Aims and Objectives
This group explores various issues related to climate and sustainable development of the earth with the help of Machine Learning and Data Sciences. The grand aim of the group is to develop a model of the world (with a special focus on India), through which we can simulate the impact of global climate change on regional weather and climate, human settlements, agricultural productions, environment and health. Development of such an integrated simulation framework involves three major vertical components:
V1) Detailed and accurate mathematical representation of Earth System processes (eg. causal relationships)
V2) Large-scale, high-resolution and highly accurate Earth System observation (eg. accurate mapping of human activities and environmental parameters)
V3) Development of efficient Machine Learning-based models that can augment or replace existing process-based Earth System Models
Some of the horizontal topics or specific research questions (which may intersect with one or more of these verticals) that we have investigated or are currently investigating include:
H1) Identifying causal relationships between climatic variables
H2) Analysis of Earth System Simulations by process-based models like CMIP6 (for climate) or APSIM (for agriculture)
H3) Spatial/temporal downscaling and bias correction of simulations by process-based models
H4) Development of computationally efficient ML-based surrogate models (emulators) for process-based models
H5) Development of high-resolution and accurate maps of entities related to environment (eg. groundwater quality) and human activities (eg. agriculture, settlements, industrial productions)
The toolkits that we are using to solve the above vertical and horizontal challenges include state-of-the-art concepts, algorithms and models based on Machine Learning. We are keen to develop new ML models, while also using off-the-shelf models with necessary tweaks. Some of the ML concepts, algorithms and models that we have used or are currently using include:
M1) Spatial Neural models (eg. variants of CNNs)
M2) Spatio-temporal Deep Generative models (eg. GANs, VaEs, Diffusion Models)
M3) Causal and Graphical Models with Inference Algorithms
M4) Deep Learning-based Object Detection and Segmentation (eg. YOLO, SAM)
M5) Advanced learning paradigms (eg. Contrastive Learning, Multi-task and Meta-Learning)
M6) Explainable AI techniques (eg. GradCAM, Shap)
Research Fellows
1) Dr. Laxmi Pandey: Predictive Models for Groundwater Salinization (Postdoctoral Research Associate)
2) Ankita Manekar: Analysis of groundwater salinization in India (Contractual Research Staff)
3) Sumanta Chandra Mishra Sharma: Deep Learning for Downscaling and Bias Correction of Climatic Variables (PhD student, Thesis Submitted)
4) Deepayan Chakrabarty: Causality Analysis and Emulation of CMIP6 models (PhD student since 2019)
5) Anjali Raj: Machine Learning in Remote Sensing for Mapping Vulnerable Human Settlements and Industries (PhD student since 2020, jointly with Dr. Manjira Sinha)
6) Priyanka: Machine Learning in Remote Sensing for Crop Mapping and yield prediction (PhD student since 2021, jointly with Dr. Manjira Sinha)
7) Sudip Kumar Bhattacharya: Climate Model Emulation (PhD student since 2020, jointly with Prof. Jiaul Paik)
Ongoing Collaborations
1) Dr. Manjira Sinha (TCS Research) - Remote Sensing
2) Prof. Auroop R. Ganguly (Northeastern University, USA) - Climate Model analysis
3) Prof. Ravi Sundaram (Northeastern University, USA) - Explainable AI
4) Prof. B. N. Goswami (Cotton University, Guwahati) - Causality and Forecasting of Indian Monsoon, Climate Model Emulation
5) Drs. Bipin Kumar, Roxy Mathew and Subodh Saha (Indian Institute of Tropical Meteorology, Pune) - Climate Model analysis and bias correction
6) Dr. Meirav Cohen (Dead Sea and Arava Science Centre, Israel) - Groundwater Salinization
7) Prof. Abhijit Mukherjee (IIT Kharagpur) - Groundwater quality
8) Dr. Amit Kumar Srivasatava (Leibniz-Centre for Agricultural Landscape Research (ZALF), Germany) - Crop mapping and yield prediction
9) Dr. Mriganka Biswas (Indian Institute of Tropical Meteorology, Pune) - Industrial emission mapping
10) Prof. Somsubhra Chakraborty (IIT Kharagpur) - Crop yield prediction
11) Drs. Jagobondhu Hazra, Ranjini Guruprasad, Kamal Das (IBM India Research Lab) - Flood prediction, emission mapping
Publications
[J] Sumanta Chandra Mishra Sharma, Bipin Kumar, Adway Mitra, Subodh Kumar Saha (2024), Atmospheric Research, Deep learning-based bias correction of ISMR simulated by GCM
