Turkish Journal of Agriculture and Forestry




Broad interest in reducing greenhouse gas emissions requires a better understanding of controls on carbon dioxide (CO_2) release under different agricultural management practices. The objective of this study was to investigate and model seasonal variation of soil CO_2 emissions from an apple orchard field (Malus domestica L. 'Starkrimson'). Soil CO_2 emissions from an apple orchard managed with common practices were measured weekly over a 3-year period (May 2008 to May 2011) from both under the crowns of trees (CO_2-UC) and between rows (CO_2-BR) using a soda lime technique and were modeled using available environmental data. The study area is located in the Harran Plain of southeastern Turkey and has a semiarid climate. The weekly soil CO_2 emissions ranged from 87.8 to 1428 kg week^{-1} ha^{-1}, from 74.6 to 835 kg week^{-1} ha^{-1}, and from 88.6 to 1087 kg week^{-1} ha^{-1} for CO_2-UC, CO_2-BR, and the average of both (CO_2-AVG), respectively, and showed a pronounced seasonal pattern with the lowest emissions in winter (January and February) and the highest emissions during the growing season (April to December). Relative to 2008 emissions, 2009 CO_2 emissions increased by approximately 75%, and 2010 emissions increased by approximately 88%. Comparison of 3 models (multiple linear regression, principal component regression, and multivariate adaptive regression splines) showed that multivariate adaptive regression splines provided the best performance in modeling soil CO_2 emissions, explaining overall variation of 64%, 56%, 76%, and 53% in CO_2-AVG for the first, second, third, and all three 3 periods, respectively. In conclusion, overall findings showed that soil CO_2 emissions could be modeled by available environmental data such as air and soil temperature.


Soil CO_2 emission, multivariate statistical analysis, principal component analysis, principal component regression, multivariate adaptive regression splines

First Page


Last Page