key_drivers.Rmd

---
title: "Key drivers of change in Arctic fjord ecosystems"
author: "Robert Schlegel"
date: '`r format(Sys.Date(), "%d %B %Y")`'
output: workflowr::wflow_html
editor_options:
  chunk_output_type: console
# css: acid.css
csl: frontiers.csl
bibliography: FACE-IT.bib
---

```{r global_options, include = FALSE}
knitr::opts_chunk$set(fig.width = 8, fig.align = 'center',
                      echo = FALSE, warning = FALSE, message = FALSE, 
                      eval = TRUE, tidy = FALSE)

# Libraries
library(tidyverse) # The tidy dialect of R
library(kableExtra) # For formatting static tables
```

<center>
![](assets/FACE-IT_Logo_900.png){ width=70% }
</center>

# Overview

It is well known that the Arctic is undergoing rapid systemic changes, and that these changes are likely to increase in both rate and extent. The FACE-IT project is primarily interested in how these changes will be impacting the socio-ecological fjord and adjacent coastal systems at seven study sites in the Arctic. There is a steadily growing body of data from these sites that can be used to document and investigate the changes that would be of scientific and societal interest. It is therefore necessary, at the outset of the FACE-IT project, to identify the key drivers of change in Arctic biodiversity and to identify the availability of data relating to these drivers.

This website was created for D1.1 of the [Horizon2020](https://ec.europa.eu/programmes/horizon2020/en/home) project [FACE-IT](https://www.face-it-project.eu/).

## Drivers vs Variables

The use of the word "driver", rather than "variable", to describe data is counterintuitive to some research groups. The decision to use the word "driver" is based on the fact that FACE-IT is a consortium of many teams that employ a range of different methods of scientific inquiry. The technical vocabulary used across these groups may differ in some important ways so it was decided to use the unifying terminology found in the IPCC reports. It must therefore be noted here that a "driver" does not necessarily mean that something is a physical (or otherwise) driver of another variable. In the context of FACE-IT most of the types of data that are being measured are indeed drivers of other types of data, but it is not necessary for them to be in order to be labelled as a "driver". For some it may be easier to think of the data outlined here as "variables", rather than "drivers".

## Selection criteria for key drivers

At the outset of the FACE-IT project a Google Sheet was created that allowed FACE-IT experts to fill-in which drivers were relevant for their work/study sites and where any corresponding data may be available. This sheet was left open for several months, and contributors were contacted by WP1 to ensure that all listed drivers were of critical importance. Those that were not were removed, and similar drivers were combined into single line items. Additionally, drivers that were specific to a single site, or considered outside of the scope of FACE-IT were also removed. The items remaining on the Google sheet therefore had to meet the following three criteria in order to be classified as key drivers: 1) confirmed importance by expert(s), 2) within scope of FACE-IT, 3) data availability. The key drivers identified in that spreadsheet were then classified into one of the five categories: <span style="background:mintcream; border-width:1px; border-style:solid; border-color:black"> Cryosphere</span>, <span style="background:skyblue"> Physical</span>, <span style="background:#F6EA7C"> Chemistry</span>, <span style="background:#A2ED84"> Biology</span>, <span style="background:#F48080"> Social</span>.

## 18 month report update

During the 18 month report meeting it was noted that the key drivers presented there had been updated from those listed below. This is because the updated list of key drivers was made in order to address the challenges of D1.3: _Review article on past and future changes of key drivers_, and are therefore a part of D1.3, and not D1.1. The key drivers for D1.3 were further focussed and refined after 18 months of extensive literature review and data collection. Many discoveries were made about the original list of key drivers that excluded many of them from being discussed in the review paper. One good example of this is the oxygen concentration in fjord waters. While it is generally agreed upon that this is an important driver (and is listed as such below in this report), when actually going through the literature it turns out that little is actually known about the effects of climate change on oxygen in Arctic fjord waters. Indeed, what little has been published shows that the importance of this driver in the Arctic is likely much less than the global ocean generally.

