Skip to content

Train a multi modal deep learning scheme (combining CNN and FNN) to predict cloudburst events using INSAT3D images and ERA5 hourly temperature and precipitation data

License

Notifications You must be signed in to change notification settings

Pramit726/Predict-Cloudburst

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Advancing Cloudburst Prediction with Mixed-Input Deep Learning: Combining CNN and Feed-Forward Neural Networks

Train a multi-modal deep learning scheme (combining CNN and FNN) to predict cloudburst events using INSAT3D images and ERA5 hourly temperature and precipitation data.

Input data

  • Satellite images from INSAT3D collected from Meteorological and Oceanographic Satellite Data Archival Centre(https://mosdac.gov.in/) during the occurrence of cloud bursts, as well as 6 hours prior and 6 hours after the cloudburst. Images of random noncloudburst events
  • satellite images were labeled into 4 categories: 'cloudburst,' 'precloudburst', 'postcloudburst', and 'no cloudburst
  • ERA5 Climate reanalysis data (hourly temperature and precipitation) from CDS store (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels?tab=form) converted into .csv format

Functionality

  • Preprocess data: the script preprocess corresponding spatial satellite images and tabular climate data from ERA5 and merge them
  • Splitting into train, validation, and test set : the processed data is shuffled and split into train, validation, and test dataset
  • Model architecture : a model architecture is defined by combining a CNN (MobileNetV2) and FNN
  • Model optimization: the defined model is optimized using ADAM algorithm in 10 epochs and model loss and accuracy were collected for each epoch
  • Visualization : model loss and accuracy are plotted to visualize if model accuracy is getting better with each epoch
  • Model testing : the best performing trained model is tested on the test dataset and model accuracy and loss on test dataset are printed

Future scope

  • The trained multi-model DL scheme can be fed with real-time satellite images and weather data to predict cloudburst risk.
  • If a cloudburst risk is detected, automatic notifications can be sent to concerned authorities for disaster readiness and evacuation planning.

Dependencies

  • Pandas
  • Numpy
  • cv2
  • Matplotlib
  • TensorFlow

Contributor

About

Train a multi modal deep learning scheme (combining CNN and FNN) to predict cloudburst events using INSAT3D images and ERA5 hourly temperature and precipitation data

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published