Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Carbon Emission Reduction Strategies using OCO and ICOS data

Oskar Åström, Carina Geldhauser, Markus Grillitsch, Ola Hall, Alexandros Sopasakis

Abstract

We propose a methodology to enhance local CO2 monitoring by integrating satellite data from the Orbiting Carbon Observatories (OCO-2 and OCO-3) with ground level observations from the Integrated Carbon Observation System (ICOS) and weather data from the ECMWF Reanalysis v5 (ERA5). Unlike traditional methods that downsample national data, our approach uses multimodal data fusion for high-resolution CO2 estimations. We employ weighted K-nearest neighbor (KNN) interpolation with machine learning models to predict ground level CO2 from satellite measurements, achieving a Root Mean Squared Error of 3.92 ppm. Our results show the effectiveness of integrating diverse data sources in capturing local emission patterns, highlighting the value of high-resolution atmospheric transport models. The developed model improves the granularity of CO2 monitoring, providing precise insights for targeted carbon mitigation strategies, and represents a novel application of neural networks and KNN in environmental monitoring, adaptable to various regions and temporal scales.

Enhancing Carbon Emission Reduction Strategies using OCO and ICOS data

Abstract

We propose a methodology to enhance local CO2 monitoring by integrating satellite data from the Orbiting Carbon Observatories (OCO-2 and OCO-3) with ground level observations from the Integrated Carbon Observation System (ICOS) and weather data from the ECMWF Reanalysis v5 (ERA5). Unlike traditional methods that downsample national data, our approach uses multimodal data fusion for high-resolution CO2 estimations. We employ weighted K-nearest neighbor (KNN) interpolation with machine learning models to predict ground level CO2 from satellite measurements, achieving a Root Mean Squared Error of 3.92 ppm. Our results show the effectiveness of integrating diverse data sources in capturing local emission patterns, highlighting the value of high-resolution atmospheric transport models. The developed model improves the granularity of CO2 monitoring, providing precise insights for targeted carbon mitigation strategies, and represents a novel application of neural networks and KNN in environmental monitoring, adaptable to various regions and temporal scales.
Paper Structure (13 sections, 3 equations, 7 figures, 1 table, 1 algorithm)

This paper contains 13 sections, 3 equations, 7 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. State of the art
  3. Challenges in CO2 emissions estimation
  4. OCO-2 and OCO-3: Advances and Challenges in CO2 Monitoring
  5. Materials and Methods
  6. Datasets
  7. Machine Learning Models
  8. Results
  9. Ground Level CO2 Estimation
  10. Multi-scale Predictions
  11. Feature Importance
  12. Discussion
  13. Conclusions

Figures (7)

  • Figure 1: Measurements from the Hyltemossa ICOS station (blue) and nearby OCO observations (orange) within a 25km radius for the time frame of data availability.
  • Figure 2: In blue, difference between the actual ICOS and actual OCO measurements. In red, difference between actual ICOS and predicted ICOS. Predictions are produced by the neural network regression model. The dashed lines highlight that most of the predicted values lie within an error of $\pm$5ppm.
  • Figure 3: Correspondence plot between the predicted and true CO2 levels for the test dataset. The x-axis shows the predicted CO2 level from each respective model, while the y-axis shows to the true ICOS measurement. The locations for each station's acronym listed in the legend can be found at https://www.icos-cp.eu/observations/station-network.
  • Figure 4: Interpolated predictions for Europe’s 2015 ground level CO2 concentrations using different parameterizations of the weighted K-nearest neighbor interpolation. The ablation study shows results as the nearest neighbor parameter $K$ ranges from 10, 200, 1000, to $\infty$ (left to right), while the decay rate $p$ is varied through 1, 0.2, and 0 (top to bottom). See Algorithm \ref{['alg:interpolation']} for details.
  • Figure 5: Predictions on the municipality of Lund, in southern Sweden, using the neural network regression model presented in Section \ref{['sec:Models']}. The KNN interpolation is made using $K=100$ and $p=0.05$. The estimated CO2 levels are highest close to the city of Lund (top left) and further south near the airport (center bottom) at the city of Malmö while they are lowest around the Häckeberga natural park reserve (middle right).
  • ...and 2 more figures