Reassessment of the temperature-emissivity separation from multispectral thermal infrared data: Introducing the impact of vegetation canopy by simulating the cavity effect with the SAIL-Thermique model

Authors: JACOB, FREDERIC - Institute For Research And Development (IRD); AUDREY, LESAIGNOUX - Institute For Research And Development (IRD); OLIOSO, ALBERT - Institut National De La Recherche Agronomique (INRA); WEISS, MARIE - Institut National De La Recherche Agronomique (INRA); CAILLAULT, KARINE - Onera/dota; JACQUEMOUD, STEPHANE - Universite Paris Descartes; NERRY, FRANCOISE - National Council For Scientific Research-Cnrs; French, Andrew; SCHMUGGE, THOMAS - New Mexico State University; BRIOTTET, XAVIER - Onera/dota; LAGOUARDE, JEAN-PIERRE - Institut National De La Recherche Agronomique (INRA)

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/24/2017
Publication Date: 9/1/2017
Citation: Jacob, F., Audrey, L., Olioso, A., Weiss, M., Caillault, K., Jacquemoud, S., Nerry, F., French, A.N., Schmugge, T., Briottet, X., Lagouarde, J. 2017. Reassessment of the temperature-emissivity separation from multispectral thermal infrared data: Introducing the impact of vegetation canopy by simulating the cavity effect with the SAIL-Thermique model. Remote Sensing of Environment. 198:160-172.

Interpretive Summary: Land surface temperature is an important environmental variable that is an indicator of plant water use, plant heat and drought stress, and soil moisture content. Land surface temperatures obtained with remote sensing aircraft and satellites are increasingly being used to estimate agricultural uses of water. However, accuracy of these estimates is constrained by the difficulty of discriminating between true and apparent land surface temperatures. This difficulty is caused in part by the effect of the hard-to-determine emissivity, a surface property that controls thermal radiative efficiency. When the surface emissivity is less than 100%- always the case- an apparent remote sensing temperature will be less than the true temperature, sometimes by several degrees Celsius. Determining accurate emissivities is therefore essential for computing meaningful water use and soil moisture estimates. Previous research has shown that surface emissivities can be estimated by correlating laboratory-scale leaf and soil samples with their spectral range of emissivities, and then applying the resulting correlations to meter to kilometer scale observations. The method is known as Temperature-Emissivity separation (TES). Unfortunately, these correlations commonly result in under-estimated emissivities, a consequence of which is an over-estimation of land surface temperatures. To address this problem a new approach was taken to model the land surface materials in a more realistic way by considering the scattering properties of the vegetation canopy. The model, ‘SAIL-Thermique’, was used to simulate a variety of soil and plant conditions for existing and prospective sensors. Simulation results led to an accurate and robust TES formulation that responds to growing vegetation. Results from this research will be important for scientists and engineers using multispectral thermal remote sensing data to retrieve land surface temperatures from farm to continental scales.

Technical Abstract: We investigated the use of multispectral thermal imagery to retrieve land surface emissivity and temperature. Conversely to concurrent methods, the temperature emissivity separation (TES) method simply requires single overpass without any ancillary information. This is possible since TES makes use of an empirical relationship that estimates the minimum emissivity e-min from the emissivity spectral contrast captured over several channels, so-called maximum-minimumdifference (MMD). In previous studies, the e-min -MMD empirical relationship of TES was calibrated and validated for various sensor spectral configurations, where the proposed calibrations involved single or linearly mixed spectra of emissivity at the leaf or soil level. However, cavity effect should be taken into account at the vegetation canopy level, to avoid an underestimation of emissivity, especially for intermediate vegetation conditions between bare soil and full vegetation cover. The current study aimed to evaluate the performances of the TES method when applied to vegetation canopies with cavity effect. We used the SAIL-Thermique model to simulate a library of emissivity spectra for a wide range of soil and plant conditions, and we addressed the spectral configurations of recent and forthcoming sensors. We obtained good results for calibration and validation over the simulated library, except for full cover canopies because of the TES gray body problem. Consistent with previous studies, the calibration/validation results were better with more channels that capture emissivity spectral contrast more efficiently. Our TES calibrations provided larger e-min values as compared to former studies, especially for intermediate vegetation cover.We explained this trend by the simulated spectral library that involved numerous vegetation canopies with cavity effect, thereby shifting up the e-min - MMD empirical relationship. Consequently, our TES calibration provided larger (respectively lower) estimates of emissivity (respectively radiometric temperature) that were likely to be more realistic as compared to previous calibrations. Finally, SAIL-Thermique simulations permitted to show that increasing Leaf Area Index induced a displacement of the (e-min, MMD) pairs along the empirical relationship. This was consistent with the TES underlying assumption, where any change in e-min induces changes in MMD since e-max is bounded on [0.98–1]. Further investigations should focus on validating the outcomes of the current study against ground-based measurements, and on assessing TES performances when accounting for instrumental and atmospheric perturbations.