Forecasting Interannual Space-based CO2 Concentration using Geostatistical Mapping Approach

Abstract

NASAs Orbiting Carbon Observatory-2 (OCO-2) is a recent satellite mission primarily aimed at measuring the column concentration of the carbon dioxide (CO2) in the atmosphere. The atmospheric CO2 concentration is measured as continuous swaths of the parallelogram footprints which are available as the Level-2 samples with a swath width of 10.3 km approximately. The temporal frequency of the retrieval at one place is 16 days approximately. This work attempts to forecast the OCO-2 samples at an interannual time scale from the available past samples at the nearby locations using geostatistical spatiotemporal kriging-based mapping approaches. This forecasting is needed to understand the future seasonal behavior of CO2 beforehand. For the validation, we have used the XCO2 swaths of OCO-2 in a study region from 2015 - 2019 and foretasted in the year of 2018 and 2019. One of the variant approaches found to produce 1.52 ppm root mean square error (RMSE), which is a good result with limited samples. This approach is capable of spatio-temporal prediction and forecasting of other products of OCO-2 and might be improved further by considering correlated auxiliary variables in the study region.