Improved field method for determining land surface emissivity Frank-M. Göttsche1, Folke S. Olesen1, Glynn C. Hulley2, 1 Karlsruhe Institute of Technology (KIT), Postfach 3640, 76021 Karlsruhe, Germany 2 Jet Propulsion Laboratory, California Institute of Technology, Pasadena, USA
[email protected],
[email protected],
[email protected] ABSTRACT Especially over arid regions the relatively high uncertainty in land surface emissivity (LSE) limits the accuracy with which land surface temperature (LST) can be retrieved from thermal infrared (TIR) radiance measurements. LSE uncertainty affects LST obtained from satellite measurements and in-situ radiance measurements alike and an accurate validation of LST products requires accurate knowledge of emissivity for the areas observed by the ground radiometers and the satellite sensor. Additionally, direct comparisons between satellite sensors and ground based sensors are complicated by spatial scale mismatch: ground radiometers usually observe some 10 m2, whereas satellite sensors typically observe between 1 km 2 and 100 km2. Therefore, LST validation sites have to be carefully selected and characterized on the scale of the ground radiometer as well as on the scale of the satellite pixel. Gobabeb (Namibia; hyper-arid desert climate) and Dahra (Senegal; hot-arid steppe-prairie climate) are two of KIT's four dedicated, permanent LST validation stations. At both stations insitu emissivities of the dominant surface cover types were determined with an improved ‘emissivity box method’ and from emissivity spectra obtained for soil samples. The measurements performed with the box method were recorded at a sampling rate of 1 Hz, which significantly eases the identification of invalid data and allows the picking of undisturbed temperatures just before and after each change in box configuration. The results obtained with the box method are in good agreement with those from the emissivity spectra and give further confidence in the in-situ LST determined at KIT’s validation sites.
1 INTRODUCTION Accurate validation of LST satellite products, e.g. LST retrieved from MSG/SEVIRI by the Land Surface Analysis – Satellite Application Facility (LSA-SAF), requires knowledge of emissivity at the spatial scale of the in-situ measurements as well as at the – usually considerably larger – satellite spatial scale. At two of KIT's four dedicated LST validation stations, namely Gobabeb (Namibia; hyper-arid desert climate) and Dahra (Senegal; hot-arid steppe-prairie climate), LSE was determined from in-situ measurements performed with the so-called ‘emissivity box method’ (Combs et al., 1965)(Rubio et al., 1997). Outside their respective rainy seasons, KIT’s validation sites Gobabeb and Dahra have relatively large fractions of bare ground. Therefore, the two sites are particularly prone to be misrepresented in satellite-retrieved LSEs (Göttsche and Hulley, 2012) (Jimenez-Munoz et al., 2014) and in-situ validation of LST strongly depends on accurate in-situ LSE. The emissivity box method consists of a sequence of thermal infrared radiance measurements and employs a box with highly reflective inner walls (in the TIR) to control the radiation from the environment (Combs et al., 1965) (Rubio et al., 1997) (Rubio et al., 2003). In the field method employed by KIT all measured
radiances are recorded at a sampling rate of one second, which allows the picking of undisturbed temperatures directly before and after each change in box configuration (Göttsche and Hulley, 2012). The fast sampling also allows the later identification of erroneous measurements, e.g. caused by incorrect handling of the box during the experiment. Here, we describe the determination of in-situ LSEs with the one-lid emissivity box method and compare results obtained at Gobabeb and Dahra with emissivity spectra of soil and grass samples obtained in the laboratory using a Fourier Transform Infra-Red (FTIR) spectrometer. 2 ONE-LID EMISSIVTY BOX METHOD (Rubio et al., 1997) studied the ‘one-lid’ and the ‘twolid emissivity box’ methods in detail and derived correction terms for both. Gobabeb LST validation site has a very high frequency of clear sky conditions and the gravel plains form an open and unobstructed area. Furthermore, clear sky brightness temperature is regularly below -50°C while surface temperature can exceed 40°C: such high temperature differences improve the signal to noise ratio of the ‘one-lid emissivity box’ method considerably. KIT's emissivity box has inner walls of polished aluminium and the same dimensions as in (Rubio et al., 1997) (Rubio et
n.1
Figure 1 Radiometric measurements performed for the ‘one-lid emissivity box method’. Left: radiance emitted by the sample plus reflected down-welling radiance. Centre: radiance over the sample with box on top (temperature of the sample). Right: radiance inside the box when its bottom is also covered by highly reflective aluminium (temperature of inner box walls for correction). al., 2003) and has been used to validate several satellite-retrieved emissivity products (Göttsche and Hulley, 2012) (Jimenez-Munoz et al., 2014). Using the nomenclature of (Rubio et al., 2003), first the LSE uncorrected for the effect of the box is obtained from the measurements in Figure 1 and simultaneously measured down-welling sky-radiance
L↓a:
(1)
Where LBB is the sample radiance measured under clear sky conditions (Figure 1, left) and L2 is the radiance measured through the bottomless box when it is placed on the sample (Figure 1, centre). LSE corrected for the influence of the box given by (2)
with correction
R = 0.265 is a box-specific correction factor, which depends on box geometry and the spectral response of
