HAB-Prediction-Research

Salinty.csv and Temperature.csv aquired from SIO https://shorestations.ucsd.edu/publications/data/ scoos_HAB_SIO.csv from https://habs.sccoos.org/scripps-pier

Bloom Forecast Inference Guide

---

1. Project Directory Structure

Code/
└─ Scripts/
   ├─ configs/
   │  └─ biolum_config.yaml   ← example configuration
   ├─ outputs/
   │  └─ bloom_forecast.json  ← generated after running inference
   └─ forecast.py             ← inference script

---

2. Create a Configuration File

1. Navigate to Code/Scripts/configs/.
2. Copy the sample file or create a new one, e.g.:

cp biolum_config.yaml my_config.yaml

3. Edit the new file (my_config.yaml) and fill in each variable:

Key	Description
data_path	Path to the “back-up” dataset used for predictions
parameters_path_1wk	Path to the 1-week model parameters
parameters_path_2wk	Path to the 2-week model parameters
parameters_path_3wk	Path to the 3-week model parameters
target	Name of the variable to forecast
json_key_path	Google Drive service-account JSON key file
bloom_thresh	Threshold: if a single model’s forecast ≥ this value, classify as “bloom”
samp	Use every samp-th model to reduce over-fitting
n	Total number of models in the ensemble
p	Percentage of models that must predict “bloom” for the final label Likely

Tip: Keep paths relative to Code/Scripts/ when possible to simplify execution.

Example snippet:

# my_config.yaml
data_path: "../Data/d1carter_data_w_gaps.csv"
parameters_path_1wk: "/Users/athulithparaselli/Documents/GitHub/HAB-Prediction-Research/Code/Results/1000_sample_biolum_1_wk.csv"
parameters_path_2wk: "/Users/athulithparaselli/Documents/GitHub/HAB-Prediction-Research/Code/Results/1000_sample_biolum_2_wk.csv"
parameters_path_3wk: "/Users/athulithparaselli/Documents/GitHub/HAB-Prediction-Research/Code/Results/1000_sample_biolum_3_wk.csv"
target: "Lingulodinium_polyedra"
json_key_path: "/Users/athulithparaselli/Documents/GitHub/HAB-Prediction-Research/Code/APIKey/sacred-truck-461716-t5-2a6499bf0869.json"
bloom_thresh: 11733.399999999963
samp: 40
n: 750
p: 0.3

---

3. Obtain & Add a Google Service-Account Key

1. Visit https://console.cloud.google.com.
2. Create (or select) a project and enable the Google Drive API.
3. Navigate to APIs & Services → Credentials → + Create Credential → Service account key.
4. Choose JSON as the key type and download the file.
5. Save the file to a secure location (e.g., Code/Scripts/secrets/) and update json_key_path in your YAML.

---

4. Run Inference

From Code/Scripts/:

python3 forecast.py configs/my_config.yaml

• Replace my_config.yaml with the name of your configuration file.
• The script will read the config, load models, and produce forecasts.

---

5. Output

• A JSON file named bloom_forecast.json is written to:

Code/Scripts/outputs/bloom_forecast.json

This file contains the predicted bloom likelihoods and confidence intervals for the upcoming weeks.

---

You’re all set! Customize your YAML, run the command, and check the output JSON for your bloom forecasts.