---
title: "Subject 1: CO2 concentration in the atmosphere since 1958"
author: "NJ"
date: "11 July 2025"
output: html_document
Link: https://scrippsco2.ucsd.edu/data/atmospheric_co2/primary_mlo_co2_record.html
---


```{r}
data_url <- 
  "https://scrippsco2.ucsd.edu/assets/data/atmospheric/stations/in_situ_co2/monthly/monthly_in_situ_co2_mlo.csv"
data <- read.csv(data_url, skip=1, na.strings=c("-"))
```

```{r}
D <- is.data.frame(data)
D
```

```{r}
df <- read.csv(
  data_url,
  skip = 64,
  header = FALSE,
  col.names = c("Year", "Month", "ExcelDate", "DecimalDate", "CO2", 
                "CO2_Adjusted", "Fit", "Fit_Adjusted", 
                "Filled", "Filled_Adjusted", "Station")
)

head(df)
```

```{r}
library(ggplot2)

ggplot(df, aes(x = DecimalDate, y = CO2)) +
  geom_line(color = "steelblue") +
  labs(title = "Mauna Loa Atmospheric CO2",
       x = "Year",
       y = "CO2 (ppm)") +
  theme_minimal()
```

```{r}
##1. Make a plot that shows the superposition of 
#a periodic oscillation and a slower systematic evolution.

# Remove missing values (like -99.99)
df_clean <- df[df$CO2 > 0, ]

# Plot
ggplot(df_clean, aes(x = DecimalDate, y = CO2)) +
  geom_line(color = "darkblue", size = 0.6, alpha = 0.8) +
  geom_smooth(span = 0.3, se = FALSE, color = "red", size = 1.2) +
  scale_x_continuous(
    limits = c(1958, 2025),  # x-axis range
    breaks = seq(1960, 2025, by = 5)  # tick marks every 5 years
  ) +
  labs(
    title = "Superposition of Seasonal Oscillation and Long-Term CO2 Trend",
    subtitle = "Mauna Loa Observatory (MLO)",
    x = "Year",
    y = "CO2 concentration (ppm)"
  ) +
  theme_minimal()
```

```{r}
#2. Separate these two phenomena. 
#2.1. Characterize the periodic oscillation. 

# Create a time variable in years and months
df_clean$time <- df_clean$DecimalDate
```

```{r}
# Fit a quadratic model (simplified but effective)
trend_model <- lm(CO2 ~ poly(time, 2), data = df_clean)

# Add predicted trend to dataframe
df_clean$trend <- predict(trend_model)

df_clean$seasonal <- df_clean$CO2 - df_clean$trend

```

```{r}
# create the date like..."YYYY-MM-15" # 15 = mid month
df_clean$Date <- as.Date(paste(df_clean$Year, df_clean$Month, "15", sep = "-"))

df_clean$month <- as.numeric(format(df_clean$Date, "%m"))
```

```{r}
p1 <- ggplot(df_clean, aes(x = DecimalDate)) +
  geom_line(aes(y = CO2), color = "gray60", size = 0.5) +
  geom_line(aes(y = trend), color = "blue", size = 1.2) +
  scale_x_continuous(
    limits = c(1958, 2025),  # x-axis range
    breaks = seq(1960, 2025, by = 5)  # tick marks every 5 years
  ) + 
  labs(
    title = "Long-Term Trend in CO2 Concentration",
    x = "Year", y = "CO2 (ppm)"
  ) +
  theme_minimal()

p1
```

```{r}
#2.2. Find a simple model for the slow contribution, estimate its parameters, 
#and attempt an extrapolation until 2025 (for validating the model using future observations).

# Fit sine curve with 1-year periodne
season_model <- nls(seasonal ~ A*sin(2*pi*DecimalDate) + B*cos(2*pi*DecimalDate),
                    data = df_clean, start = list(A = 1, B = 1))

summary(season_model)

df_clean$time <- df_clean$DecimalDate
```


```{r}
# Create new data from last year to 2030
future <- data.frame(time = seq(from = max(df_clean$time), to = 2030, by = 1/12))

# Predict trend
future$trend <- predict(trend_model, newdata = future)

# Predict seasonal using fitted sine model
A <- coef(season_model)[1]
B <- coef(season_model)[2]
future$seasonal <- A * sin(2 * pi * future$time) + B * cos(2 * pi * future$time)

# Combine to get total CO2 estimate
future$CO2_predicted <- future$trend + future$seasonal

CO2_predicted <- future$trend + future$seasonal

```

```{r}
# Trend of the CO2 (ppm) predicted from 2025 to 2030 year
ggplot() +
  geom_line(data = future, aes(x = time, y = trend), color = "blue") +
  geom_line(data = future, aes(x = time, y = CO2_predicted), color = "red") +
  labs(title = "CO2 Projection with Trend + Seasonal Model",
       y = "CO2 (ppm)", x = "Year") +
  theme_minimal()
```

```{r}
# the inceasing of the CO2 (ppm) predicted within 2030 year
ggplot() +
  geom_line(data = df_clean, aes(x = DecimalDate, y = CO2), color = "black") +
  geom_line(data = future, aes(x = time, y = CO2_predicted), color = "darkgreen") +
  geom_vline(xintercept = 2023, linetype = "dashed", color = "red") +
  annotate("text", x = 2023.5, y = max(df_clean$CO2), 
           label = "Prediction zone", color = "red", hjust = 0) +
  scale_x_continuous(
    limits = c(1958, 2030),  # x-axis range
    breaks = seq(1960, 2030, by = 5)  # tick marks every 5 years
  ) + 
  labs(
    title = "CO2 Prediction Until 2030",
    x = "Year", y = "CO2 (ppm)"
  ) +
  theme_minimal()

```













Note: write your solution in the file module3/exo3/exercice_en.Rmd. 
Generate a PDF file using the KnitR tool and store it as module3/exo3/YourFileName.pdf (don't forget to commit it).