Keywords

gams, gamms, water quality, nonlinear regression

Start Date

1-7-2010 12:00 AM

Description

An on-going challenge for decision makers is the interpretation of temporal trends frommonitoring data given that environmental processes often generate complex data that aremultivariate and potentially nonlinear. Generalized additive models (GAMs) is a wellsuitedmodelling framework for uncovering such trends and unifying datasets. Thisapproach allows flexible specification of regression splines to represent the functionalrelationships between a response variable (the parameter of interest) and a suite of temporaland spatial covariates that can be continuous or discrete using a link function and smoothfunctions of the covariates. We highlight the utility of using GAMs through three casestudies. The first highlights the use of a GAM to unify the findings of an established longtermwater quality-monitoring program with those of a focused short-term monitoringprogram. In the second, a GAM is used to evaluate the spatial patterns in a biomonitoringdataset whilst simultaneously accounting for variability in oyster size, which can have aconfounding effect on such data. The final case study focuses on a 12 month continuousmonitoring program of oceanographic data as part of an evaluation of the environmentalconditions for a desalination plant intake pipe. The context for these studies ispredominantly water quality in the coastal zone, however the benefits and widespreadapplication to other research areas is clearly evident.

Share

COinS
 
Jul 1st, 12:00 AM

Using Generalized Additive Models to Assess, Explore and Unify Environmental Monitoring Datasets

An on-going challenge for decision makers is the interpretation of temporal trends frommonitoring data given that environmental processes often generate complex data that aremultivariate and potentially nonlinear. Generalized additive models (GAMs) is a wellsuitedmodelling framework for uncovering such trends and unifying datasets. Thisapproach allows flexible specification of regression splines to represent the functionalrelationships between a response variable (the parameter of interest) and a suite of temporaland spatial covariates that can be continuous or discrete using a link function and smoothfunctions of the covariates. We highlight the utility of using GAMs through three casestudies. The first highlights the use of a GAM to unify the findings of an established longtermwater quality-monitoring program with those of a focused short-term monitoringprogram. In the second, a GAM is used to evaluate the spatial patterns in a biomonitoringdataset whilst simultaneously accounting for variability in oyster size, which can have aconfounding effect on such data. The final case study focuses on a 12 month continuousmonitoring program of oceanographic data as part of an evaluation of the environmentalconditions for a desalination plant intake pipe. The context for these studies ispredominantly water quality in the coastal zone, however the benefits and widespreadapplication to other research areas is clearly evident.