BloomSail_surface.Rmd

---
title: "MLD"
author: "Jens Daniel Müller"
date:  "`r format(Sys.time(), '%d %B, %Y')`"
output: 
  workflowr::wflow_html:
    number_sections: true
    toc_depth: 3
    toc_float:
      collapsed: false
editor_options:
  chunk_output_type: console
---


```{r global_options, include = FALSE}
knitr::opts_chunk$set(warning=FALSE, message=FALSE)
```

```{r packages}
library(tidyverse)
library(seacarb)
library(oce)
library(marelac)
library(patchwork)
```


```{r ggplot_theme, include = FALSE}
theme_set(theme_bw())
```

# Approach

In order to test how (and how well) the depth-integrated NCP estimates can be reproduced if only surface CO2 data were available, the BloomSail observations were restricted to those made in surface water and two reconstruction approaches were tested:

1. Integration of surface observation across the MLD, assuming homogenious vertical patterns
2. Vertical reconstruction of incremental CT changes based on profiles of incremental changes in temperature

# Sensor data

1m gridded, downcast profiles were used.

```{r read_Tina-V_Sensor_data}

ts_profiles_ID <-
  read_csv(here::here("Data/_merged_data_files", "ts_profiles_ID_long_cum_MLD.csv"))

ts_profiles_ID <- ts_profiles_ID %>% 
  select(ID, date_time_ID, dep, CT = value, CT_diff = value_diff, CT_cum = value_cum, sign, rho_lim, MLD)

ts_profiles_ID <- ts_profiles_ID %>% 
  mutate(CT = if_else(dep == 3.5, CT, NaN),
         rho_lim = as.factor(rho_lim))

```


# MLD approach

MLD calculation was previously described in the [CT dynamics](https://jens-daniel-mueller.github.io/BloomSail/CT_dynamics.html) chapter.

## Timeseries

For comparison to MLD, the effective penetration depth of NCP, z~eff~, was calculated as the ratio of the inremental, depth-integrated change of C~T~, divided by the change in surface C~T~, for all cases where the change in surface C~T~ was negative.

```{r NCP_penetration_depth}

ts_profiles_ID_surface <- ts_profiles_ID %>% 
  filter(sign=="neg",
         rho_lim == "0.5",
         dep == 3.5) %>% 
  group_by(ID, date_time_ID) %>% 
  summarise(CT_diff_surf = mean(CT_diff, na.rm = TRUE)) %>% 
  ungroup()

ts_profiles_ID_i <- ts_profiles_ID %>% 
  filter(sign=="neg",
         rho_lim == "0.5") %>% 
  group_by(ID, date_time_ID) %>% 
  summarise(CT_diff_i = sum(CT_diff, na.rm = TRUE))

zeff <- inner_join(ts_profiles_ID_surface, ts_profiles_ID_i)
rm(ts_profiles_ID_surface, ts_profiles_ID_i)

zeff <- zeff %>% 
  mutate(zeff = CT_diff_i / CT_diff_surf)

```



```{r MLD_time_series, fig.asp=0.4}

ts_profiles_ID %>% 
  ggplot()+
  geom_hline(yintercept = 0)+
  geom_line(data = zeff, aes(date_time_ID, zeff, linetype="zeff"))+
  geom_line(aes(date_time_ID, MLD, col=rho_lim, linetype="MLD"))+
  scale_y_reverse()+
  scale_color_viridis_d(name="Rho limit")+
  scale_linetype(name="Estimate")+
  labs(y="Depth (m)")+
  theme(axis.title.x = element_blank())

rm(zeff)

```

## iC~T~ calculation

Integrated C~T~ depletion was calculated as the product of observed incremental C~T~ changes in surface waters and the respective mixed layer depth. 

```{r NCP_calculation, eval=FALSE}

ts_profiles_ID_surface <- ts_profiles_ID %>% 
  filter(dep==3.5) %>% 
  group_by(rho_lim) %>% 
  arrange(date_time_ID) %>% 
  mutate(iCT_diff = CT_diff * MLD / 1000,
         iCT_cum = cumsum(replace_na(iCT_diff, 0))) %>% 
  ungroup()
  
```


## Incremental and cumulative timeseries

Total incremental and cumulative C~T~ changes inbetween cruise dates were calculated.


```{r plot_integrated_NCP_timeseries, fig.asp=1.5, eval=FALSE}

p_iCT <- ts_profiles_ID_surface %>% 
  ggplot(aes(date_time_ID, iCT_diff, fill= rho_lim))+
  geom_hline(yintercept = 0)+
  geom_col(col="black", position = "dodge")+
  scale_y_continuous(breaks = seq(-100, 100, 0.2))+
  scale_fill_viridis_d()+
  labs(y="integrated CT changes [mol/m2]")+
  theme(axis.title.x = element_blank())

p_iCT_cum <- ts_profiles_ID_surface %>% 
  ggplot(aes(date_time_ID,  iCT_cum, 
             col=rho_lim))+
  geom_line()+
  geom_hline(yintercept = 0)+
  scale_color_viridis_d()+
  scale_y_continuous(breaks = seq(-100, 100, 0.2))+
  theme(strip.background = element_blank(),
        strip.text = element_blank())+
  labs(y="integrated, cumulative CT changes [mol/m2]", x="date")


(p_iCT / p_iCT_cum)+
  plot_layout(guides = 'collect')

rm(p_iCT, p_iCT_cum)
```