In NERC-CEH/ukghg: Greenhouse Gas Fluxes from the UK

library(rmarkdown)
system.time(render("ukghg_timeDisaggn.Rmd", output_file = "ukghg_timeDisaggn.html"))

Introduction

A key aim of the NERC Greenhouse Gas Emissions and Feedbacks Programme is to compare bottom-up inventory estimates with top-down estimates from atmospheric inverse modelling, to deliver improved GHG inventories and analyses for the UK. The output required by the inverse modelling groups from "Deliverable D" WP1 is a data set (and/or algorithm) for daily anthropogenic and biogenic emission estimates of CO$_2$, CH$_4$, N$_2$O with a spatial resolution of 1 km$^2$ for the UK. This provides the prior estimate (spatially- and temporally-resolved) for the inverse modelling, and the bottom-up data set against which to compare top-down model results.

There are three greenhouse gases to consider (CO$_2$, CH$_4$, N$_2$O), but the following applies equally to all of them. We distinguish two types of fluxes, which we model separately: anthropogenic ($A$) and biogenic ($B$), and the total flux is the sum of these two components. Both $A$ and $B$ vary in space and time. Spatially, we represent the fluxes $A$ and $B$ as two-dimensional grids with 1-km resolution and covering the extent of the UK.

Anthropogenic GHG emissions

Data on anthropogenic GHG emissions in the UK come from national statistics collated by DECC, AEA and NAEI, and are divided into ten SNAP sectors. These sectors are: Sexpr{sectorName}. For each of these sectors, we have a single spatial map of annual emissions from a reference year (denoted $A_\mathrm{ref}$), so for the moment, we consider the relative spatial pattern for each sector to be fixed. The modelling applied here extrapolates and disaggregates emissions to time, $t$, so that we can predict a spatial map of emissions, $A$, for any sector in any hour, of any weekday, on any day, of any year. Within each sector, the model has the following form:

\begin{align} A_t =& A_\mathrm{ref} \times \alpha_t \ \text{which can be decomposed to:} \ \alpha_t =& \alpha_\mathrm{year} \, \times \, \alpha_\mathrm{yday} \, \times \alpha_\mathrm{wday} \, \times \, \alpha_\mathrm{hour} \ \ \text{where:} \ \ \alpha_\mathrm{year} =& \mathrm{f(year)} \ \alpha_\mathrm{yday} =& \mathrm{f(yday)} \ \alpha_\mathrm{wday} =& \mathrm{f(wday)} \ \alpha_\mathrm{hour} =& \mathrm{f(hour)}. \end{align}

$\alpha$ terms act as multipliers representing temporal variation on four time scales: inter-annual (year 1990 to 2016), intra-annual day-of-the-year (yday 1 to 365), within-week-day (wday 1 to 7), and hourly (hour 1 to 24). The sub-annual $\alpha$ functions f(yday), f(wday) and f(hour) are cubic splines. The sub-annual $\alpha$ functions have an average value of 1, i.e.:

\begin{align} \frac{1}{365}\sum_{j=1}^{365}\alpha_\mathrm{yday}(j) =& 1 \ \frac{1}{7} \sum_{d=1}^{7} \alpha_\mathrm{wday}(d) =& 1 \ \frac{1}{24} \sum_{h=1}^{24} \alpha_\mathrm{hour}(h) =& 1 \end{align}

so they so they decribe the temporal variation around the annual mean. $\alpha_\mathrm{year}$ accounts for the inter-annual variation in the annual mean, relative to the reference year. This model is applied to each sector independently, and summed to give the total anthropogenic flux.

\begin{align} A(t) = \sum_{i=1}^{n_\mathrm{sector}} A_{i,\mathrm{ref}} \times \alpha_{i,t} \end{align}

There are thus 40 parameters in this model: values of $\alpha$ for 10 sectors at four temporal scales.