There are many other such examples like that of oxygen, and many of them were discussed in the [review paper](https://www.cambridge.org/core/journals/cambridge-prisms-coastal-futures/article/drivers-of-change-in-arctic-fjord-socioecological-systems-examples-from-the-european-arctic/1129E92425F6012F9610D8C9C172B2F0), which was published January, 2023. It must also be noted that many of the key drivers in the updated list represent a combination of several similar key drivers from this list. An example being the key driver “Carbonate system” in the updated list, which is a combination of the several CO2 related drivers in the “Chemistry” category of drivers below (e.g. Partial pressure of CO2, pH, total alkalinity). The drivers in the report below have therefore not been replaced by the refined list for D1.3 because the key drivers listed here follow a consistent methodology that has not changed. The drivers listed here continue to serve as the operational base for investigations by the other WPs because the refined list created for D1.3 is not meant to be restrictive on the other research projects. That being said, the refined D1.3 list is much more concise, so is displayed first in this document, with the D1.1 list provided at the end. One may use the table of contents (TOC) menu on the left to jump there directly.

# <span style="background:mintcream; border-width:1px; border-style:solid; border-color:black"> Cryosphere</span>

The FACE-IT project is focused on fjord and adjacent coastal socio-ecological systems, which means it is concerned primarily with marine data, but changes in the terrestrial cryosphere have a direct impact on coastal systems and so are key variables when considering drivers of change. These terrestrial variables include any measure of glaciers, such as mass balance and changes to their terminating edges. Also of interest are changes to permafrost and snow cover thickness. All of these terrestrial changes are important to fjords as they have direct impacts on the rates and composition of runoff and river discharge in the coastal zone, which have in turn very large impacts on water quality variables such as salinity, nutrients concentration and turbidity. The importance of changes to water quality are discussed again in the <span style="background:skyblue"> Physical</span> and <span style="background:#F6EA7C"> Chemistry</span> sections.

One of the aspects of the cryosphere that most directly impacts fjord and adjacent coastal systems is sea-ice. This is because the presence of ice can have a dramatic effect on what may be considered the three most fundamental important <span style="background:skyblue"> Physical</span> variables of sea water: temperature, salinity, and light. There are many ways of measuring sea-ice including: concentration, extent, thickness, and age. Whereas only one of these variables can be used when necessary, it is almost always preferable to have as many different measures of sea-ice as possible. It has been documented that there is a reliable correlation between fast-ice (ice attached to the coast) and local air and sea temperatures. This means that it may be possible to use air and sea temperatures as a proxy for sea-ice when there is a paucity of these data.

```{r}
knitr::kable(data.frame(Category = "Cryosphere",
                        Driver = c("sea ice", "glacier mass balance", "terrestrial runoff")), 
             caption = "Table 1: The key cryosphere drivers of changes for Arctic fjords.") %>% 
  kable_styling(bootstrap_options = c("striped", "hover", "responsive")) %>% 
  row_spec(1:3, background = "mintcream") #%>% # Cryosphere
  # row_spec(7:20, color = "black", background = "skyblue") %>% # Physical
  # row_spec(21:29, background = "#F6EA7C") %>% # Chemistry
  # row_spec(30:36, background = "#A2ED84") %>% # Biology
  # row_spec(37:42, background = "#F48080") # Social
```

# <span style="background:skyblue"> Physical</span>

One could consider the <span style="background:mintcream; border-width:1px; border-style:solid; border-color:black"> Cryosphere</span>  drivers to be physical drivers, but they are categorised separately due to their importance in the Arctic. The physical aspects of coastal systems adjacent to fjords are generally well documented as part of global gridded datasets. However, the resolution of gridded data is often too coarse to capture well the local scale variability inside fjords. It is therefore preferable to identify local sources of these key drivers wherever possible. Bathymetry being a primary example of this issue.

As mentioned in the <span style="background:mintcream; border-width:1px; border-style:solid; border-color:black"> Cryosphere</span>  section, potentially the three most important variables are seawater temperature, salinity, and light. This is because these three variables have systemic effects on the <span style="background:#A2ED84"> Biology</span> of the foundational species and ecosystems from which the other socio-ecological functioning of the fjords and adjacent coastlines are based. There are a number of global products that provide these data as well as a host of locally collected in situ data. While these three variables alone would generally suffice for the physical characterisation of drivers of change, there are many other variables that are of interest to FACE-IT. For example, seawater temperatures do not change uninvited. They are acted on by air-sea heat flux and so it is also desirable to know what changes may be occurring in: net heat flux, long+shortwave radiation, latent+sensible heat flux. As part of the balance of flux across the air-sea border, one would also want to know about: the mixed layer depth, sea level pressure, precipitation, evaporation, and wind speed+direction. All of these are important variables in their own right as they tell a more specific part of the story of a changing climate. Additionally, the historic location, direction, and volume of ocean currents have been changing, which have had published impacts on fjord and coastal systems, so it is also important to measure this driver.