[J] Sumanta Chandra Mishra Sharma, Adway Mitra (2024), Neural Computing and Applications, MAUNet: A max-average neural network architecture for precipitation downscaling
[J] Sumanta Chandra Mishra Sharma, Adway Mitra (2022), Environmental Data Science, ResDeepD: A Residual Super-Resolution Network for Deep Downscaling of Daily Precipitation over India
[J] B. N. Goswami, Deepayan Chakraborty , P. V. Rajesh and Adway Mitra (2022), NPJ Climate and Atmospheric Science. Predictability of South-Asian monsoon rainfall beyond the legacy of Tropical Ocean Global Atmosphere program (TOGA)
[J] Adway Mitra (2021), Frontiers in Climate (Predictions and Projections). A Comparative Study on the Skill of CMIP6 Models to Preserve Daily Spatial Patterns of Monsoon Rainfall Over India
[J] Madhumita Chakraborty, Soumyajit Sarkar, Abhijit Mukherjee, Mohammad Shamsudduha, Kazi Matin Ahmed, Animesh Bhattacharya , Adway Mitra (2020), Science of the Total Environment [Impact Factor: 6.6]. Modeling regional-scale groundwater arsenic hazard in the trans-boundary Ganges River Delta, India and Bangladesh: infusing physically-based model with supervised machine learning
[C] Priyanka, Adway Mitra, Manjira Sinha (2024). ACM International Conference on Data Science and Management of Data (CODS-COMAD). SepHRNet: Generating High-Resolution Crop Maps from Remote Sensing imagery using HRNet with Separable Convolution
[C] Deepayan Chakraborty, Adway Mitra (2024), International Conference on Climate Informatics, Simulation of Global Sea Surface Temperature Maps using Pix2Pix GAN
[C] Priyanka Goyal, Adway Mitra, Manjira Sinha (2024), IEEE International Geoscience and Remote Sensing Symposium (IGARSS), ASYMM-UNET: A CONCISE MODEL FOR SEGMENTATION OF SATELLITE IMAGERY FOR CROP MAPPING (short paper)
[C] Anjali Raj, Shubham Agarwal, Adway Mitra, Manjira Sinha (2023), IEEE International Geoscience and Remote Sensing Symposium (IGARSS), MAPPING SLUMS FROM SATELLITE IMAGERY USING DEEP LEARNING (short paper)
[C] Sumanta Chandra Mishra Sharma, Adway Mitra (2022), International Conference on Climate Informatics, ResDeepD: A Residual Super-Resolution Network for Deep Downscaling of Daily Precipitation over India
[C] Adway Mitra (2020), International Conference on Climate Informatics. A Hybrid Deep Generative Approach to Simulate Spatial Patterns of Daily Temperature and Rainfall
Workshop Presentations
4 presentations in AGU24!
i) Sumanta Chandra Mishra Sharma: "Deep Learning in Bias Correction of Daily and Seasonal Precipitation Estimates Over India" (session: Hydrology—Surface Hydrology)
ii) Deepayan Chakraborty: "Simulation of Monthly Global Sea Surface Temperature Data using Ensemble Pix2Pix Conditional GAN"
(session: AI-Driven Innovations in Earth and Climate Sciences)
iii) Anjali Raj: "Satellite-Based Monitoring of Industrial Emissions: A Contrastive Learning Approach for Climate Change Mitigation"
(session: Global Environmental Change)
iv) Priyanka: "Transformative Agricultural Monitoring: Enhancing Crop Classification with Machine Learning"
(session: Global Environmental Change)
2 presentations in EGU22
i) Deepayan Chakraborty: "Identification of Global Drivers of Indian Summer Monsoon using Causal Inference and Interpretable AI"
ii) Sumanta Chandra Mishra Sharma: "Super-Resolution based Deep Downscaling of Precipitation"
Pedagogical Activities
I have designed a course called "Machine Learning for Earth System Science", one of the earliest such courses in India, if not the world. In this course, I start with the basics of environmental data sciences and then shift to various case studies related to the application of AI/ML for earth observations, climate, hydrology etc. The course was offered three times in IIT Kharagpur as an elective. It was subsequently recorded as a MOOC in India's NPTEL program.
Link to the MOOC: https://archive.nptel.ac.in/courses/106/105/106105238/
Events
1) Adway Mitra and Dr. Amit Kumar Srivastava hosted R-FARM: Indo-German Workshop on Resilient Food Systems with AI, Remote Sensing and Crop Models in Harmony at University of Bonn, Germany during 21-22 February 2024, funded by IGSTC (Indo-German Science and Technology Council)
2) Adway Mitra and Prof. Auroop R. Ganguly hosted COEAI-SPARC Workshop on Hybrid Physics-AI Models for Climate, Weather and Water at IIT Kharagpur during 19-21 June 2024, funded by Ministry of Education under SPARC program