speaker

Dr. Uttam Kumar

Assistant Professor

International Institute of Information Technology Bangalore

email: uttam[at]iiitb.ac.in

Brief Introduction

Dr. Uttam Kumar holds a Bachelor’s Degree in Computer Science from VTU Belgaum, Diploma in Advanced Computing from CDAC Pune, Master’s Degree in Geoinformation Science from University of Twente, The Netherlands and Ph.D. in Algorithms for Geospatial Data Analysis from Indian Institute of Science (IISc), Bangalore. He moved to NASA (National Aeronautics and Space Administration) Ames Research Center, California, USA as a NASA Postdoctoral Fellow and subsequently served as a Visiting Scientist at NASA Ames/USRA for almost 4 years.

He has published 30 research papers in national/international journals, 6 book chapters, 48 papers in conference proceedings and 8 technical reports. He has delivered 53 invited guest lectures and is the reviewer of 23 national and international scientific journals.

He has received several awards including The Institute of Engineers India (IEI) Young Engineers Award 2016; NASA Group Achievement Award by NASA, Washington, D.C.; Certified Sentinel of Science Award 2016 in the Earth and Planetary Sciences; 5 Best Paper Awards and the Young GeoSpatial Scientist Award 2011, New Delhi.

Abstract

Subpixel Learning Algorithms for Estimating Global Land Cover Fractions from Satellite Data Using High Performance Computing: The physical surface of the landscape is undergoing transformation either naturally or due to human interference, giving rise to Land Cover Land Use Change (LCLUC). Expanding urban regions and consequent LCLUC have emerged as one of the major anthropogenic sources of global environmental degradation, bringing numerous stresses to landscapes, vegetation, natural habitats, soil, air, water, etc. LCLUC at a sub-continent or country level can be monitored through high temporal and low spatial resolution data, such as those obtained from Landsat at 30 m or MODIS Terra/Aqua at 250 m spatial resolution. These satellites improve the ability to map large areas of Earth’s surface quickly due to their wider IFOV and inexpensively. However, different land cover (LC) types jointly occupy a single pixel, and the resulting spectral measurement is a composite of the individual spectra. The intrinsic scale of spatial variation in LC is usually finer than the scale of sampling imposed by the image pixels. Due to scale-resolution mismatch, the spatial resolution of the details on the ground is less than what is required, leading to sub-pixel heterogeneity, imposing limitations in ecological modelling with these data sets. The talk will present some attempts to resolve the mixed pixel problem.