Premium
Partial Least Squares Regression for Determining Factors Controlling Winter Wheat Yield
Author(s) -
Hu Yutong,
Wei Xiaorong,
Hao Mingde,
Fu Wei,
Zhao Jing,
Wang Zhe
Publication year - 2018
Publication title -
agronomy journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.752
H-Index - 131
eISSN - 1435-0645
pISSN - 0002-1962
DOI - 10.2134/agronj2017.02.0108
Subject(s) - partial least squares regression , agronomy , fertilizer , yield (engineering) , nutrient , mathematics , precipitation , regression analysis , sowing , crop yield , biology , statistics , ecology , geography , materials science , metallurgy , meteorology
Core Ideas Importance of factors on wheat yield was tested by partial least squares regression. Nitrogen fertilizer was the most important factor on wheat yield in all four groups. Climate factors, precipitation, and soil nutrients were also major control factors. Partial least squares regression is a useful tool to reveal the control factors on wheat yield.Wheat ( Triticum aestivum L.) yield is influenced by many independent factors including precipitation, fertilization, soil nutrients, and crop variety. Due to high correlations of these factors, it is difficult to analyze their relative importance on wheat yield. This study quantified the effects of independent factors on wheat yield and identified the most important control factors through a long‐term experiment on the Loess Plateau, China. The experiment consisted of 17 treatments, including five different levels of N and P fertilizer. Partial least squares regression (PLSR) was used to evaluate the factors on wheat yield in four variety groups‐ Qinmai4 (1985–1986), Changwu131 (1987–1996), Changwu134 (1997–2015), and 31‐yr planting across the three varieties (1985–2015). Variable importance in projection (VIP) value revealed that N fertilizer had the greatest effect on wheat yield in all four groups (VIP = 1.266–2.313). The second most important factors were climate factors for Qinmai4 (VIP = 1.060), precipitation (February, annual, and fallow season) for Changwu131 ( W 1 = 0.335–0.351, VIP = 1.381–1.474), and soil nutrients (total nitrogen [TN], soil organic matter [SOM], and available potassium [AK]) for Changwu134 ( W 1 = –0.231–0.514, VIP = 1.084–2.317). When tested across varieties, TN and SOM were the second most important factors for 31‐yr planting ( W 2 = 0.455 and 0.313; VIP = 1.908 and 1.370, respectively). These results indicate that PLSR can reveal the control factors on wheat yield in the study area and provide a reference tool for analyses in other crops or areas.