Ten Days Online FDP on Advanced Remote Sensing and Machine Learning for
Environmental Sustainability
Session III : AI & ML Applications for agricultural management - Dr Anand S, Scientist D, Space Application Centre – ISRO, Government of India
Highlights of the Session - III
-
Remote Sensing data for agricultural management- Optical(hyperspectral, multispectral, LIDAR), Microwave(synthetic Aperture Radar), Thermal and their usage were discussed.
-
To integrate the entire idea into Machine Learning, represent the data in different forms such as spatial, spectral, feature space and temporal changes were explained graphically.
-
The need of Machine Learning was elaborated with the reasons such as crop yield prediction, price forecasts, etc.
-
In ML taxonomy , ML algorithms and its agricultural applications were discussed. Forexample, feature extraction is used to find the spectral index for vegetation, etc. Feature extraction was explained using Landsat-8 reflectance curves from X.Wang et al., 2018 also Modified Normalized Difference Water Index was calculated.
-
Segmentation/Clustering - K means clustering was explained with appropriate figures and algorithms.
-
Classification and Regression – its differences was elaborated. Model selection was briefed out with an example from a published journal paper.
-
Classification Accuracy measures were briefed by creating a confusion matrix for a water body area and validating it with formulae.
-
Multi-class classification and Regression accuracy measures were also demonstrated and the application of classification and regression were explained with mapping of chlorophyll.
-
Sub pixel detection, change detection, Temporal analysis, Data fusion and estimation of LAI using RT model was also elaborated.
-
Ml application on precision agriculture, future scope of Remote sensing and Machine Learning in agricultural management and finally steps for start-up was discussed.
Session IV : Cyber physical systems - Dr Deepak Mishra , University of Georgia
Highlights of the Session - IV
-
Dr.Deepak Mishra has presented a detailed lecture on solving environmental monitoring challenges in the age of big data, sensing integrated cyber-physical systems(CPS).
-
Relevance of CPS with long term environmental trends like acute events, potential synergistic and antagonistics, effects across ecosystems and identifying existing resources.
-
New tools and methods of CPS such as cross calibrated remote sensing, proximal or in-situ sensing, citizen science, developing data centric models, implementing modes on all available satellite data, disseminating on open source platforms such as google earth Engine, etc. were also discussed.
-
Basics of recent trends in sensing integrated CPS, integration of computational algorithms and physical components, developing core system science to engineer complex CPS, transitioning science and technology, etc.
-
Explained cyber innovated environmental sensing, water resources management, wetland health and productivity and sea level rise induced coastal flooding with appropriate examples.
-
Narrated how water quality management is carried out using CPS with the example of NSF CPS project: Cyano tracker, how to develop the early warning systems for phytoplankton, cyanobacteria and the effects of algal bloom on ecology, public health, economic and so on.
-
Showcased few images and discussed the case studies namely, Impacts of cyanobacterial blooms in lake Taiha, China and Gulf of Finland, Eye of an algal storm (Cyano HAB) using Sentinel-2A(10m) image, 2016 Toxic Algae bloom and fish kill in Indian river Lagoon using Sentinel 2A image.
-
Illustrated the pictures of Absorption of Chl-a and Phycocyanin commonly used in PC algorithms, Absorption spectra of Chl-a and Phycocyanin with the examples of 2016 Okeechobee bloom(June-July,2016), Nutrient pollution plus climate change and how it became a public crisis, Effect of Bonnet Carrie Spillway opening on water quality parameters.
-
Discussed the Plume and bloom concepts with the example of sediment plume in Mississippi River from April – June,2008, Multispectral satellite sensors (Chl-a using NDCI) and also explained them in detail.
-
Summarized the major objective of cyano tracker to combine social networks, he has listed out the problems with satellite based remote sensing such as, poor spatial and temporal resolutions, lack of 620nm band center, cloud cover/ atmospheric attenuation, etc crowd sourcing, advanced remote sensing and cloud enabled big data computing.
-
He has exhibited the architecture of Cyano tracker, which included computation cloud, community cloud and sensor cloud. He explained how community cloud is built using phone apps, flyers, social media, videos and sensor cloud which uses in-situ hyperspectral sensors, UAS integration, AquaBlinkR sensor that can measure temperature, salinity and other water quality parameters.
-
Insight into CubeSat Revolution, which is cheap and rapid, UGA SSRL- SPOC satellite and has described the computation cloud of Cyanotracker and its functions like development of robust models and the use of Quasianalytical Inversion Algorithm which can give the concentration of cyanobacteria in micrograms per litre.
-
He has listed out the operation earth observing Satellite sensors like Aqua, sentinel 2, sentinel 3, and cloud platforms such as global water quality dashboards. He has described about Phenocam and its uses like Flood flagging via vegetation Indices, upland forest mortality after hurricane Mathew in Oct.2016, wetland dieback research, Flooding frequency mapping using phenocam satellite model, etc.
-
Remote sensing of marsh biophysical characteristics, MC models, Eddy flux tower at Grand Bay for monitoring long term wetland carbon dynamics, Phenological variation in tidal wetland of Teerebonne Parish, LA from 2000 – 2014 were discussed.
-
Case study of acute event of Impact of COVID-19 Anthropause on coastal water Resources with a hypothesis stating the reasons as overall watershed nutrient, input increased from water amplified spring green up in marshes and the other hypothesis with reasons such as fluctuations in dissolved N (changes in atmospheric deposition, shifting waste water disposal) affected below and above ground marsh biomass.
-
Soil coring method with its steps such as, collection of soil cores, analysis of soil core in lab using ting sensors and satellite sensor datas, investigating low cost field sensor for soil properties (color, salinity, electrical conductivity, pH and redox potential), field deployment and developing SOC sensor platform.
-
Case study of New CPS project named, A sensor cloud base approach for hyperlocal mapping or urban heat hazards using SCOUTS ( Smart community
urban thermal sensing framework) , urban heat exposure measurement using human borne sensing on backpack and mobile apps and future applications of CPS systems.