In addition to changes in the air-sea fluxes and currents, it is also of great importance to understand what changes to water quality may be occurring. Many of these drivers have been grouped into the <span style="background:#F6EA7C"> Chemistry</span> section below, so in this section we focus on river discharge and the more physical aspects of water quality: sedimentation, suspended matter, suspended solids, suspended organics. These drivers are important because they may have a very large impact on light penetration at depth, which can change the competitiveness of different biological groups and have disruptive influences on local ecosystems. For example, changes in the timing and intensity of river runoff can unduly promote the early growth of phytoplankton blooms, which will then deplete nutrients earlier than expected, making it more difficult for kelp forests to grow, thereby limiting benthic primary production, which in turn may cascade up the trophic food-web.
  
```{r}
knitr::kable(data.frame(Category = "Physics",
                        Driver = c("seawater temperature", "salinity", "light")), 
             caption = "Table 2: The key physics drivers of changes for Arctic fjords.") %>% 
  kable_styling(bootstrap_options = c("striped", "hover", "responsive")) %>% 
  # row_spec(1:3, background = "mintcream") #%>% # Cryosphere
  row_spec(1:3, color = "black", background = "skyblue") #%>% # Physical
  # row_spec(21:29, background = "#F6EA7C") %>% # Chemistry
  # row_spec(30:36, background = "#A2ED84") %>% # Biology
  # row_spec(37:42, background = "#F48080") # Social
```

# <span style="background:#F6EA7C"> Chemistry</span>

The first of the two primary areas of interest for the chemistry drivers is the changes to the nutrient load of seawater. The main nutrients of interest are: nitrate, nitrite, ammonium, phosphate, and silicate. These are important to monitor as dramatic short-term, or significant long-term changes to the nutrient profiles in fjord and adjacent coastal waters can cause bottom up forcing of ecological structures. Potentially fundamentally restructuring the species composition/biodiversity of a study site. Also of interest for water quality measurements are dissolved organic carbon (DOC), dissolved organic nitrogen (DON), and dissolved O~2~. These variables are important for similar reasons to the nutrients.

The other primary area of interest for this category is the amount of inorganic carbon and related minerals such as aragonite and calcite as changes to these minerals are related to shifts in the pH of seawater. This is of concern because the acidification of the ocean has a wide range of well documented effects on sea life. Primarily on calcifying organisms such as coralline algae, coccolithophores, molluscs and sea urchins, which are a foundational species in some pelagic and benthic ecosystems. Parameters of the carbonate system which are of interest are pCO~2~, pH, total alkalinity, and the concentration of dissolved inorganic carbon (DIC). 

```{r}
knitr::kable(data.frame(Category = "Chemistry",
                        Driver = c("carbonate chemistry", "nutrients")), 
             caption = "Table 3: The key chemistry drivers of changes for Arctic fjords.") %>% 
  kable_styling(bootstrap_options = c("striped", "hover", "responsive")) %>% 
  # row_spec(1:3, background = "mintcream") #%>% # Cryosphere
  # row_spec(1:3, color = "black", background = "skyblue") #%>% # Physical
  row_spec(1:2, background = "#F6EA7C") #%>% # Chemistry
  # row_spec(30:36, background = "#A2ED84") %>% # Biology
  # row_spec(37:42, background = "#F48080") # Social
```

# <span style="background:#A2ED84"> Biology</span>

This category is comprised of individual species and the broader ecosystems that they populate. These variables are controlled by a lot of chemical, physical, and biological drivers and their change is itself a driver of changes, for example on upper trophic levels of relevance for humans, including emblematic and species of commercial interest. Foremost amongst the drivers for these categories is the very broad concept of biodiversity. One of the FACE-IT work packages (WP2) focuses on this driver primarily oriented around their target species (e.g. kelps and seabirds). To this end the biology category could contain a near limitless list of drivers if every aspect of biological change for the myriad of potential target species found within fjord and adjacent coastal systems were to be considered. As the FACE-IT researchers are experts in their field of study, the specifics of a given target species are left to them to track. In most cases this is of no concern as they will have been the one who generated the original data. Variables that have a broader interest across focal species and ecosystems are also considered as key drivers of change. These include biogeochemical variables such as photosynthesis, respiration, primary production, calcification and nitrogen fixation.

