Spearman's rank correlation analysis of phenology and PCA weather scores
I performed a PCA with the weather variables to get components which would reflect best combinations of the weather conditions. According to the component importance graph (Figure 18), the first 5 explain most of the variation in the dataset – 88%. Components 1 through 5 correspond to El Niño (negative correlation), cold and clear conditions, hot and dry conditions, El Niño (positive correlation), and dry and clouded conditions, respectively. The Spearman’s rank correlation analysis in figure 19 reveals that the flowering in different regions has stronger correlations with different weather patterns. Sites in all of Peninsular Malaysia correlate with the onset of drought (component 3) where West Borneo correlates with El Niño. The strongest negative correlation in East Borneo is with component 3 of hot and arid conditions. This suggests that flowering does not occur in dry/hot conditions in East Borneo. This is an unexpected result. I was expecting to find a correlation with El Niño in East Borneo as the correlation between El Niño and weather was the strongest there(Figure 11).
Figure 18. Standard deviation and cumulative variance explained for principle component analysis of weather trends for sites
Figure 19. Spearman's rank correlation coefficients for flowering and principle components
Classification of flowering cues by regions based on weather and El Niño using CART
Classification tree analysis suggests similar trends as from PCA. CART gives the thresholds of variables where the main partitions are made. It also gives out alternative partitions at each node marked in parenthesis in figures 20-24. Almost in every case lower values of precipitation resulted in occurrence of flowering. However the thresholds for regions are different. In Borneo, aside from impact of El Niño on flowering, colder and clearer conditions seem to be present as well. In Malaysia drop in temperatures, clear and dry conditions influence flowering. From this we can conclude that drought is the factor that influencing flowering in Malaysia, Borneo, Indonesia, and the Philippines. CART also gives us the various thresholds for the variables in different regions.
Figure 20. Classification tree for East Borneo (EB)
Figure 21. Classification tree for West Borneo
Figure 22. Classification tree for North Peninsular Malaysia
Figure 23. Classification tree for South Peninsular Malaysia
Figure 24. Classification tree for Singapore region