The Energy Balance Closure Problem

Abstract

talists often use the surface energy balance as a measure of the quality of their measurements, sometimes by examining the size of the ratio (H + AE)I(R n-G), often referred to as the "recovery ratio", or by calculating the sum (R n-G-H-AE).A dataset in which the energy balance does not close to within ±1O% is often considered to be unreliable, although, of course, an apparently good recovery ratio can mask large, compensating errors in the individual components of the balance. Atmospheric modellers, trying to use field data for calibration or validation have to deal with the inconsistency between the perfect energy closure demanded by the models and the poor closure often delivered by the data. There are many publications giving field results of energy balance closure. The differences in instrumentation, software, quality control, etc. used in the different experiments make a systematic review extremely difficult but a selection from various studies are presented here to illustrate the wide range of results that are obtained. The results from the first energy balance measurement campaigns were reported by Russian expeditions (Elagi-na et al. 1978; Elagina et al. 1973; Orlenko and Legotina 1973; Tsvang et al. 1987). They closed the energy balance to within 80% and suggested that problems with the eddy covariance measurements and horizontal advection were the probable reasons for the shortfall. Later, Bolle et al. (1993), Braud et al. (1993) and Foken (1990) showed that the storage of energy in the canopy and the upper soil layer were also contributing factors. Panin et al. (1996) combined the results of the first internationalland-surface experiments FIFE (Kanemasu et al. 1992), KUREX-88 (KURsk EXperiment; Tsvang et al.1991) and TARTEX-90 (TARTu EXperiment; Foken et al. 1993) and proposed that the difference in energy closure ranging from 90 to 60% might be caused by heterogeneities in the surrounding area. Recently, it has become possible to make energy balance measurements for much longer periods of several months to years. Greco and Baldocchi (1996) present a 290 (B.1) 280 • 250 260 170