## Sensible heat flux and evaporation for sparse vegetation using temperature-variance and a dual-source model.

##### Abstract

The high population growth rate and rapid urbanization that the world is experiencing today has aggravated the competition for the already scarce resource ¡V water ¡V between the agricultural sector and the other economic sectors. Moreover, within the agricultural sector, water is increasingly being used for commercial plantations as opposed to growing food crops, threatening food security. Therefore, it is very important that this scarce resource is managed in an efficient and sustainable manner, for now and future use. This requires understanding the process of evaporation for accurate determination of water-use from agricultural lands. In the past, direct measurements of evaporation have proven difficult because of the cost and complexity of the available equipments, and level of expertise involved. This justifies a quest for relatively simple, accurate and inexpensive methods of determining evaporation for routine field applications. Estimation of sensible heat flux (H) from high frequency air temperature measurements and then calculating latent energy flux (ƒÜE) and hence evaporation as a residual of the shortened surface energy balance equation, assuming that closure is met, is appealing in this sense. Concurrent net irradiance (Rn) and soil heat flux (G) measurements can be conducted with relative ease for use in the energy balance equation. Alternately, evaporation can also be mathematically modelled, using single- or multi-layer models depending on vegetation cover, from less expensive routine meteorological observations. Therefore, the ultimate objective of this study is to estimate and model H and ƒÜE, and thereby evaporation, accurately over sparsely vegetated agricultural lands at low cost and effort. Temperature-variance (TV) and surface renewal (SR) methods, which use high-frequency (typically 2 to 10 Hz) air temperature measurements, are employed for estimation of H. The TV method is based on the Monin and Obukhov Similarity Theory (MOST) and uses statistical measures of the high frequency air temperature to estimate H, including adjustments for stability. The SR method is based on the principle that an air parcel near the surface is renewed by an air parcel from above and, to determine H, it uses higher order air temperature differences between two consecutive sample measurements lagged by a certain time interval. Single- and double-layer models that are based on energy and resistance combination theory were also used to estimate evaporation and H from sparse vegetation. Single- and double-layer models that were extended to include inputs of radiometric temperature in order to estimate H were also used. The transmission of solar irradiance to the soil beneath in sparse canopies is variable and depends on the vegetation density, cover and apparent position of the sun. A three-dimensional radiation interception model was developed to estimate this transmission of solar irradiance and was used as a sub-module in the double-layer models. Estimations of H from the TV (HTV), SR (HSR) and double-layer models were compared against H obtained from eddy covariance (HEC), and the modelled ƒÜE (single- and double-layer) were compared with that obtained from the shortened energy balance involving HEC. Besides, long-term ƒÜE calculated from the shortened energy balance using HTV and HSR were compared with those calculated using HEC. Unshielded and naturally-ventilated fine-wire chromel-constantan thermocouples (TCs), 75 ƒÝm in diameter, at different heights above the ground over sparse Jatropha curcas trees, mixed grassland community and bare fallow land were used to measure air temperature. A three-dimensional sonic anemometer mounted at a certain height above the ground surface was also used to measure virtual temperature and wind speed at all three sites. All measurements were done differentially at 10-Hz frequency. Additional measurements of Rn, G and soil water content (upper 60 mm) were also made. The Jatropha trees were planted in a 3-m plant and inter-row spacing in a 50 m ¡Ñ 60 m plot with the surrounding plots planted to a mixture of Jatropha trees and Kikuyu grass. Average tree height and leaf area index measurements were taken on monthly and bimonthly basis respectively. An automatic weather station about 10 m away from the edge of the Jatropha plot was also used to obtain solar irradiance, air temperature and relative humidity, wind speed and direction and precipitation data. Soil water content was measured to a depth of 1000 mm from the surface at 200 mm intervals. Soil and foliage surface temperatures were measured using two nadir-looking infrared thermometers with one mounted directly above bare soil and the other above the trees. The three-dimensional solar irradiance interception model was validated using measurements conducted on different trees and planting patterns. Solar irradiance above and below tree canopies was measured using LI-200 pyranometer and tube solarimeters respectively. Leaf area density (LAD) was estimated from LAI, canopy shape and volume measurements. It was also determined by scanning leaves using either destructive sampling or tracing method. The performance of the TV method over sparse vegetation of J. curcas, mixed grassland community and fallow land was evaluated against HEC. Atmospheric stability conditions were identified using (i) sensor height (z) and Obukhov length (L) obtained from EC and (ii) air temperature difference between two thermocouple measurement heights. The HTV estimations, adjusted and not adjusted for skewness (actual and estimated) of air temperature (sk), for unstable conditions only and for all stability conditions were used. An improved agreement in terms of slope, coefficient