Another requirement for biology data is to be able to perform species distribution modelling (SDM) in order to project where the target species may currently be present, and where they may be with future projections. In order to perform SDM a researcher needs access to as much data describing the focal species as possible. Traditionally this is species presence/absence and abundance/biomass for a given site or geographical location.

```{r}
knitr::kable(data.frame(Category = "Biology",
                        Driver = c("primary production", "biomass", "species richness")), 
             caption = "Table 4: The key biology drivers of changes for Arctic fjords.") %>% 
  kable_styling(bootstrap_options = c("striped", "hover", "responsive")) %>% 
  # row_spec(1:3, background = "mintcream") #%>% # Cryosphere
  # row_spec(1:3, color = "black", background = "skyblue") #%>% # Physical
  # row_spec(1:2, background = "#F6EA7C") #%>% # Chemistry
  row_spec(1:3, background = "#A2ED84") #%>% # Biology
  # row_spec(37:42, background = "#F48080") # Social
```

# <span style="background:#F48080"> Social</span>

Putting the 'socio' in socio-ecological fjord systems, the social category contains the most relevant drivers that involve or directly affect human activities in the fjords and adjacent coastlines. Of most direct impact on fjord systems for this category is fish landings, both commercial and recreational. This is because the biomass removed from the study sites has a significant impact on the ecosystem. Also, species migration could create opportunities. Hunting (i.e. game landings) is also of importance as the removal of larger terrestrial organisms can have a trickle down effect on the pressure between trophic levels. To this end, it is necessary to document the local and national management regulations strategies for these livelihood activities, and how these are linked to science and scientific advice. It is relevant to understand how and on what basis hunting and fish catch quotas are set. This will provide insights into how the extent to which fish and marine mammals are managed in a multispecies and ecosystems frame.  Resource management is therefore another driver which will impact the species composition and abundance of commercially important fish and marine mammal species and their prey.

An indirect impact on the fjord and adjacent coastal systems comes from human activities and presence, such as the traffic from tourism and recreational vessels, and other activities in the fjords and along the coasts. Indirect impacts of human activities are harder to measure and estimate when compared to intentional direct impacts on the fjord ecosystems such as regulated fishing and hunting. Human activities create waste and pollution to such a degree that it can have disturbing effects in the natural state of the study sites. The drivers that may be tracked to determine changes to anthropogenic perturbations at a study site are: the tourist arrivals per month and their nationality, tourist vessel counts and mileage, seasonal patterns of activities and the national statistics of demography, income, coastal livelihoods, and unemployment.

```{r}
knitr::kable(data.frame(Category = "Biology",
                        Driver = c("governance", "tourism", "fisheries")), 
             caption = "Table 5: The key social drivers of changes for Arctic fjords.") %>% 
  kable_styling(bootstrap_options = c("striped", "hover", "responsive")) %>% 
  # row_spec(1:3, background = "mintcream") #%>% # Cryosphere
  # row_spec(1:3, color = "black", background = "skyblue") #%>% # Physical
  # row_spec(1:2, background = "#F6EA7C") #%>% # Chemistry
  # row_spec(1:3, background = "#A2ED84") #%>% # Biology
  row_spec(1:3, background = "#F48080") # Social
```

# Data availability

The data for the drivers outlined above are available from multiple established data sources. Large gridded data sets for most physical drivers at the Arctic scale can be accessed via: [Copernicus](https://marine.copernicus.eu/), [EMODnet](https://www.emodnet.eu/en), and [NOAA](https://www.ncei.noaa.gov/). Site specific datasets can be accessed via: [NPDC](https://data.npolar.no/home/), [NSF](https://arcticdata.io/), [openpolar](https://openpolar.no/), [PANGAEA](https://pangaea.de/), and [SIOS](https://sios-svalbard.org/metadata_search). Data for drivers pertaining more to policy may be found at: [INTAROS](https://intaros.nersc.no/) and [SAON](https://www.arcticobserving.org/). The data portals for national statistics and other social drivers are: [Grønlands statistik](https://stat.gl/dialog/topmain.asp?lang=da&subject=Fiskeri%20og%20fangst&sc=FI), [kystverket](https://ais-public.kystverket.no/ais-download/), [SSB](https://www.ssb.no/), and [statistikknett](https://statistikknett.no/). The specific data of relevance to the key drivers that may be found within each of these sources are given in greater detail in [D1.2: Meta-database](https://face-it-project.github.io/WP1/metadatabase.html) which was delivered on month 12 (Nov 2021).

