Vyom Pathak1 | Brijesh Bhatt1 | Arvind Sahay2 | Mini Raman2
Dharmsinh Desai University, Nadiad1 | Indian Space Research Organization, Ahmedabad2
Abstract
Inherent optical properties (IOPs) of the coastal oceans are modulated independently by the in-water optical constituents, which cause variations in the water leaving radiances or re-mote sensing reflectances. Accurate determination of IOPs and the optical constituents from water-leaving radiances or reflectances using conventional empirical ratio approaches fail in the coastal oceans. Alternate non-parametric approaches such as neural network (NN) based approaches can be developed to derive the parameters of interest using training datasets. Further NN based approaches can deal with the non-linearity of functional dependence between optical constituents and the IOPs.To retrieve the IOPs, earlier NN models used Levenberg-Marquardt with Bayesian Regularization as an optimizer for learning the weights of the model, which has a slow learning rate. Moreover, with low-resource datasets while retrieving IOPs till the third level, the probability of error propagation becomes high. To overcome these two problems, we present a Modified Neural Network (MNN) algorithm (modification of NN model Ioannou et. al. to retrieve Inherent Optical Properties (IOPs) of ocean waters, in which three Neural Networks (NN) were developed in parallel. Our method is based on the approach where we use the Adam optimizer, instead of the Levenberg-Marquardt since it has a faster training time. Also, the error propagation is observed to be very less even with low-resource data while retrieving IOPs at the third level, with a decent R 2 score.Results of the MNN algorithm indicate that MNN retrieves IOPs with an R 2 = 0.99 between measured and predicted values for b bp (443) and R 2 = 0.99 for a pg (443) at Level 1. Level-2 products give R 2 = 0.98 and R 2 = 0.99 between measured and predicted values for a pg (443) and a dg (443) respectively. Similarly Level-3 products give R 2 = 0.97 and R 2 = 0.51 between measured and predicted values for a g (443) and a d (443) respectively. The algorithm retrieves better R 2 score for all parameters except a d (443) compared to Semi Analytical Algorithm, Quasi Analytical Algorithm and NN algorithm by Ioannou et. al.. The new technique has the advantage of faster convergence and better generalization capacity for deriving IOPs from complex waters. The new algorithm is also able to separate gelbstoff and detrital absorption.
If you find this work useful, please cite this work using the following BibTeX:
@inproceedings{pathak2021neural,
title={Neural Network Based Retrieval of Inherent Optical Properties (IOPs) Of Coastal Waters of Oceans},
author={Pathak, Vyom and Bhatt, Brijesh and Sahay, Arvind and Raman, Mini},
booktitle={2021 IEEE International India Geoscience and Remote Sensing Symposium (InGARSS)},
pages={285--288},
year={2021},
organization={IEEE}
}
- Linux OS
- Python-3.6
- TensorFlow-2.2.0
git clone https://github.com/01-vyom/NN_Based_IOP_Retrieval_From_Coastal_Waters.git
python -m venv nn_iop_env
source $PWD/nn_iop/bin/activate
Change directory to the root of the repository.
pip install --upgrade pip
pip install -r requirements.txt
Change directory to the root of the repository.
To train the model in the paper, run this command:
python ./src/model.py
Note:
- The training data is stored in
./data
directory. - The results are stored in
./results
directory. - The model is stored in
./model
directory.
Our algorithm achieves the following performance:
Technique name | b_bp(443) | a_pg(443) | a_pg(443) | a_dg(443) | a_g(443) | a_d(443) |
---|---|---|---|---|---|---|
Semi Analytical Algorithm | 0.99 | 0.96 | 0.97 | 0.96 | do not exist | do not exist |
Quasi Analytical Algorithm | 0.98 | 0.93 | 0.98 | 0.93 | do not exist | do not exist |
NN algorithm by Ioannou et. al. | 0.99 | 0.92 | 0.98 | 0.93 | 0.92 | 0.89 |
Proposed Modified-NN algorithm | 0.99 | 0.98 | 0.99 | 0.98 | 0.97 | 0.51 |
This work was supported by the Indian Space Research Organization (ISRO), Ahmedabad, India.
Licensed under the MIT License.