of determination (r2) and root mean square error (RMSE), almost over all surfaces, was obtained when the temperature difference rather than the z/L means of identifying stability conditions was used. The agreement between the HTV and HEC was improved for estimations adjusted for actual sk than not adjusted for sk. Improved agreement was also noted when HTV was adjusted using estimated sk compared to not adjusting for sk over J. curcas. The TV method could be used to estimate H for surfaces with varying homogeneity with reasonable accuracy. Long-term water-use of a fetch-limited sparse vegetation of J. curcas was determined as a residual of the shortened surface energy balance involving HTV and HSR and compared with those estimated using HEC. Concurrent measurements of Rn and G were also performed. The long-term water-use of J. curcas trees calculated from the shortened surface energy balance involving HTV and HSR agreed very well when compared with those obtained from HEC. The seasonal HTV and HSR also agreed very well when compared with HEC. Changes in structure of the canopy and environmental conditions appeared to influence partitioning of the available energy into H and ƒÜE. The seasonal total evaporation for the EC, TV and SR methods amounted to 626, 640 and 674 mm respectively with a total rainfall of 690 mm. Footprint analysis also revealed that greater than 80% of the measured flux during the day originates from within the surface of interest. The TV and SR methods, therefore, offer a relatively low-cost means for long-term estimation of H, and ƒÜE, hence the total evaporation, using the shortened surface energy balance along with measurements of Rn and G. Evaporation and biomass production estimations from tree crops requires accurate representation of solar irradiance transmission through the canopy. A relatively simple three-dimensional, hourly time-step tree-canopy radiation interception model was developed and validated using measurements conducted on isolated trees, hedgerows and tree canopies arranged in tramline mode. Measurements were obtained using tube solarimeters placed 0.5 m from each other starting from the base of a tree trunk in four directions, along and perpendicular to the row up to mid-way between trees and rows. Model-simulations of hourly radiant transmittance were in good agreement with measurements with an overall r2 of 0.91; Willmott.s index of agreement of 0.96; and general absolute standard deviation of 17.66%. Agreement between model-estimations and measurements, however, was influenced by distance and direction of the node from the tree trunk, sky conditions, symmetry of the canopy, and uniformity of the stand and leaf distribution of the canopy. The model could be useful in planning and management applications for a wide range of tree crops. Penman-Monteith (PM) equation and the Shuttleworth and Wallace (SW) model, representing single- and dual-source models respectively, were used to determine the total evaporation over a sparse vegetation of J. curcas from routine automatic weather station observations, resistance parameters and vegetation indices. The three-dimensional solar irradiance interception model was used as a sub-module in the SW model. The total evaporation from the sparse vegetation was also determined as a residual of the shortened surface energy balance using measurements of Rn, G and HEC. The PM equation failed to reproduce the .measured. daily total evaporation during periods of low LAI, with improved agreement with increased LAI. The SW model, however, produced total evaporation estimates that agreed very well with the .measured. with a slope of 0.96, r2 of 0.91 and RMSE of 0.45 mm for a LAI ranging from 0 (no leaves) to 1.83 m2 m-2. The SW model also estimated soil evaporation and plant transpiration separately, and about 66 % of the cumulative evaporation was attributed to soil evaporation. These findings suggest that the PM equation should be replaced by the SW model for surfaces that assume a range of LAI values during the growing season. The H was estimated using (i) SW model that was further developed to include surface radiometric temperature measurements; (ii) one-layer model, but linked with a two-layer model for estimation of excess resistance, that uses surface radiometric temperature; and (iii) the SW model (unmodified). The agreement between modelled and measured H, using 10-min data, was in general reasonably good with RMSE (W m-2) of 45.11, 43.77 and 39.86 for the three models respectively. The comparative results that were achieved from (iii) were not translated into the daily data as all models appeared to have a tendency to underestimate H. The resulting RMSEs for the daily H data for the three models were (MJ m-2) 1.16, 1.17 and 1.18 respectively. It appears that similar or better agreement between measured and estimated H can be forged without the need for surface radiometric temperature measurements. The study showed, in general, that high frequency air temperature measurements can be used to estimate H with reasonable accuracy using the simple and relatively low-cost TV and SR methods. Moreover, these methods can be used to calculate ƒÜE, hence ET, as a residual of the shortened surface energy balance equation along with measurements of Rn and G assuming that energy balance closure is met. The simple and low-cost nature of these methods makes replication of measurements easier and their robust nature allows long-term measurements of energy fluxes. The study also showed that H and ƒÜE can be modeled using energy and resistance combination equations with reasonable accuracy. It also reiterated that the SW-type models, which treat the plant canopy and soil components separately, are more appropriate for estimation of H and ƒÜE over sparse vegetation as opposed to the PM-type models.