Extending the Time Series of Satellite-Based Lightning Observations
We used 15 years of OTD/LIS monthly lightning flash count data with 2.5 by 2.5 degree spatial resolution. Starting with North America, we examined the correlations in total lightning at different latitudes with scatter plots of total lightning at any given OTD latitude (north of 40N) against its most correlated OTD/LIS latitude (south of 40N). From this plot, we determined the best-fit line and use this linear model to extrapolate the time series from 2000 to present day for latitudes north of 40N. The model calculates total lightning at latitudes north of 40N by using data available through the present day at latitudes south of 40N. We distributed the modeled lightning in the east-west direction using the 1995-2000 mean monthly longitudinal distribution of lightning counts. The final product is a 2.5 x 2.5 degree, monthly lightning count time series for North America from 1995 to present day.
For 83% of latitudes, the linear model provided a statistically significant (p < 0.05) fit to the data and, on average, captured 76 +/- 22% of the variability in the data. Furthermore, 67% of the latitudes exhibited model uncertainty of less than 40% as compared to the 5 years of OTD observations. In North America, most of the lightning north of 40N tends to occur close to 40N, and our linear model performed best at these latitudes. Our model is able to capture most of the interannual variability of lightning north of 40N using the relationship between OTD and LIS observations. To reduce model uncertainties and to better capture interannual variability, we are testing additional model forms. Preliminary results suggest the model uncertainties are reduced by using a power-law function to model the relationship between lightning at different latitudes. We are beginning to test the methods for lightning in boreal Russia.
From this research, a completed time-series of lightning count data for North America will be made available for applications such as modeling global and regional fire distributions and improving the input to chemical transport models. For example, because lightning data is limited to March 2000, global fire models currently use the mean monthly climatology of lightning to simulate interannual variability in fire activity. In boreal regions of Canada and Russia where lightning acts as the primary natural source of fire ignitions, a mean monthly climatology may miss significant peaks in fire activity that a full time series would not. The lightning dataset will also be useful for chemical transport models because lightning is the primary natural source of nitrogen oxides (NOx) in the atmosphere and NOx is important for air quality and ozone depletion.