Conservation Tillage Effects on European Crop Yields: A Metadata Analysis
Description
We were hypothesizing that ridge-till and strip-till can improve crop yields under European agroecosystems in contrast with the often observed reduction in yields under No Tillage and Reduced Tillage.To answer the hypothesized question, and based on our data, it can be stated that yields are indeed not optimal under no till as opposed to RT and ST where yields are higher than those of conventional tillage. The raw data shows every parameter and data extracted from peer reviewed articles while the analyzed shows the data obtained from running the meta-analysis by fitting random and mixed effect models. Raw data Here, the data consists of all parameters and variables extracted from the peer-reviewed articles that met the inclusion criteria. Also, there is a metadata sheet with all abbreviations and their meaning. Analyze data Here, the estimate, lower (ci.lb) and upper (ci.ub) confidence intervals are in log transform format. Thus, taking the exponent gives the response ratio. Se =standard error, zval= z correlation, pval =p value, RT (ridge-till) and RH (number) = ridge height. C =Cotton, Cr = Carrot, RD = Dandellion, T=Tulips, GM= grain maize, WC= winter cereal, SB =sugar beet, SC=spring cereal, P =potato. Also, there is a metadata sheet with all abbreviations and their meaning. Graph Fig. 1. Forest plot showing data plot of estimated average response ratios (RR) for No (NT), Ridge (RT) and Strip (ST) tillage based on random effects model. n is sample size/number of studies for the specific tillage type. The error bars represent the 95% confidence intervals CI. The square shapes represent the estimated RR and their size correspond to their weight (based on its sample size). The vertical broken line represents the line of no effect
Files
Steps to reproduce
We commenced by defining our research question and setting the criteria for which an article will be considered. These conditions included; i) A study needed to include field experimentation with side by side comparisons of one or more conservation tillage practices of no (NT), strip (ST) and ridge (RT) tillage relative to conventional (CT) tillage (as control) with the latter using mouldboard or disc ploughing and based on maximum soil disturbance; ii) the studies must have been performed on experimental sites in Europe (using the United Nations definition of geographical Europe) with locations mentioned; iii) the studies must have compared crop yields as one major component under investigation; iv) the studies must have been documented in English. Web of Science was employed for the literature search (search terms used were: ridge* OR strip* OR mulch* OR no* OR direct drill* OR conservation OR till* in combination with the different European states). Other publications like conference papers, notes, meeting abstracts, proceeding papers, book chapters, letters and editorial material were thus eliminated. The majority of the studies were easily accessible on the Elsevier publishing site through our institutional (Ghent University) access. This service also made it possible to access articles from other publishers like Cambridge University Press, Wiley and Sons, Taylor and Francis, among others. Some studies were not freely accessible and effort was made to contact the authors for the provision and possible clarifications by sending emails and using the request option on publishing sites like ResearchGate. Zotero 5.0.88 (https://www.zotero.org/) was employed for quick study retrieval, identifying duplicated studies, creating reference lists, and bibliographies. Data mining and entry were done using the Microsoft Excel package. Plot Digitizer (http://plotdigitizer.sourceforge.net/) was used to extract data values present in the form of charts. FAO New_LocClim (http://www.fao.org/land-water/databases-and-software/en/) tool was used to access climatic data for those studies that did not report them. R software environment (https://cran.r-project.org/) was used for the analysis. In order to compute the effect size, the response ratio (RR) of annual mean NT, ST, RT yield to mean CT yield was used. Majority of the studies included did not report standard deviations thus we followed the procedure of Nunes et al. (2020) whereby, 1/10 of each mean yield was taken as the standard deviation. Together with the sample sizes, the data was then ready to call on the mefor package. From this stage, the metafor package was utilized for the meta-analysis. Starting out by fitting a random model to get the estimates and then proceeding to a meta-regression by fitting a mixed effect model and including the different variables. The output from fitting the models are presented in the analysed data table.