Study on Phytoplankton variability of Sundarban Estuarine river using Random Forests Classifier

Abstract

Statistical Classifier Random Forests (RF) is nowadays extensively used by ecologists for accurate classification, determination of variable importance, understanding of complex interactions in ecosystem studies. In the present study, extensive observation data collected from a river of SES, Sundarban Estuarine System (River Jagaddal, situated in the southernmost part of western Sundarbans and in the closed proximity to Northern boundary of Bay of Bengal) has been used to perform rigorous data analysis using Random Forests. The basic objective of the study is to identify variability, importance and associated interactions for phytoplankton as well as chlorophyll concentrations along the river stretch. This study enables us to identify status of productivity pertinent to availability as well as growth of fish production of this region. It may be noted that most of the people residing in these areas have sole dependency of earning from fishing. RF model has a high predictive power. The interpretations of RF model can be visualized by feature importance graphs which evaluate a feature as a whole which means, contribution of all the properties to a certain phenomena can be observed. In this approach, Random Forest (RF) classification algorithm has been used to classify different properties of estuarine water according to their individual roles on the phytoplankton density and chlorophyll-a concentrations. The properties are classified as significant, moderately significant and insignificant on the basis of their roles (as obtained from RF model) in modulating phytoplankton density and chlorophyll-a concentration. The present study reveals that the phytoplankton density is strongly influenced by the distance from the sea. Surface salinity is the other important factor to be considered as per our findings. Surface phosphate and surface nitrate are marked to be the other important dominating factors on phytoplankton density. In fact, nitrogen and phosphorus are the primary factors that control phytoplankton abundance in estuaries. The bottom temperature has little significance on phytoplankton density as per model investigation. But, the roles of the other factors like surface temperature (st), bottom phosphate (bp) and bottom salinity (bs) have found to be literally insignificant. The chlorophyll-a concentrations found in the study area during study period are found to be strongly correlated to phytoplankton density. This is quite expected since chlorophyll-a concentration can be used as a direct measure of phytoplankton density. The surface phosphate and bottom salinity also observed to have significant influences on chlorophyll-a concentration. It can also be concluded from the model output that, surface phosphate is the most important limiting nutrient on the chlorophyll-a concentration in this specific study area. Usually, there is a general consensus that there are seasonal and spatial variations of the limiting nutrients which in turn are the most important factors responsible for variation in phytoplankton density. Phosphorus is known to be the principal limiting nutrient of phytoplankton growth. Phosphorus limitation is often associated with periods of high river runoff whereas, N or N+P limitation was associated with low river runoff, with comparatively greater influence of sea water. The contribution of other parameters like distance from the sea (d), surface temperature, bottom temperature, bottom phosphate, surface nitrate, bottom nitrate and surface salinity are observed to have more or less insignificant effect on chlorophyll-a concentrations.