R Notebook

How senstive are the fits to differnt values of sigma? (simga from the mesa function that is ues to define the cmip range for heat flux).

library(dplyr)
library(tidyr)
library(ggplot2)
library(hector)
library(hectorcal)
# How much does changning the sigma value change the param results??? 
BASE_DIR <- "/Users/dorh012/Documents/2019/hectorcal"
BESTFIT_DIR <- file.path(BASE_DIR, 'analysis', 'best_fit')
best_fits_og       <- read.csv(list.files(file.path(BESTFIT_DIR, 'bestFit_conc_tempAll_heatFutRange'), '.csv', full.names = TRUE))
best_fits_og$sigma <- '.15'
best_fits_half       <- read.csv(list.files(file.path(BESTFIT_DIR, 'bestFit_conc_tempAll_heatFutRange_sigmaHalf'), '.csv', full.names = TRUE))
best_fits_half$sigma <- '0.075'
best_fits_doubble       <- read.csv(list.files(file.path(BESTFIT_DIR, 'bestFit_conc_tempAll_heatFutRange_sigmaDoubble'), '.csv', full.names = TRUE))
best_fits_doubble$sigma <- '.3'
best_fits <- bind_rows(best_fits_og, best_fits_half, best_fits_doubble) 

ggplot(data = best_fits) + 
    geom_point(aes(S, model, color = sigma), size = 2, alpha = 0.7) + 
    labs(y = NULL, 
         x = 'S deg C',
         title = 'S values calibrated with the range method + different sigma heat flux values')

Note for best fits run with sigma = 0.075 not all of the runs worked. So not every model had a blue dot. BUt for some models there is considerable spread in the S values based on the sigma is definfed, I thiunk that half a degree in S a fair a mount. For the most part, although not always the case the S values solved with sigma = 0.3 had smaller values of S compared to the runs that used simga = 0.15.

ggplot(data = best_fits) + 
    geom_point(aes(diff, model, color = sigma), size = 2, alpha = 0.7) + 
    labs(y = NULL, 
         title = 'diff values calibrated with the range method + different sigma heat flux values')

The best fit value of diff changes by a small amount with respect to sigma but it is not consistent across models.

ggplot(data = best_fits) + 
    geom_point(aes(alpha, model, color = sigma), size = 2, alpha = 0.7) + 
    labs(y = NULL, 
         title = 'S values calibrated with the range method + different sigma heat flux values')

Alpha was not impacted much by the change in sigma which is not surprising.

ggplot(data = best_fits) + 
    geom_point(aes(volscl, model, color = sigma), size = 2, alpha = 0.7) + 
    labs(y = NULL, 
         title = 'S values calibrated with the range method + different sigma heat flux values')

volscl best fit values did not change that much with sigma.

What happen if we run the Hector cores with the best fit values?? Using the best fits from MIROC-ESM, MIROC-ESM is one of the models that has a higher than expected S value and happens to fall on the outter edge of the heatflux cmip range envlope. I would expect that increasing the size of simga would mean that we would allow heat flux to increase even more.

The paramterizations did not make much of a difference for Tgav which is not suprising. But it did for future heat flux. But I would have expected that the larger sigma would have had a laxer heat flux edges therefore allow for the heatflux to increase. However this is not the case. Does it have anything to do with the fact that we are looking at the -log of the mesa function? So relaxing the edges of the cmip envolope would happen with smaller values of sigma compared to the larger ones?