🌊 SST Prediction App 🌡️

Introduction

Welcome to the SST Prediction App! 🎉

This application enables users to predict Sea Surface Temperatures (SST) based on historical data. With a specialized ConvLSTM model, the app processes 5 days of SST data to predict the SST for the 6th day.

Model Overview:

🔧 Trained by Boris

Layer (type)	Output Shape	Param #
ConvLSTM2D	(None, 401, 451, 32)	38,144
BatchNormalization	(None, 401, 451, 32)	128
Dropout	(None, 401, 451, 32)	0
Conv2D	(None, 401, 451, 64)	18,496
BatchNormalization_1	(None, 401, 451, 64)	256
Dropout_1	(None, 401, 451, 64)	0
Conv2D_1	(None, 401, 451, 32)	18,464
BatchNormalization_2	(None, 401, 451, 32)	128
Dropout_2	(None, 401, 451, 32)	0
Conv2D_2	(None, 401, 451, 1)	289

Total params: 75,905
Trainable params: 75,649
Non-trainable params: 256

The data used hails from the MUR-JPL-L4-GLOB-v4.1 dataset on earthdata.nasa.gov.

Setup Steps 🚀:

1. Clone the repo:

git clone https://github.com/NaNa7Miiii/upwelling_prediction.git

2. Navigate to the directory：

cd upwelling_prediction

3. Install the dependencies:

pip install -r requirements.txt

4. Launch the app:

streamlit run app.py

5. Please comment out this chunk of code which aims to create a .netrc file to access NASA's data, but if you want to deploy the repo in streamlit cloud please don't do anything on that:

NETRC_PATH = os.path.expanduser("~/.netrc")

def create_netrc(machine, login, password, path=NETRC_PATH):
    netrc_content = f"""
    machine {machine}
    login {login}
    password {password}
    """

    with open(path, "w") as file:
        file.write(netrc_content)

    os.chmod(path, 0o600)

secrets = st.secrets["earthdata_test"]
create_netrc(secrets["machine"], secrets["login"], secrets["password"])

After all that, you are ready to go!