Gap filling - replacing missing flux data from the Straits Flux monitoring station

 

Our aim is to run the measurement system constantly, but a certain amount of data loss is inevitable. The instillation of mains power in October 2005 has reduced gaps caused by battery failure however some data are still lost during system maintenance and through the quality control procedure.

A gap-filling procedure has been developed to replace missing data and enable carbon budgets to be prepared. This procedure is independent of the process model ForestFlux, since the model is dependent on flux data for its parameterisation. The gap filling procedure uses look-up tables, as described in Falge et al. (2001). Separate look up tables, based on temperature and global radiation, have been prepared for the summer and winter time periods of each year, thus accounting for inter-annual variability. Temperature is categorised as 2 degree bands over the range -6 to 30 degrees Celsius whilst global radiation is categorised as 50 W/m2 bands over the range 0-950 W m/m2. Gaps during the summer and winter data sets have filled directly from the corresponding lookup table.

Gap filling during Spring and Autumn

As a result of canopy dynamics during spring development and Autumn decline, it is not possible to derive a single look-up table for each period that can provide a comprehensive reference. An approach dependent on canopy structure (spring) and activity (autumn) has been developed. The period during which each approach was applied in 2003 is given in the table below.

Breakdown of canopy developmental period by day number for 2003

Canopy state	Day number
Canopy developing (Spring)	121-151
Full canopy (Summer)	152-273
Canopy decline (Autumn)	274-305
No Canopy (Winter)	306-365 & 1-120

Spring

During spring, the rate of canopy development is quantified as a function of mean daily global radiation interception, as measured by tube solarimeters located above and below the canopy). The data set was restricted to periods when the radiation above the canopy was > 300 W/m2. The points at which canopy development began (developmental index=0) and was complete (developmental index=1) are identified 'by eye' and the slope given by linear regression (period of development denoted by red data-points in the graph below).

Individual regressions are developed for each year. The developmental index is then applied to the difference between winter and summer look-up tables from the previous winter and following summer.

Autumn

In order to fill the gaps in the data-set during the period of canopy decline at the end of the growing season, a different approach is used. The is because the autumn decline in canopy activity is a result of physiogical rather than structural (and physiological) dynamics. The alternative method used in autumn is a function of mean daily carbon flux between 11.00-15.00 hrs. A linear regression (canopy development index vs day of year) is again fitted to the data-set and the slope used to modify the difference between summer and winter look-up table as shown in the graph below.