Main content area

Accuracy of the Temperature–Vegetation Dryness Index using MODIS under water-limited vs. energy-limited evapotranspiration conditions

Garcia, M., Fernández, N., Villagarcía, L., Domingo, F., Puigdefábregas, J., Sandholt, I.
Remote sensing of environment 2014 v.149 pp. 100-117
air temperature, algorithms, arid lands, climatic factors, drought, ecosystems, eddy covariance, energy balance, evapotranspiration, growing season, heat transfer, models, moderate resolution imaging spectroradiometer, monitoring, normalized difference vegetation index, relative humidity, remote sensing, shrublands, solar radiation, spatial variation, surface temperature, transpiration, Spain
Water deficit indices based on the spatial relationship between surface temperature (Ts) and NDVI, known as triangle approaches, are widely used for drought monitoring. However, their application has been recently questioned when the main factor limiting evapotranspiration is energy. Even though water is the main control in dryland ecosystems, these can also undergo periods of energy and temperature limitation. In this paper we aimed to: (i) evaluate the TVDI (Temperature–Vegetation Dryness Index) to estimate water deficits (e.g. ratio between actual and potential evapotranspiration), and heat surface fluxes using MODIS data; and (ii) provide insights about the factors most affecting the accuracy of results. Factors considered included the type of climatic control on evapotranspiration, λE, (i.e. water-limited vs. energy-limited), the quality of Tair estimates, the heterogeneity of land cover types and climatic variables in the region, or the algorithm to extract hydrological boundaries from the images.The TVDI was compared with eddy covariance (EC) data from two shrublands with different climatic controls for λE in South Spain. Evaluations showed that it could be used to estimate the water deficit when water was the main limiting factor (R=0.81–0.88; Mean Average Error, MAE=0.16–0.17) but not in energy-limited situations (R<0.2; MAE=0.10–0.2). Spatial heterogeneity in climatic variables also had a different impact on accuracy depending on limiting factors. Relative humidity was significant at the water-limited site while solar irradiance and air temperature were more important at the energy-limited site. The skill of the TVDI to estimate surface fluxes at the water-limited site was confirmed for the dominant sensible heat flux, H (R2=0.93; Mean Absolute Percentage Error, MAPE=12.85%) but not for λE (R2=0.01, MAPE=115.22%) as λE fluxes at this site are just slightly above the error of the eddy covariance system. At the energy-limited site, λE (R2=0.74; MAPE=31.83%) and H estimates (R2=0.80; MAPE=26.85%) were better than those from the SEBAL (Surface Energy Balance Algorithm for Land) or the PML (Penman–Monteith–Leuning) models. However, the skill to predict surface fluxes in this case was due to the net radiation inputs and not by the TVDI input.Our analyses also suggest: (1) to apply the TVDI excluding energy-limited sites/periods based on climatic knowledge of limiting factors on λE; (2) the best conditions for TVDI performance correspond to the situation when the controlling factors are less limiting, e.g. during the growing season (higher SWC and lower VPD); and (3) to account better for the role of vegetation controls on transpiration.