# D1.1 driver list

```{r key-drivers-table, eval=TRUE}
# Easiest to show each variable as an smol individual data.frame

# Convenience wrapper for drivers
make_driver <- function(category, driver, why, sites, reference){
  res <- data.frame(Category = category, Driver = driver)#, Why = why, Sites = sites, Reference = reference)
}

# The order of the info:
## Category, Driver, Why, Sites, Reference

## Cryosphere
cryo_coastal_ice <- make_driver("Cryosphere", "Coastal ice: formation/break up date, thickness")#, "", "", "")
cryo_fast_ice <- make_driver("Cryosphere", "Fast ice: extent, thickness", "Decreasing (fast) ice cover can have a range of ecological effect.")#, "", "Muckenhuber et al. (2016)")
cryo_glacier <- make_driver("Cryosphere", "Glacier: land mass balance, terminating edge")#, "An indicator of the change to other drivers.", "", "")
cryo_permafrost <- make_driver("Cryosphere", "Permafrost: temperature, thickness")#, "", "", "")
cryo_sea_ice <- make_driver("Cryosphere", "Sea ice: concentration, extent, freeboard, thickness")#, 
                            # "Affects pelagic and benthic production.", "EU Arctic", "Stroeve and Notz (2018)")
cryo_snow_cover <- make_driver("Cryosphere", "Snow cover: thickness")#, "", "EU Arctic", "")
cryo_ALL <- rbind(cryo_coastal_ice, cryo_fast_ice, cryo_glacier, cryo_permafrost, cryo_sea_ice, cryo_snow_cover)
# nrow(cryo_ALL)

## Physical
phys_bathy <- make_driver("Physical", "Bathymetry")#, "", "", "")
phys_current <- make_driver("Physical", "Current: direction, location, volume")#, "", "", "")
phys_evap <- make_driver("Physical", "Evaporation")#, "", "", "")
phys_heatflux <- make_driver("Physical", "Heatflux: net, latent/sensible, long/shortwave radiation")#, 
                             # "Responsible for the changing of seawater temperature in various important ways.", "EU Arctic", "")
phys_extinct <- make_driver("Physical", "Light extinction coefficient")#, "Indicates the amount of light to benthic producers.", "EU Arctic", "")
phys_mld <- make_driver("Physical", "Mixed layer depth")#, "", "EU Arctic", "Popova et al. (2014)")
phys_precip <- make_driver("Physical", "Precipitation")#, "Affects the amount of fresh water runoff into coasts.", "EU Arctic", "")
phys_river_discharge <- make_driver("Physical", "River discharge")#, 
                                    # "Delivers carbon and nutrient rich freshwater to coast.", "EU Arctic", "Delpech et al. (2021)")
phys_salinity <- make_driver("Physical", "Salinity")#, "", "EU Arctic", "")
phys_mslp <- make_driver("Physical", "Sea level pressure")#, "", "", "")
phys_sst <- make_driver("Physical", "Seawater temperature: surface, mid, bottom")#, "", "EU Arctic", "Popova et al. (2014)")
phys_sedimentation <- make_driver("Physical", "Sedimentation rate")#, "", "", "")
phys_suspended <- make_driver("Physical", "Suspended matter: organic, mineral")#, "", "", "")
phys_wind <- make_driver("Physical", "Wind: direction, speed")#, "", "", "")
phys_ALL <- rbind(phys_bathy, phys_current, phys_evap, phys_heatflux, phys_extinct, phys_mld, phys_precip, phys_river_discharge, 
                  phys_salinity, phys_mslp, phys_sst, phys_sedimentation, phys_suspended, phys_wind)
# nrow(phys_ALL)

## Chemistry
chem_CACO3 <- make_driver("Chemistry", "CaCO3 saturation state")#, "", "EU Arctic", "")
chem_DIC <- make_driver("Chemistry", "Dissolved inorganic carbon")#, "", "EU Arctic", "Popova et al. (2014)")
chem_DOC <- make_driver("Chemistry", "Dissolved organic carbon")#, "", "", "")
chem_DON <- make_driver("Chemistry", "Dissolved organic nitrogen")#, "", "", "")
chem_DO2 <- make_driver("Chemistry", "Dissolved 02")#, "", "", "")
