Please login first
Trend assessment for a CO2 and CH4 data series in the North of Spain
* , , , ,
1  Faculty of Sciences. University of Valladolid. Department of Applied Physics, Paseo de Belén, 7. 47011 Valladolid (Spain)


The main objective of this paper is to implement different methods to assess the salient features of the data trend for a CO2 and CH4 data series. Said series was obtained at the Low Atmosphere Research Centre (41°48′49″ N, 4°55′59″ W) using a Picarro analyzer (G1301). Semi-hourly measurements from 15 October 2010 to 29 February 2016 were considered and divided into diurnal and nocturnal records.


Different functions with two terms, one for the trend and another for the annual cycle, were employed. The first was a harmonic function based on a third-degree polynomial and a series of four harmonics that considers the amplitude as a fixed and variable term over time. An increasing trend was reported, and was below 2.30 ppm year-1 for CO2 and below 11.90 ppb year-1 for CH4. Nocturnal amplitudes were higher than diurnal ones for both gases, except in winter due to anthropogenic emissions and lower assimilation rates. The second function was based on the kernel procedure. Epanechnikov, Gaussian, biweight, tricubic, rectangular and triangle kernels, were applied with a 500-day bandwidth for the trend. The best fit was obtained by the biweight kernel (r>0.38), with an increasing trend around 1.80 ppm year-1 for CO2 and around 7.15 ppb year-1 for CH4. The final analysis, which included local linear regression functions also applying a 500-day bandwidth, revealed increasing trends for both CO2, around 1.98 ppm year-1, and CH4, around 10.85 ppb year-1. Trend values were far more accelerated in the latter years of the series regardless of the chosen function.


Finally, the application of these functions to other terrestrial ecosystems would improve current knowledge of these gases in different environments, thereby increasing the effectiveness of climate change policies.

Keywords: trend, harmonic function, kernel functions, local linear regressions, daytime, night-time.