Plant growth data analysis is done in two ways: one where growth curves are generated and another where formal analysis following the same approach as harvest yield was carried out.
Here, dose response curve or drc package in R is used according to the below line of code to generate a graph as shown. The shape of the growth curves provides insight in nutrient effects in all stages of crop growth. In Figure 8 for example, N becomes more limiting than P only after about 50 days since planting. The full R script (DoseResponse.r) for these analyses is available here.
drm(Biovolume~DAP,curveid=TrtDesc,fct=LL.3(),data=site,na.action=na.omit)# DAP=days after planting
Figure 8. Plant growth as observed in Kiberashi Tanzania in 2010.
Plant growth data is used to determine treatment effects in the same way as yield. Here, since no biovolume data is available (due to basal diameter not being collected) for the two sorghum sites i.e., Koloko and Kontela, only maize sites are included in the plant growth analysis. The response variable of interest for the crop growth data is the maximum biovolume for each plot irrespective of the time point at which the maximum biovolume was observed. The analysis focuses on the difference between the maximum biovolume for each plot and the average maximum control biovolume (averaged across the two control replicates). The R scripts for this analysis are contained in the “Maximum Biovolume Analysis.r” available here which, as in the case of yield, calls the function code “SiteMixed_resent.r” and executes the models presented earlier. Histograms of the response are first produced to assess the suitability of modeling on the original scale (the difference scale). Having assessed the histograms, the modeling was conducted for every site, and the residuals were checked to further examine this assumption. For three of the sites namely Kasungu, Pampaida and Mbinga, data were not normally distributed and it was decided to refit the model on the difference between the arithmetic log maximum biovolume and the arithmetic log average maximum control biovolume (taking the difference of two logs is the equivalent of working on the log ratio scale). The average maximum control biovolumes have also been included in the models to adjust for any ‘control’ effect; this term has been removed from models where there was no evidence of an effect (i.e., P value <0.05). Figure 9 shows the effects of treatment on maximum biovolume in Mbinga, Tanzania.
Figure 9. Effect of treatment on plant biovolume in Mbinga, southern Tanzania
Considering both harvest yield and biovolume is important and can reveal important information. In Kasungu for example, we found that applications of multinutrients over NPK did not have positive effect on biovolume, unlike grain yield. The significant effect of this treatment on yield alone is related to cobs and grains that were heavier compared to other treatments.