chem_nutrients <- make_driver("Chemistry", "Nutrient: nitrate, nitrite, ammonium, phosphate, silicate")#, "", "", "")
chem_pCO2 <- make_driver("Chemistry", "Partial pressure of CO2")#, "", "EU Arctic", "Popova et al. (2014)")
chem_pH <- make_driver("Chemistry", "pH")#, "", "EU Arctic", "")
chem_alk <- make_driver("Chemistry", "Total alkalinity")#, "", "EU Arctic", "")
chem_ALL <- rbind(chem_CACO3, chem_DIC, chem_DOC, chem_DON, chem_DO2, chem_nutrients, chem_pCO2, chem_pH, chem_alk)
# nrow(chem_ALL)

## Biology
bio_calc <- make_driver("Biology", "Calcification")#, "", "", "")
bio_N_fix <- make_driver("Biology", "Nitrogen fixation")#, "", "", "")
bio_photo <- make_driver("Biology", "Photosynthesis")#, "", "", "")
bio_prim_prod <- make_driver("Biology", "Primary production")#, "", "", "")
bio_resp <- make_driver("Biology", "Respiration")#, "", "", "")
bio_species <- make_driver("Biology", "Species: presence/absence, abundance/biomass")#, "", "", "")
bio_ALL <- rbind(bio_calc, bio_N_fix, bio_photo, bio_prim_prod, bio_prim_prod, bio_resp, bio_species)
# nrow(bio_ALL)

## Social
soc_fish_landings <- make_driver("Social", "Fish landings: commercial, recreational, quotas, seasonality")#, 
                                # "", "Porsangerfjorden + Isfjorden + Disko Bay", "")
soc_game_landings <- make_driver("Social", "Game landings: quotas, seasonality")#, "", "Porsangerfjorden + Isfjorden + Disko Bay", "")
soc_local_management <- make_driver("Social", "Local and national resource management")#, "", "", "")
soc_national_statistics <- make_driver("Social", "National statistics: demography, income, unemployment")#, "", "EU Arctic", "")
soc_tourist_arrivals <- make_driver("Social", "Tourist arrivals: per month, nationality")#, "", "Isfjorden + Disko Bay", "")
soc_tourist_vessels <- make_driver("Social", "Tourist vessels: count, mileage")#, 
                                   # "Indicative of direct anthropogenic disturbance.", "Isfjorden + Disko Bay", "")
soc_ALL <- rbind(soc_fish_landings, soc_game_landings, soc_local_management, 
                 soc_national_statistics, soc_tourist_arrivals, soc_tourist_vessels)
# nrow(soc_ALL)

# Combine and print
knitr::kable(rbind(cryo_ALL, phys_ALL, chem_ALL, bio_ALL, soc_ALL), 
             caption = "Table 6: The key drivers of changes on Arctic fjord and adjacent coastal systems as identified by the experts in FACE-IT for D1.1. The drivers have been separated into the following categories: Cryosphere, Physical, Chemistry, Biology, Social. The 'Crysophere' category contains any drivers related to ice, glaciers, and/or snow. This category contains both marine and terrestrial drivers that impact the marine environements of Arctic fjords and adjacent coastlines. The 'Physical' category contains atmospheric and oceanic drivers. The 'Chemistry' category contains primarily water quality variables that are no longer physical, but not yet biological. This includes inorganic as well as organic nutrients. The 'Biology' category contains drivers related to biology and ecology. The 'Social' category contains drivers that are primarily related to anthropogenic impacts on the natural world, as well as drivers that directly affect the well-being of local stakeholders. This category therefore contains all drivers relating to fisheries.") %>% 
  kable_styling(bootstrap_options = c("striped", "hover", "responsive")) %>% 
  row_spec(1:6, background = "mintcream") %>% # Cryosphere
  row_spec(7:20, color = "black", background = "skyblue") %>% # Physical
  row_spec(21:29, background = "#F6EA7C") %>% # Chemistry
  row_spec(30:36, background = "#A2ED84") %>% # Biology
  row_spec(37:42, background = "#F48080") # Social
```