Transfer Learning with ResNet3D-101 for Global Prediction of High Aerosol Concentrations

Dušan P. Nikezić,Dušan S. Radivojević,Ivan M. Lazović,Nikola S. Mirkov,Zoran J. Marković

DOI: https://doi.org/10.3390/math12060826

IF: 2.4

2024-03-12

Mathematics

Abstract:In order to better predict the high aerosol concentrations associated with air pollution and climate change, a machine learning model was developed using transfer learning and the segmentation process of global satellite images. The main concept of transfer learning lies on convolutional neural networks and works by initializing the already trained model weights to better adapt the weights when the network is trained on a different dataset. The transfer learning technique was tested with the ResNet3D-101 model pre-trained from a 2D ImageNet dataset. This model has performed well for contrail detection to assess climate impact. Aerosol distributions can be monitored via satellite remote sensing. Satellites can monitor some aerosol optical properties like aerosol optical thickness. Aerosol optical thickness snapshots were the input dataset for the model and were obtained from NASA's Terra-Modis satellite; the output images were segmented by comparing the pixel values with a threshold value of 0.8 for aerosol optical thickness. Hyperparameter optimization finds a tuple of hyperparameters that yields an optimal model that minimizes a predefined loss function on given independent data. The model structure was adjusted in order to improve the performance of the model by applying methods and hyperparameter optimization techniques such as grid search, batch size, threshold, and input length. According to the criteria defined by the authors, the distance domain criterion and time domain criterion, the developed model is capable of generating adequate data and finding patterns in the time domain. As observed from the comparison of relative coefficients for the criteria metrics proposed by the authors, dcc and dtc, the deep learning model based on ConvLSTM layers developed in our previous studies has better performance than the model developed in this study with transfer learning.

mathematics

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve This paper aims to better predict high aerosol concentrations related to air pollution and climate change through a machine learning model. Specifically, the researchers developed a model using transfer learning and the segmentation process of global satellite images. #### Main Objectives - **Predict High Aerosol Concentrations**: Predict global high aerosol concentrations using the pre-trained ResNet3D-101 model through transfer learning techniques. - **Monitor Aerosol Distribution**: Monitor Aerosol Optical Thickness (AOT) using satellite remote sensing technology and use this data as input. - **Establish an Early Warning System**: Predict which areas will have AOT concentrations exceeding 0.8 in the next 8 days to assist decision support systems. #### Method Overview - **Data Source**: The input dataset comes from AOT snapshots obtained by NASA's Terra-Modis satellite, taken every 8 days from February 18, 2000, to November 17, 2023, totaling 1094 images. - **Model Structure**: The pre-trained ResNet3D-101 model is used and fine-tuned to adapt to the new dataset. The output is a binary segmentation image where pixels with AOT values greater than 0.8 are marked as 1, and those with values less than 0.8 are marked as 0. - **Evaluation Criteria**: Distance Domain Criterion (DDC) and Time Domain Criterion (DTC) are defined to evaluate model performance. #### Experimental Results - **Hyperparameter Optimization**: The model was optimized through methods such as grid search and batch size adjustment. - **Classification Threshold**: The optimal classification threshold was found through the Receiver Operating Characteristic (ROC) curve, ultimately determined to be 0.38. Through the above methods, this study developed a model capable of predicting high aerosol concentrations, providing a powerful tool for environmental monitoring and climate change.

Transfer Learning with ResNet3D-101 for Global Prediction of High Aerosol Concentrations

Symmetric U-Net Model Tuned by FOX Metaheuristic Algorithm for Global Prediction of High Aerosol Concentrations

Deep Learning Model for Global Spatio-Temporal Image Prediction

DAerosol-NTM: applying deep learning and neural Turing machine in aerosol prediction

Machine Learning Approach for Spatially and Temporally Resolved PM2.5 Exposures in the Continental United States

Deep Learning with Pretrained Framework Unleashes the Power of Satellite-Based Global Fine-Mode Aerosol Retrieval

Transfer Learning-Driven Hourly PM2.5 Prediction Based on a Modified Hybrid Deep Learning

Deep learning for Aerosol Forecasting

An Improved Hybrid Transfer Learning-Based Deep Learning Model for PM2.5 Concentration Prediction

BresNet: Applying Residual Learning in Backpropagation Neural Networks to Predict Ground Surface Concentration of Primary Air Pollutants

Improving Prediction of Marine Low Clouds with Cloud Droplet Number Concentration and a Deep Learning Method

Advancing Spatiotemporal Pollutant Dispersion Forecasting with an Integrated Deep Learning Framework for Crucial Information Capture

Real-time early warning and the prediction of air pollutants for sustainable development in smart cities

A hybrid deep learning model for regional O3 and NO2 concentrations prediction based on spatiotemporal dependencies in air quality monitoring network

Wide and Deep Learning Model for Satellite-Based Real-Time Aerosol Retrievals in China

Deep Transfer Learning on Satellite Imagery Improves Air Quality Estimates in Developing Nations

The Capability of Deep Learning Model to Predict Atmospheric Compositions Across Spatial and Temporal Domains

Design a regional and multistep air quality forecast model based on deep learning and domain knowledge

Deep learning the atmospheric boundary layer height

Air pollutant prediction model based on transfer learning two-stage attention mechanism

A Self-Organizing Memory Neural Network for Aerosol Concentration Prediction