Aplicação do modelo aditivo generalizado para analisar a dinâmica da concentração de nutrientes no canal de bombeamento de água para a aquacultura – Chuabo - Dembe, Quelimane

Abstract

This study aims to develop generalized additive models that determine the nutrient concentration dynamics (NH4 +, PO43-, NO3-) for application in the water pumping channel for aquaculture at the Chuabo Dembe in Quelimane. The data were collected during two days corresponding to June 23 and December 13, 2017. The exploratory analysis of the covariates was performed by testing the collinearity, normality and homogeneity of the response variable. GAM was used in program R version 3.5.0 2018 to predict those of [NH4 +], [NO3-] and [PO43]. The GAM selection was based on the elimination of non-significant variables (p-value> 0.05). To validate the GAM, it was based on the AIC (Akaike Information Criterion), GCV (General Cross Validation) and the graphical method. GAM explained the deviation of 99.30%, 97.20% and 90.40% for Ammonia, Nitrate and Phosphate respectively. For [NH4 +], the predictive variables selected by order of significance were: EC, ATM, Time, VC, TS and Tb; for [NO3-]: TS, CE, VC, Time, MOD and Tb; and for [PO43-]: MOD, CE, TS, Tb, Time and VC. From the analysis made it was verified that GAM can be applied to model the nutrient concentration because, preliminary results showed that the nutrient-rich water can be pumped in the afternoon, although the data were not sufficient to obtain a robust conclusion.