Forecast_Assessment.Rmd

---
title: "Forecast_Assessment"
output: html_document
date: "2024-05-08"
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

```{r cars}
library(arrow)

```

```{r}
all_results <- arrow::open_dataset("s3://anonymous@bio230014-bucket01/challenges/forecasts/parquet/project_id=neon4cast/duration=P1D/variable=oxygen/model_id=AquaticEcosystemsOxygen?endpoint_override=sdsc.osn.xsede.org")
df <- all_results |> dplyr::collect()


```

```{r}
target <- readr::read_csv("https://data.ecoforecast.org/neon4cast-targets/aquatics/aquatics-targets.csv.gz", guess_max = 1e6)
oxygen <- target |> 
  dplyr::filter(variable == "oxygen")
```

```{r}

```

```{r}
# Step 1: Filter the prediction data for site "WLOU" and reference_datetime "2024-05-01"
filtered_predictions <- subset(df, site_id == "WLOU" & reference_datetime == as.Date("2024-05-01"))

# Step 2: Merge the filtered predictions with the observation data
comparison_data <- merge(filtered_predictions, oxygen, by = c("site_id", "datetime"))

# Step 3: Calculate the difference between predicted and observed values
comparison_data$prediction_error <- comparison_data$prediction - comparison_data$observation

# Optionally, you can view the first few rows to check the results
head(comparison_data)

# Further analysis, like summary statistics
summary(comparison_data$prediction_error)

# Plotting the difference can also be insightful
library(ggplot2)
ggplot(comparison_data, aes(x = datetime, y = prediction_error)) +
  geom_line() +
  labs(title = "Prediction Error Over Time", x = "Datetime", y = "Prediction Error")

```

```{r}

# Load the ggplot2 library
library(ggplot2)

# Create a line plot with both predictions and observations
ggplot(comparison_data, aes(x = datetime)) +
  geom_line(aes(y = prediction, colour = "Predicted"), size = 1) +
  geom_line(aes(y = observation, colour = "Observed"), size = 1) +
  labs(title = "Predicted vs Observed Values Over Time",
       x = "Datetime",
       y = "Value",
       colour = "Legend") +
  theme_minimal() +
  scale_colour_manual(values = c("Predicted" = "blue", "Observed" = "red"))

```