Geostatistical Analysis of CH 4 Columns over Monsoon Asia Using Five Years of GOSAT Observations

The aim of this study is to evaluate the Greenhouse gases Observation SATellite (GOSAT) column-averaged CH4 dry air mole fraction (XCH4) data by using geostatistical analysis and conducting comparisons with model simulations and surface emissions. Firstly, we propose the use of a data-driven mapping approach based on spatio-temporal geostatistics to generate a regular and gridded mapping dataset of XCH4 over Monsoon Asia using five years of XCH4 retrievals by GOSAT from June 2009 to May 2014. The prediction accuracy of the mapping approach is assessed by using cross-validation, which results in a significantly high correlation of 0.91 and a small mean absolute prediction error of 8.77 ppb between the observed dataset and the prediction dataset. Secondly, with the mapping data, we investigate the spatial and temporal variations of XCH4 over Monsoon Asia and compare the results with previous studies on ground and other satellite observations. Thirdly, we compare the mapping XCH4 with model simulations from CarbonTracker-CH4 and find their spatial patterns very consistent, but GOSAT observations are more able to capture the local variability of XCH4. Finally, by correlating the mapping data with surface emission inventory, we find the geographical distribution of high CH4 values correspond well with strong emissions as indicated in the inventory map. Over the five-year period, the two datasets show a significant high correlation coefficient (0.80), indicating the dominant role of surface emissions in determining the distribution of XCH4 concentration in this region and suggesting a promising statistical way of constraining surface CH4 sources and sinks, which is simple and easy to implement using satellite observations over a long term period.