Day 4

Detailed paper information

Back to list

Paper title Artificial Intelligence models in solving Ill-posed Inverse problems of Remote Sensing GHG emission
  1. Sergiy Sylantyev Space Research Institute National Academy of Sciences of Ukraine and State Space Agency of Ukraine Speaker
  2. Hanna Yailymova Educational and Research Institute of Physics and Technology NTUU "Igor Sikorsky Kyiv Polytechnic Institute"
  3. Andrii Shelestov National Technical University of Ukraine «Igor Sikorsky Kyiv Polytechnic Institute»
Form of presentation Poster
  • C1. AI and Data Analytics
    • C1.07 ML4Earth: Machine Learning for Earth Sciences
Abstract text Earth’s atmosphere and surface is undergoing rapid changes due to urbanization, industrialization and globalization. Environmental problems such as desertification, soil depletion, water shortages, greenhouse gas (GHG) emissions warming the atmosphere, are increasingly significant and troubling consequences of human activities. UNEP forecast that under current policy, GHG emission will reach 60 gigatons CO2 per year by 2030. On the COP26 António Guterres said, what “We must accelerate climate action to keep alive the goal of limiting global temperature rise to 1.5 degrees”, and it is time to go “into emergency mode”.
Before today a total of 33 relevant satellite missions with spectrometers like SAM, SAGE, GRILL, ATMOS, HALOE, POAM, GOMOS, MAESTRO was used for GHG monitoring capabilities from space underpinning the dynamic analysis and forecasting by solving ill-posed inverse problems on the bases GHG atmospheric measurement.
Most of the practice science problems of atmospheric measurement emission gases are formally reduced to the Fredholm integral equations of the first kind.
When numerically solving the Fredholm integral equation of a first kind, with which ill-posed inverse problems are associated, most of the problems of forecasting the dynamics of greenhouse gas emission, and other problems of forecasting the dynamics of atmospheric gases are reduced to solving a system of algebraic equations.
In most cases, direct calculation of the kernel function of the Fredholm integral equation is impossible due to the lack of information on the parameters of the interaction of the spectrometer with the atmospheric measurement environment.
As a consequence, the algorithm for solving the inverse ill-posed problem associated with forecasting the dynamics of greenhouse gas emission can be based on the use of machine learning and artificial intelligence methods.
Moreover, taking into account the stochastic nature of the behavior of both atmospheric parameters and the measurement errors of spectrometers in inverse ill-posed problems, it is necessary to search not a single solution, but the distribution of the probabilities of solutions.