Could satellite measurement of CO help us to better understand sources of atmospheric CO2?
Seasonal pattern of spatial average (CO/CO2 measurement & GFED)
Presented are seasonal patterns of spatial average of measurement of CO and CO2 for three regions. Each point in the figure represents the average value for specific month in each region. The size of the points represents the number of data used for the calculation (missing data are caused either by limited observations of satellite or poor quality of retrievals). Smaller the point, more is the missing data for the specific month in the region. Therefore, larger point represents a more persuasive result than the smaller point.
For the three regions, seasonal patterns are detected in the MOPITT CO measurement and SCIAMACHY CO2 measurement, the former starts to decrease in late spring while the latter decreases after summer. However, we hardly see a clear pattern in CO column measurement from SCIAMACHY datasets, probably due to the large amount of missing data after April. .
When comparing the amount of CO in three regions, the figures suggest that North Africa gets the lowest while central China gets the highest. As to CO2 concentration, the central US shows the lowest while North Africa gets the highest.
For the three regions, seasonal patterns are detected in the MOPITT CO measurement and SCIAMACHY CO2 measurement, the former starts to decrease in late spring while the latter decreases after summer. However, we hardly see a clear pattern in CO column measurement from SCIAMACHY datasets, probably due to the large amount of missing data after April. .
When comparing the amount of CO in three regions, the figures suggest that North Africa gets the lowest while central China gets the highest. As to CO2 concentration, the central US shows the lowest while North Africa gets the highest.
Presented are the seasonal patterns of spatial average of fire emission data (GFED) for the three regions. It is clear that North Africa has the highest CO and CO2 emission from fires, especially in winter and spring, which agrees to the MOPITT CO measurement, where highest CO measurement in spring among the three regions also appears in North Africa.
The seasonal patterns for the three regions are quite different. In central US, fire emission increases in late summer, while in central China, fire emission starts to increase in late spring.
Another interesting finding is that, the seasonal patterns of CO and CO2 from fire emission datasets are almost the same. As we mentioned previously, GFED datasets are based on satellite data combined with biogeochemical model, so how strong is the correlation between CO and CO2 values in GFED datsets?
The seasonal patterns for the three regions are quite different. In central US, fire emission increases in late summer, while in central China, fire emission starts to increase in late spring.
Another interesting finding is that, the seasonal patterns of CO and CO2 from fire emission datasets are almost the same. As we mentioned previously, GFED datasets are based on satellite data combined with biogeochemical model, so how strong is the correlation between CO and CO2 values in GFED datsets?
The figures above answer the question: how strong is the correlation between CO and CO2 emission from GFED?
The left figure shows the correlation coefficents of GFED CO and CO2 data in each month with all the zeros removed (zeros in GFED means no fires detected). The right figure uses the same method except retaining all the zeros.
The correlation between GFED CO and CO2 emission is very high, even only for the areas with detected fires.
The left figure shows the correlation coefficents of GFED CO and CO2 data in each month with all the zeros removed (zeros in GFED means no fires detected). The right figure uses the same method except retaining all the zeros.
The correlation between GFED CO and CO2 emission is very high, even only for the areas with detected fires.
Figures above show the monthly correlation coefficients between fire emission datasets and satellite measurement of CO and CO2 during the whole year. Large p value is indicated by solid circle, and again, size of the point represents the numbers of data used for calculation. Missing data only comes from satellite measurement for the reason that GFED has no missing data (zeros in GFED mean no fires detected).
The figures suggest that in North Africa, strong correlation between MOPITT CO measurement and fire emission data is detected in winter and spring, while CO2 concentration has a poor relationship with the fire emission datasets.
As to Central US and China, the correlation coefficients are too low to reach any conclusion.
The figures suggest that in North Africa, strong correlation between MOPITT CO measurement and fire emission data is detected in winter and spring, while CO2 concentration has a poor relationship with the fire emission datasets.
As to Central US and China, the correlation coefficients are too low to reach any conclusion.
Spatial correlation of annual average (CO/CO2 measurement & EDGAR)
In the following we analyze the relationship between anthropogenic emission and satellite measurement of CO/CO2.
