**Inter-Comparison of Field- and Laboratory-Derived Surface Emissivities of Natural and Manmade Materials in Support of Land Surface Temperature (LST) Remote Sensing**

**Mary F. Langsdale 1,\* , Thomas P. F. Dowling 1,2 , Martin Wooster <sup>1</sup> , James Johnson <sup>1</sup> , Mark J. Grosvenor <sup>1</sup> , Mark C. de Jong <sup>1</sup> , William R. Johnson <sup>3</sup> , Simon J. Hook <sup>3</sup> and Gerardo Rivera <sup>3</sup>**


Received: 31 October 2020; Accepted: 11 December 2020; Published: 17 December 2020

**Abstract:** Correct specification of a target's longwave infrared (LWIR) surface emissivity has been identified as one of the greatest sources of uncertainty in the remote sensing of land surface temperature (LST). Field and laboratory emissivity measurements are essential for improving and validating LST retrievals, but there are differing approaches to making such measurements and the conditions that they are made under can affect their performance. To better understand these impacts we made measurements of fourteen manmade and natural samples under different environmental conditions, both in situ and in the laboratory. We used Fourier transform infrared (FTIR) spectrometers to deliver spectral emissivities and an emissivity box to deliver broadband emissivities. Field- and laboratory-measured spectral emissivities were generally within 1–2% in the key 8–12 micron region of the LWIR atmospheric window for most samples, though greater variability was observed for vegetation and inhomogeneous samples. Differences between laboratory and field spectral measurements highlighted the importance of field methods for these samples, with the laboratory setup unable to capture sample structure or inhomogeneity. The emissivity box delivered broadband emissivities with a consistent negative bias compared to the FTIR-based approaches, with differences of up to 5%. The emissivities retrieved using the different approaches result in LST retrieval differences of between 1 and 4 ◦C, stressing the importance of correct emissivity specification.

**Keywords:** land surface temperature; land surface emissivity; measurement uncertainties; emissivity box method; Fourier transform infrared spectrometer; portable spectrometer

#### **1. Introduction**

Emissivity is a spectrally varying property of a material, describing at any particular wavelength the efficiency at which an object emits electromagnetic radiation as a function of its temperature. It is mathematically defined as the ratio between the electromagnetic radiation actually emitted by the object at the wavelength in question, and that emitted by a black body at the object's thermodynamic (or kinetic) temperature [1]. Kirchhoff's law of thermal radiation furthermore states that at any particular wavelength, the absorptivity of a surface is equal to the emissivity of the surface if it is in thermal equilibrium with its surroundings, meaning for example that a perfect blackbody absorbs all the arriving electromagnetic radiation and re-emits the absorbed energy according to Planck's radiation law [2]. However, natural materials are not perfect blackbodies, and most are selective radiators, which may emit electromagnetic radiation according to Planck's radiation law at certain wavelengths, but not others. It is therefore important to understand their spectrally varying emissivity across the electromagnetic spectrum, including within the longwave infrared (LWIR) spectral atmospheric window (8–13 µm) where most remote sensing of land surface temperature (LST) is conducted. This is particularly the case when estimating LST remotely, where knowledge of the target's surface emissivity in the LWIR is essential when converting infrared brightness temperature (BT) measurements into accurate estimates of LST [3].

Emissivity depends on the chemical makeup of a material, and its geometry, surface roughness, and moisture content and as such can show strong seasonally varying cycles and land use/land cover variability [4]. Most soils and vegetation emissivities vary between a minimum of around 0.6 to a maximum of at or close to 1, while pure metals for example can have far lower values [2]. Unfortunately, relatively small errors in the assumed emissivity of a surface can induce quite large impacts on the finally estimated LST. For typical earth surface conditions, Jiménez-Muñoz and Sobrino [5] calculated that emissivity uncertainties of 0.01 typically result in LST uncertainties of around 0.6 K. Given the many applications of LST–such as deriving evapotranspiration and monitoring droughts [6]–recent years have seen an increase in interest in improving the accuracy of LST retrieval as evidenced by the development of new thermal infrared (TIR) sensors capable of LST retrieval such as the ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) [7] and the classification of LST as an essential climate variable (ECV) by the World Meteorological Organisation's Global Climate Observing System (GCOS) [8]. The correct specification of surface spectral emissivity has been identified as the greatest source of error in current satellite-based measurements of LST [9] and it therefore is essential to try to minimise emissivity uncertainties in order to maximise the accuracy of remotely sensed LST estimates.

Multiple field and laboratory techniques for measuring emissivity have been developed, enabling both spectral emissivity measurement and broadband emissivity retrieval [10–14]. Although unable to perfectly capture field conditions known to impact surface spectral emissivity, such as soil moisture [15] or canopy structure [16], laboratory-based emissivity measurements are often preferred to field-based measurements (for samples that can be transported without modifying the sample and its emissivity). This is because, unlike field measurements, laboratory measurements can be collected under highly controlled conditions, thus reducing errors that might result from changing atmospheric or thermal conditions in the field for example [17]. Online spectral emissivity libraries consist predominately of such laboratory-derived emissivity spectra [18–22]. Data from these spectral libraries have been used extensively to "ground-truth" airborne and satellite LST and emissivity outputs [14,23–25], for the derivation LST algorithm coefficients [26–28] and in the calibration of LWIR satellite and airborne sensors [29]. However, a recent inter-comparison of laboratory emissivity measurements of the same samples reported some quite significant differences in emissivity values from different laboratory measurement setups [30]. For example, they found standard deviations of ±2.52% (0.024) in the emissivities derived for distilled water within the LWIR atmospheric window (8–14 µm). These uncertainties are much larger than those previously reported for laboratory setups [15] and larger than those typically reported with field measurements [11,31,32], thus highlighting the continuing importance of field-based surface emissivity measurements. This is particularly true given that such in situ measurement approaches allow measurements of emissivity under "natural conditions"—for example for samples such as vegetation that is difficult to preserve while transported.

Given that the correct specification of surface spectral emissivity is the greatest source of error in current satellite-based measurements of LST [9], and the discrepancies that have been found both within laboratory and between field and laboratory measurements detailed above, there is a need for further rigorous examination of the degree of agreement between current approaches to emissivity measurement. With this in mind, we conducted a study to compare different field and laboratory spectral emissivity measurement approaches, using the same targets to better understand the emissivity differences that can result from use of different measurement approaches and/or different measurement conditions. We have focused on Fourier transform infrared (FTIR) spectrometer-based emissivity measurement systems since these are the most common type used to provide spectral emissivity measurements, applying to the measured spectra a variety of different post-processing approaches to derive the surface emissivity information. We also include a comparison of these spectrally resolved data to the broadband emissivities produced using an "emissivity box", a popular low-cost method of broadband field emissivity determination that uses a sequence of LWIR radiometer measurements and a specially constructed box [33]. The impact of the emissivity measurement uncertainties from these methods on calculation of in situ LSTs is assessed as the last stage of our investigation.
