern Colorado. Transactions of the ASABE, 50(5):1565-1578.

Interpretive Summary: Exploring alternative agricultural management systems in the semi-arid Great Plains is receiving increasing attention. GPFARM is a farm/ranch decision support system (DSS) designed to assist in strategic planning for land management units up to the whole farm scale. This study evaluated GPFARM simulation model performance across a range of dryland no-till experimental sites in eastern Colorado and compared differences in dry mass grain yield, total soil profile water content, crop residue, and total soil profile residual NO3-N. Evaluation data were collected from 1987 through 1999 from an on-going, long-term experiment at three eastern Colorado locations (Sterling, Stratton, and Walsh) along a potential evapotranspiration gradient. Relative error (RE) of simulated mean, root mean square error (RMSE), and index of agreement (d) model evaluation statistics were calculated to evaluate model performance. A one-way, fixed effect ANOVA was used to test the effect of location on both measured data and GPFARM simulation model output responses. GPFARM simulations generally agreed with observed trends and showed the model was able to simulate location differences for the majority of model output responses. GPFARM appears to be adequate for use in strategic planning of alternative cropping systems across eastern Colorado dryland locations; however, further improvements in the crop growth and environmental components (including improved parameterization) of the simulation model should also improve its applicability for strategic planning scenarios.

Technical Abstract: Exploring alternative agricultural management systems in the semi-arid Great Plains is receiving increasing attention. GPFARM is a farm/ranch decision support system (DSS) designed to assist in strategic planning for land management units up to the whole farm scale. Important components include simulation of crop production, economic and environmental impact analysis, and site-specific database generation, from which alternative agricultural management systems can be evaluated and compared. This study evaluated GPFARM simulation model performance across a range of dryland no-till experimental sites in eastern Colorado to assess regional applicability and efficacy in comparing differences in dry mass grain yield, total soil profile water content, crop residue, and total soil profile residual NO3-N. Evaluation data were collected from 1987 through 1999 from an on-going, long-term experiment at three locations in eastern Colorado along a gradient of low (Sterling), medium (Stratton), and high (Walsh) potential evapotranspiration. Simulated crops were winter wheat (Triticum aestivum L.), corn (Zea mays L.), sorghum (Sorghum bicolor L.), proso millet (Panicum miliaceum L.), and fallow. Relative error (RE) of simulated mean, root mean square error (RMSE), and index of agreement (d) model evaluation statistics were calculated to adequately evaluate model performance since each statistic can give different indications of model performance. Additional statistical analysis was performed to determine interactions among locations with a one-way, fixed effect ANOVA used to test the effect of location on both measured data and GPFARM simulation model output responses. For overall model performance, the ranges (best to worst) of RE, RMSE, d were: grain yield (RE: -3% to 35%, RMSE: 495 to 2188 kg ha-1, d: 0.81 to 0.13); total soil profile water content (RE: 6% to 8%, RMSE: 54 to 61 mm water, d: 0.79 to 0.71); crop residue (RE: -4% to 32%, RMSE: 1530 to 2099 kg ha-1, d: 0.76 to 0.62); and total soil profile residual NO3-N (RE: -14% to 25%, RMSE: 32 to 52 kg ha-1, d: 0.69 to 0.45). For trend analysis (magnitudes and location differences), GPFARM simulations generally agreed with observed trends and showed the model was able to simulate location differences for the majority of model output responses. GPFARM appears to be adequate for use in strategic planning of alternative cropping systems across eastern Colorado dryland locations; however, further improvements in the crop growth and environmental components (including improved parameterization) of the simulation model should also improve its applicability for strategic planning scenarios.

Last Modified: 12/22/2014