Gas dispersion trials: a surface area source enclosed by a windbreak

摘要

As a simple alternative to direct measurements, it may sometimes be useful to diagnose the strength (Q) of a finite surface area source of a trace gas, from nearby measurement(s) of concentration (C) at a point P. This is sometimes called "inverse dispersion", and requires the use of a suitable dispersion model, which must be provided measurements of the atmospheric state: at a minimum, the friction velocity u_*, Obukhov length L, roughness length Z_0, and mean wind direction β. The procedure could be represented symbolically as: Q~((est)) = Q~((est)) (C_P|u_*, K, z_0, β) (1) A particularly flexible instance of this approach was introduced by Flesch et al. (1995), based on (an ensemble of N) backward traiectories. Q~((est)) = U C_P/ n n = 1/ N ∑_i 2/w_(0i)/U where U is a reference windspeed (whose explicit inclusion renders the "magic number" n dimensionless). The summation runs over all "touchdowns" of trajectories on the source, and w_(0i) is the magnitude of the vertical velocity of the ith touchdown. We performed experiments to test the accuracy of the backwards Lagrangian stochastic ("bLS") procedure, by releasing methane from a 6m x 6m source on ground, and detecting line-average concentration nearby using lasers. Two papers at this conference describe the outcome: paper 9.8 covers the case where the source was on open terrain (undisturbed winds), and this paper covers the case where the source lay within the windbreak described 'in paper 2.5. Here we enquire whether the naive use of inverse dispersion (eqns 1, 2), using a dispersion model appropriate (only) for undisturbed surface layer winds, would have some value even if applied to estimate a source in a region of disturbed winds (Fig. 1). It is important to emphasize that, even in the horizontally-uniform case, source diagnosis for short periods (15-30 mins) by "inverse dispersion" carries an uncertainty of (roughly) +- 25% (the bias when successive short-term estimates are summed is smaller). This is because the micro-meteorological state (deduced from observations) and the dispersion model are built on a set of assumptions about the atmosphere: eg. Monin-Obukhov "universal" functions φ_m, φ_h for wind and temperature profiles, the ratio σ_w/u_*, and other "universal" constants.