Development of mechanistic energy balance models for grassland

Abstract

Grasslands are responsible for a significant percentage of the world's food supplies and, due to their large geographical extent, have an important influence on global climate. Major determinants of grassland productivity include near-surface soil and air temperatures and plant water use. Therefore it is important to understand how grasslands partition the energy they receive into heat and vapour fluxes, and soil heat flux which regulates the underlying soil temperatures. The project described in this thesis was aimed at measuring and modelling the energy balance at two contrasting grassland sites, with special emphasis on predicting soil temperature, and assessing the use of the models for determining plant water stress, dry matter production and impacts of climate change.

Two-layer models were developed, solving simultaneously for the energy balances at a) the soil surface and b) an equivalent canopy layer. Two distinct version of this basic two-layer model were developed. The simpler non-interacting (NACT) model assumes the heat and vapour fluxes from the soil surface transfer directly to the atmosphere, with no interaction with the canopy. The interacting (IACT) version exhibits diffusive exchange and hence greater coupling between the canopy and soil surfaces. However, both IACT and NACT versions are radiatively interacting, i.e. include full shortwave and longwave radiative exchanges between the two layers. These versions are further divided into sub-versions with constant (average) and diurnally-varying windspeed as inputs.

Key features of the models are as follows: retention of mechanistic simplicity, which aids intuitive understanding, as well as being 'reductionist'; the description of multi-day (typically 7-day) average diurnal variations, i.e. of the 'climate' rather than the detailed day-to-day 'weather'; and the use of analytical rather than numerical solutions (which is assisted by the previous two features). The models described in this thesis combine aspects of previous energy balance models, including those developed for 1) estimating evapotranspiration from 'sparse canopies', and 2) determining the moisture and thermal regimes below mulched surfaces.

The two sites used for model development and validation were as follows. Site 1 was an intensive pasture located on the Canterbury Plains and site 2 a high country site with a mixture of tussock and introduced species such as browntop. A set of climatic instruments was used at both sites, including a solarimeter, net radiometer, wet and dry bulbs in a Stevenson screen, and anemometer, with half-hourly averages recorded by a datalogger. Sets of soil temperature profiles and soil heat flux measurements were also recorded, along with periodic data sets for canopy temperature, and heat and vapour fluxes. Pertinent physical parameters of the grassland system were measured regularly and/or derived from other system parameters.

Values for the aerodynamic and canopy resistances, and the 'coupling factor' were obtained at both sites and compared. The canopy resistance (rc) was correlated with soil water content (r² = 0.3 and 0.69 for sites 1 and 2 respectively). At low water contents and at both sites, rc showed a degree of sensitivity to air saturation deficit. The average daytime magnitudes of rc were 71.2 and 83.5 s m⁻¹ for sites 1 and 2 respectively. The aerodynamic resistance for heat and vapour transport ranged between 45 - 75 s m⁻¹ and 20 - 60 s m⁻¹ for sites 1 and 2 respectively, showing the higher degree of roughness at site 2. The coupling factor (Ω) was found to have a weakly positive relationship with soil water content, with site 1 and 2 averages of 0.52 and 0.43 respectively. These values for Ω, mid-way between its extremes of 0 and 1, show a combination of energy-limited and plant-controlled evapotranspiration rate (Et).

Model performances with regard to the main model outputs (soil surface temperature, Ts, and canopy temperature, Tc) were compared with the help of a sensitivity analysis. Overall the performances were reasonably good, with the average errors less than 1.2 K for Ts and 1.4 K for Tc. Comparison of the measured and modelled soil temperatures showed that the IACT versions performed better at site 1, while all models performed similarly at site 2. Uncertainties in the moisture status of the soil surface, and the in-canopy resistance parameters were considered as the main causes of predictive error. In estimating Tc, the models performed similarly at site 1, however the NACT versions outperformed IACT versions at site 2. The neglect of free moisture on the canopy and of non-neutral atmospheric stability were proposed as the two main factors contributing to errors in modelling Tc. Soil heat flux at 5 cm depth was modelled reasonably well at site 1 but poorly at site 2, due partly to the stratified nature of the soil surface layers at site 2, and partly to large contrast in thermal conductivities between the heat flux plates and the surrounding soil.

Reasonable relationships were found at both sites between maximum daylight (between the hours of 1 and 4pm) canopy-air temperature difference (ΔTac) and a) soil water content and b) evapotranspiration, Et. The models were able to predict ΔTac well while the atmosphere was not too unstable, a condition which occurs typically in early to mid spring and mid to late autumn.

An annual average value for water-use efficiency at site 1 was derived viz 11.5 kg ha⁻¹ mm⁻¹. The models predicted well the cumulative evapotranspiration for periods between 1-3 weeks in duration, and hence should be capable of predicting dry matter production via use of the WUE. However in the context of future climate change, the possible effects on the WUE of increased atmospheric CO₂ concentration are not well known. Hence grassland productivity changes are equally poorly known. However, the increased temperatures and decreased rainfalls forecast for the Canterbury region, are likely to increase plant water-stress through simultaneous increase in evaporative demand, and decrease in soil water supply.