In USGS-R/streamMetabolizer: Models for Estimating Aquatic Photosynthesis and Respiration

knitr::opts_chunk$set(echo = TRUE)
options(width=100)

Overview

This vignette demonstrates a few of the options for relating GPP and ER to light and/or temperature.

Setup

Load streamMetabolizer and dplyr.

library(streamMetabolizer)
library(dplyr)

Get some data to work with: here we're requesting three days of data at 15-minute resolution. Thanks to Bob Hall for the test data.

dat <- data_metab('3', '15')

GPP and ER functions

Here's a basic model with GPP proportional to light and ER constant over time.

# the Classic: linear GPP, constant ER (also the default)
mm_classic <- 
  mm_name('mle', GPP_fun='linlight', ER_fun='constant') %>% 
  specs() %>%
  metab(dat)
mm_classic

Here's one where GPP is a saturating function of light. ER is still constant.

# the Saturator: GPP saturating with light, constant ER
mm_saturator <- 
  mm_name('mle', GPP_fun='satlight', ER_fun='constant') %>% 
  specs() %>%
  metab(dat)
mm_saturator

The Saturator produces fitting warnings, which are condensed to 'w' and a summary in the above print-out. They can be inspected in detail by looking directly at the fitted daily parameters:

get_params(mm_saturator) %>% select(date, warnings, errors)

Similary, you can inspect the warnings and errors that arise during prediction by pulling out the daily metabolism predictions (but there aren't any, so those columns are empty):

predict_metab(mm_saturator) %>% select(date, warnings, errors)

You can predict and/or plot instantaneous DO values from the fitted daily parameters.

predict_DO(mm_saturator) %>% head
plot_DO_preds(mm_saturator)

Yep, that fitting warning on day 2 was meaningful! We can encourage the model toward a good fit by adjusting the initial values of Pmax and alpha from which the fitting function should explore likelihood space. There are two ways to do this - as date-specific values in data_daily, or as values that apply to every date in specs(). The two methods can even be combined.

mm_saturator2 <- 
  mm_name('mle', GPP_fun='satlight', ER_fun='constant') %>% 
  specs() %>%
  metab(dat, data_daily=select(get_params(mm_saturator), date, init.Pmax=Pmax, init.alpha=alpha))
get_params(mm_saturator2)
mm_saturator3 <- 
  mm_name('mle', GPP_fun='satlight', ER_fun='constant') %>% 
  specs(init.Pmax=6.2, init.alpha=0.008) %>%
  metab(dat)
get_params(mm_saturator3)
mm_saturator4 <- 
  mm_name('mle', GPP_fun='satlight', ER_fun='constant') %>% 
  specs(init.Pmax=6.2, init.alpha=0.008) %>%
  metab(dat, transmute(get_params(mm_saturator), date, init.Pmax=Pmax[1], init.alpha=alpha[1])[2,])
get_params(mm_saturator4)

Despite the remaining warnings, DO predictions from the saturating GPP-light function do look better than from the classic model for this particular dataset.

plot_DO_preds(mm_classic)
plot_DO_preds(mm_saturator4)

See the full list of available functions for gross primary productivity (GPP) and ecosystem respiration (ER) in ?mm_name.

USGS-R/streamMetabolizer documentation built on Aug. 15, 2023, 7:50 a.m.