A new method to detect long term trends of methane (CHsub4/sub) and nitrous oxide (Nsub2/subO) total columns measured within the NDACC ground-based high resolution solar FTIR network

摘要

pstrongAbstract./strong Total columns measured with the ground-based solar FTIR technique are highly variable in time due to atmospheric chemistry and dynamics in the atmosphere above the measurement station. In this paper, a multiple regression model with anomalies of air pressure, total columns of hydrogen fluoride (HF) and carbon monoxide (CO) and tropopause height are used to reduce the variability in the methane (CHsub4/sub) and nitrous oxide (Nsub2/subO) total columns to estimate reliable linear trends with as small uncertainties as possible. The method is developed at the Harestua station (60?° N, 11?° E, 600 m a.s.l.) and used on three other European FTIR stations, i.e. Jungfraujoch (47?° N, 8?° E, 3600 m a.s.l.), Zugspitze (47?° N, 11?° E, 3000 m a.s.l.), and Kiruna (68?° N, 20?° E, 400 m a.s.l.). Linear CHsub4/sub trends between 0.13 ?± 0.01-0.25 ?± 0.02 % yrsup−1/sup were estimated for all stations in the 1996-2009 period. A piecewise model with three separate linear trends, connected at change points, was used to estimate the short term fluctuations in the CHsub4/sub total columns. This model shows a growth in 1996a??1999 followed by a period of steady state until 2007. From 2007 until 2009 the atmospheric CHsub4/sub amount increases between 0.57 ?± 0.22a??1.15 ?± 0.17 % yrsup−1/sup. Linear Nsub2/subO trends between 0.19 ?± 0.01a??0.40 ?± 0.02 % yrsup−1/sup were estimated for all stations in the 1996-2007 period, here with the strongest trend at Harestua and Kiruna and the lowest at the Alp stations. From the Nsub2/subO total columns crude tropospheric and stratospheric partial columns were derived, indicating that the observed difference in the Nsub2/subO trends between the FTIR sites is of stratospheric origin. This agrees well with the Nsub2/subO measurements by the SMR instrument onboard the Odin satellite showing the highest trends at Harestua, 0.98 ?± 0.28 % yrsup−1/sup, and considerably smaller trends at lower latitudes, 0.27 ?± 0.25 % yrsup−1/sup. The multiple regression model was compared with two other trend methods, the ordinary linear regression and a Bootstrap algorithm. The multiple regression model estimated CHsub4/sub and Nsub2/subO trends that differed up to 31 % compared to the other two methods and had uncertainties that were up to 300 % lower. Since the multiple regression method were carefully validated this stresses the importance to account for variability in the total columns when estimating trend from solar FTIR data./p.