Surface fluxes over natural landscapes using scintillometry

Motivated by the demand for reliable area-averaged fluxes associated with natural landscapes this thesis investigates a relative new measurement technique known as the scintillation method. For homogeneous areas the surface fluxes can be derived with reasonable accuracy. However, fluxes representative for large natural landscapes (comparable to the horizontal grid box size of numerical models or the pixel size of satellite imagers) are more difficult to obtain because at these scales the surface is mostly heterogeneous. At this moment only a few techniques are available that can provide flux information at spatial scales of several kilometres, such as the scintillation method. Based on the propagation statistics of EM radiation that has propagated through the atmosphere over a horizontal path of several kilometres it is possible to derive the surface fluxes of sensible heat, water vapour and momentum. In this study a Large Aperture Scintillometer (LAS) has been developed that can be used over distances up to 5 km. Since the LAS operates at a near-infrared wavelength hence it is primarily sensitive to temperature related scintillations, from which the sensible heat flux can inferred. In this thesis the following aspects regarding the LAS are investigated: The performance of the LAS over heterogeneous land surfaces The reliability of area-averaged water vapour fluxes provided by the LAS and in combination with a radio wave scintillometer over heterogeneous land surfaces Its practical applicability and usefulness in other scientific areas For the derivation of the sensible heat flux from the LAS signal one must rely on the Monin-Obukhov Similarity Theory (MOST). However, MOST requires homogeneous surface conditions. The question arises whether the LAS can be used over distances of several kilometres, since at these scales the surface is mostly heterogeneous. In order to test experimentally the applicability of the LAS over heterogeneous areas and the reliability of the derived fluxes of sensible heat a field campaign was carried out in Flevoland (The Netherlands). The general characteristics of the Flevoland area are as follows: a vast and completely flat area covered by four crops in a chessboard configuration of patches of 500 m ´ 250 m. Based on the horizontal length scale of the patches this landscape is classified as a Type A landscape, meaning that only the lower part of the surface layer is affected by the irregularities. Eddy covariance (EC) measurements were performed over the main types of farmland to provide to aggregation independent area-averaged fluxes. The EC observations reveal that the heterogeneity in the Flevoland area is primarily the result of spatial variations in the thermal properties. The fluxes of two large aperture scintillometers, installed at a height of 11.6 m and 20.4 m, respectively, show a close resemblance with the area-averaged EC fluxes, especially for the upper LAS. The lower LAS shows a slight underestimation of the sensible heat flux of approximately 7%. This underestimation is assessed using a blending height approach and an analytical footprint model for estimating the source areas and the associated fluxes. The blending height is considered as the level above the surface where the influences of the patches gradually decay. It is found, using a heuristic model that the blending height for the Flevoland area varies between 9 m and 14 m. Based on the found blending heights it is concluded that the upper LAS always measured above the blending height, which is consistent with the depicted LAS results. For the lower LAS the situation is more complicated as the individual fields influence the measurements, suggesting that the MOST may be violated. After dividing the Flevoland area into 8 wind-sectors and re-arranging the area-averaged fluxes for the entire area and for the 8 source areas, a closer agreement is found. These results indicate that from a LAS, which measures just below the blending height, still reliable area-averaged fluxes can be derived and that the violation of MOST is small. Next the performance of a combined LAS and radio wave scintillometer (LAS-RWS) over a heterogeneous land surface is studied. Although this scintillation technique, known as the two-wavelength method, provides both the sensible heat flux and the water vapour flux, most attention is focussed at the water vapour flux. The water vapour flux provided by a `stand-alone` LAS is evaluated also. In the latter case the water vapour flux is estimated as the residual of the surface energy balance equation using a simple parameterisation scheme (based on global radiation data) for estimating the area-averaged available energy (i.e. Rn - Gs). The LAS-RWS study is based also on data of the Flevoland experiment. As mentioned before the EC observations collected in Flevoland reveal that the heterogeneity in the area is primarily the result of spatial variations in the thermal properties and likewise in the buoyant production term of MOST. First, the water vapour fluxes from the combined LAS-RWS system are investigated. It was found that these fluxes agree well with the area-averaged water vapour flux aggregated from the in-situ observations. The found scatter is explained to be caused by: closure failure of the energy balance for the EC measurements, the non-linearity between the structure parameters and the inferred fluxes, and low frequency water vapour absorption fluctuations that affect the RWS. Finally, the water vapour fluxes derived from the stand-alone LAS are discussed. These results show that the `stand-alone¿ LAS can provide also reliable estimates of the area-averaged water vapour flux over heterogeneous areas (type A). In order to study the operational aspects of the LAS two LAS devices and a small micrometeorological station were installed in the Gediz Basin near Menemen (Turkey) in 1998 as part of an international experiment. The main objective of this experiment was to compare actual evapotranspiration estimates based on satellite remote sensing methods, hydrological models and field methods. This thesis deals only with the field methods, i.e. the variance method and in particular the LAS. One LAS was set-up over a transect of the valley of the Gediz river basin. For the derivation of the sensible heat flux additional wind speed and temperature data are taken from a nearby meteorological station. In addition a small micro-meteorological station was placed at an irrigated cotton field. The fluxes for this site are inferred from collected temperature fluctuation data using the variance method. Due to experimental problems with a second LAS installed at the same site these data are excluded from this study. The presented time series of 24-hour average fluxes for the valley clearly shows the seasonal trend of the sensible heat flux, including the irrigation events. This time series demonstrates from an operational perspective that the LAS, which was operational during the entire growing season, is a robust and reliable instrument that requires only occasional servicing. Finally, it is investigated whether the LAS fluxes collected in Turkey can be used as `ground-truth` data in other scientific studies such as remote sensing. For that purpose a large number of surface flux maps are generated using the SEBAL remote sensing algorithm, and are compared with the LAS results. In this validation study the in-situ fluxes and radiation measurements of the irrigated cotton field are included also. The SEBAL fluxes are derived from moderate resolution AVHRR visible and thermal-infrared images taken from the NOAA-14 satellite. Both instantaneous and daily average sensible heat fluxes are determined for the entire growing season. It is found that the SEBAL based instantaneous fluxes agree closely with the in-situ fluxes for the cotton site. However, the results for the valley site, i.e. SEBAL versus LAS, reveal a discrepancy. The following reasons are offered: the scaling mismatch between the source area of the LAS and the pixel size of the raw AVHRR images; a possible distortion by the dry slopes in the relative narrow valley compared to the resolution of the AVHRR imager. Therefore a convincing validation of SEBAL for the valley cannot be done. Finally, the daily average sensible heat fluxes for the cotton field are compared. It is found that only during the irrigation period the daily average results agree. Not actually dry reference pixels, which lead to underestimated evaporative fractions, are suggested to be the reason for the observed difference