Estimación de flujos de agua entre suelo, vegetación y atmósfera mediante teledetección

Abstract

In the boundary layer between vegetation and soil with the atmosphere different processes within the hydrologic cycle take place. When rainfall occurs, part of the water reaches the inland surface, part is infiltrated in the soil, and part is intercepted by vegetation. The fraction reaching the soil infiltrates in the unsaturated zone of the soil making it wet, dissolving the elements that will be absorbed by the plants, and changing the physical properties of the soil. Vegetation needs to open the stomata during the absorption of CO2 during photosynthesis. In this process loss of water by vegetation trough the leaves is produced. If this is loss is large, the wilting point of the plant can be reached, and therefore there is a need of water uptake which is done by the roots. Water returns to the atmosphere trough evaporation and plant transpiration. Evapotranspiration (ET), which can be defined as the process that combines the evaporation from the soil and the transpiration from the plants transferring water vapor to the atmosphere, is one the most relevant processes as it has strong influence on local climate, weather and many other biophysical processes. Finally, ET is an energy driven process that relies on land-surface energy fluxes like net radiation (Rn), sensible heat flux (H) and ground heat flux (G). It is possible to collect in situ different variables from vegetation, soil and atmosphere related to water fluxes in the boundary layer. However, these variables are difficult to measure due to its spatial and temporal variability. Remote sensing platforms on the other hand offer an alternative to in situ measurements. Nowadays, remote sensing offers a great variety of sensors with different spatial and temporal resolutions, which makes this tool very attractive to monitor these processes, although there is a need for accounting for the existing differences between remote sensing estimates and in situ measurements. This issue requires rigorous validations of remote sensing estimates as well as in situ sampling protocols. In this thesis, different variables related to water fluxes between vegetation, soil and atmosphere have been analyzed. The study focused on vegetation and soil water content, biophysical parameters of vegetation and in the behavior of Evaporative Fraction (EF), is defined as the ratio between ET and available energy (Rn-G). All data used were collected in a dehesa ecosystem at Las Majadas del Tiétar, in the province of Cáceres in Spain. In the study site there is an eddy covariance flux tower inside the FLUXNET network that has been operation since 2003. The first part of this work was focused on the estimation of biophysical and structural variables of the vegetation, being the study mainly oriented to estimate vegetation water content. A large data set collected over two full phenological years was used to relate different metrics of vegetation water content with spectral measurements at two different scales, using proximal sensing data and using MODIS surface reflectance data (500 m). Three different water content metrics simultaneously obtained from the same sample - Fuel Moisture Content (FMC), Equivalent Water Thickness (EWT) and Canopy Water Content (CWC) - were related to spectral vegetation indices (VI) calculated from MODIS and proximal sensing data. In addition estimates of FMC and CWC by inversion techniques of Radiative Transfer Model (RTM) were tested and compared against the empirical models. Dry Matter content (Dm) and Leaf Area Index (LAI) were also evaluated empirically with VI. As result from this study, several findings were revealed. The first one is related to the field data collection: results showed that temporal changes in FMC, EWT and CWC are more critical than their spatial variation within the MODIS pixel. The field protocols should hence be adapted in order sample more frequently rather than conducting extensive spatial samplings. Field data analysis also highlighted the need of being consistent with the size of the sample collected, and avoiding as much as possible subjective operator interpretations when collecting the data. Another relevant result from this study is that due to the high seasonal Dm variability, a constant annual value would not be recommended to predict EWT from FMC. Furthermore, Vegetation Atmospheric Resistant Index (VARI) provided the worst results in all cases. The empirical estimators differed between sensors, slightly better for MODIS than proximal sensing, probably due to differences in view angles, and as result the proportions of canopy observed by the sensor increases reducing the soil effect. These empirical methods still exceed RTM inversions developed for other sites to predict FMC and CWC. The second part focused on estimation of soil moisture (SM) combining optical and thermal sensors and evaluating the Temperature Vegetation Dryness Index (TVDI) calculated using the triangle method. Some of the factors involved in the triangle parameterization were studied to better understand how they affect the TVDI values and the implications in the final model performance for estimating SM. A modification was introduced in the TVDI calculation and the Normalized Difference Vegetation Index (NDVI) was substituted by the Normalized Difference Infrared Index (NDII). This modification translated in a better performance of empirical methods to estimate SM using this technique. Finally, this study investigates the behavior of EF on the site and the estimate using remote sensing. The reason why EF is investigated is that EF has been extensively used to retrieve daily ET, assuming that EF remains constant during daytime. Using the EF measured by the Eddy Covariance (EC), we have investigate how EF varies temporary and validated the EF calculated from Landsat using a modified triangle approach in which the VI was substituted by MODIS LAI downscaled to Landsat (from 1km to 30m). A novel method for selecting the highest quality days for the analysis based on the statistics of the triangle is also presented. The validation of EF estimates from remote sensing was carried out using either the pixel contribution based on the information of the EC footprint, or the single pixel located in the EC tower, showing both cases very similar results. Furthermore, daily EF and instantaneous EF at the time of the satellite overpass were also compared with the EF calculated from Landsat. Results of this comparison showed larger differences indicating that for this ecosystem the EF self-preservation assumption should be carefully taken into account if daily ET has to be obtained from EF.