Due to the coarse temporal resolution of EDGAR, the monthly measurement of CO and CO2 is averaged for each grid. The method used here is to calculate the Pearson's product moment correlation coefficient, as well as analyzing the variograms and cross-variograms of different variables.
Due to the coarse temporal resolution of EDGAR, the monthly measurement of CO and CO2 is averaged for each grid. The method used here is to calculate the Pearson's product moment correlation coefficient, as well as analyzing the variograms and cross-variograms of different variables.
Presented are the correlation coefficients between different variables. In all three regions, relatively strong correlation between CO measurement of MOPITT and SCIAMACHY can be detected.
For the relations between CO and CO2, weak or even negative correlations are showed in central US and China, while in North Africa slightly stronger correlation appears.
The interesting finding is that EDGAR (anthropogenic emission) has a relatively stronger correlation with CO concentration measurement rather than with CO2, especially in central US and central China. Moreover, MOPITT CO measurement has a stronger correlation with EDGAR data than SCIAMACHY measurement of CO in central US.
For the relations between CO and CO2, weak or even negative correlations are showed in central US and China, while in North Africa slightly stronger correlation appears.
The interesting finding is that EDGAR (anthropogenic emission) has a relatively stronger correlation with CO concentration measurement rather than with CO2, especially in central US and central China. Moreover, MOPITT CO measurement has a stronger correlation with EDGAR data than SCIAMACHY measurement of CO in central US.
Variograms and cross-variograms (CO/CO2 measurement & EDGAR)
Here are the variograms and cross-variograms for CO/CO2 measurement and EDGAR datasets for North Africa. The lag distance in kilometers is calculated based on the gridded geolocations, therefore, the distance between two points at the neighboring longitudes will change with their latitude . The angle tolerance used here is 5 degree. Variograms and cross-variograms are calculated at two directions.
As to the variograms, CO measurement from MOPITT shows the clearest layering distribution in east direction due to the fact that north-direction variogram is above the east-direction one, while the same pattern can also be found in SCIAMACHY measurements of CO and CO2. EDGAR data should follows the sparse or random distribution for the variogram almost stay the same with increasing lag distance in both two directions.
As to the cross-variograms, the correlations between CO measurement and EDGAR decrease smoothly with distance (based on the inverse relationship between variograms and correlations). This feature can also be detected between CO measurement and CO2 measurement from SCIAMACHY.
As to the variograms, CO measurement from MOPITT shows the clearest layering distribution in east direction due to the fact that north-direction variogram is above the east-direction one, while the same pattern can also be found in SCIAMACHY measurements of CO and CO2. EDGAR data should follows the sparse or random distribution for the variogram almost stay the same with increasing lag distance in both two directions.
As to the cross-variograms, the correlations between CO measurement and EDGAR decrease smoothly with distance (based on the inverse relationship between variograms and correlations). This feature can also be detected between CO measurement and CO2 measurement from SCIAMACHY.
The same method used in North Africa above applies to central US.
As to the variograms, CO measurement from MOPITT shows slightly layering distribution in north direction due to the fact that north-direction variogram is below the east-direction variogram after certain lag distances, while again, the variogram of EDGAR data almost stays the same with increasing lag distance.
As to the cross-variograms, the correlations between CO measurement from MOPITT and EDGAR decrease with distance. The cross-variogram between CO and CO2 measurement is negative.
As to the variograms, CO measurement from MOPITT shows slightly layering distribution in north direction due to the fact that north-direction variogram is below the east-direction variogram after certain lag distances, while again, the variogram of EDGAR data almost stays the same with increasing lag distance.
As to the cross-variograms, the correlations between CO measurement from MOPITT and EDGAR decrease with distance. The cross-variogram between CO and CO2 measurement is negative.
Represented are the variograms and cross-variograms for central China.
For the variograms, CO measurement from MOPITT and SCIAMACHY show north layering distribution. while the correlations between data pairs in EDGAR data do not change with lag distance.
As to the cross-variograms, the correlations between CO measurement and EDGAR decrease smoothly with distance along the east direction.
For the variograms, CO measurement from MOPITT and SCIAMACHY show north layering distribution. while the correlations between data pairs in EDGAR data do not change with lag distance.
As to the cross-variograms, the correlations between CO measurement and EDGAR decrease smoothly with distance along the east direction.