the inner walls (Rubio et al., 1997). The term Bc is the radiance measured through the box when its bottom is also covered by polished aluminium, i.e. the box is closed. 3 IMPROVEMENTS OF FIELD METHOD Two improvements of the ‘one-lid box’ field technique were made: the first concerns the ‘Narcissus effect’ while the second improves the acquisition of the radiance measurements. Reliable data acquisition is especially important just before and after the configuration of the emissivity box is changed, e.g. when it is put on or taken off the sample under investigation, since it is from these measurements that emissivity is obtained. 3.1 Avoiding the ‘Narcissus effect’ The Heitronics KT15.85 IIP radiometer used for the emissivity box measurements has an absolute accuracy of better than ±0.3K (Theocharous et al., 2010) and a full field of view of 8.5°, which - together with a box height of 80 cm - results in an observed surface area of about 110 cm2. In order to avoid the 'Narcissus effect’, i.e. the radiometer observing its own reflection on the sample or on the aluminium bottom, the opening for the radiometer in the top of the box is located 6 cm off-centre and the inserted radiometer is inclined by 5 degrees w.r.t. nadir. This ensures that the KT15.85 IIP is directed towards the centre of the sample but at the same time does not observe direct (specular) reflections of itself. Radiance emitted by
n.2
Figure 2 Brightness temperatures (BT) measured with the box method. Left axis: BT of sample and closed aluminium box. Right axis: measured ‘KT15 sky BT’ representative of the down-welling hemispherical radiance in the spectral range of the KT15.85 IIP radiometer (9.6-11.5µm). the radiometer takes at least two reflections over the sample before being reflected back onto itself: over the spectral range of the KT15.85 IIP and over natural surfaces (Lambertian reflectors; reflectance < 10%) the Narcissus effect can, therefore, be neglected. 3.2 Fast and continuous measurements It is essential for the success of the emissivity box method that the sample temperature remains (at least approximately) constant between consecutive measurements (left & centre drawings of Figure 1): this requires experimenters to take fast readings while quickly (< 5 seconds) changing the box configuration. Here all brightness temperatures are continuously recorded once per second (Figure 2) and the optimum measurements for obtaining emissivity are selected later on 'off-line'. Besides reducing errors, the fast sampling allows obtaining emissivity when the box is placed on the sample as well as when it is removed from the sample, i.e. it doubles the measurements that can be used to obtain emissivity. Furthermore, erroneous measurements, e.g. due to an incorrect position of the box on the sample, can be identified later on and excluded from the analysis.
Figure 2 shows examples of measurements performed with the different box configurations over sand dunes of the Namib Desert near Gobabeb. Emissivity is obtained from the BTs just before & after the box is put on the sample (at 17:11:25, 17:15:15, and 17:19:20 [hh:mm:ss]) as well as from the BT just before & after the box is taken off the sample (at 17:13:35, 17:17:40, and 17:21:25 [hh:mm:ss]). The key assumption of the box method is that the observed changes in BT are solely due to the different radiative environments, i.e. inside and outside the box, but do not reflect actual changes in thermodynamic temperature: this is the reason why the radiance measurements have to be performed quickly. The ‘BT of aluminium’ is used for the correction term (eq. 2) and ‘KT15 sky BT’ is the measured zenith BT (right axis, here around -80.3 °C) from which down-welling hemispherical radiance can be obtained (Rubio et al., 1997) (Rubio et al., 2003). 4 RESULTS 4.1 Gobabeb, Namibia Gobabeb station is located on vast and flat gravel plains (several 100 km2), which are mainly covered by
n.3
Figure 3 Emissivity spectra of some samples (Gobabeb) and KT15.85 IIP response function. coarse gravel, sand, and desiccated grass (Göttsche et al., 2013). The gravel plains are highly homogeneous in space and time, which makes them ideal for validating a broad range of satellite-derived products (Göttsche and Hulley, 2012) (Göttsche et al., 2013) (Jimenez-Munoz et al., 2014).
ch10.8, the different response functions of the two instruments have to be accounted for (Göttsche and Hulley, 2012). This is illustrated in Figure 3, which shows emissivity spectra obtained with a Nicolet 520 Fourier Transform Infra-red (FTIR) spectrometer for some samples from Gobabeb, Namibia.
Some emissivities obtained over different surfaces with the one-lid box method (see section 3) are presented in Table 1. Due to the high level of uniformity and the small grain size of the investigated gravel and sand, their respective emissivities have small uncertainties. The ‘disturbed gravel‘ (Table 1) was produced by briefly mixing the top layer of gravel with the sand directly underneath, which caused
Shown alongside with the emissivity spectra in Figure 3 is the KT15’s response function (refers to right axis). The different response functions of the KT15.85 IIP (9.6-11.5 µm) and MSG/SEVIRI ch10.8 (9.8-11.8 µm; see Figure 4) combined with the low emissivities of sand and gravel around the ‘Reststrahlen bands’ (8-9.5 µm; SiO2-stretching) result in slightly different channel-effective emissivities for the two sensors; pronounced Reststrahlen bands are indicative of high quartz content.
Table 1 LSE and standard deviations determined with the one-lid box method at Gobabeb with the KT15.85 IIP radiometer. N gives number of measurements.
For the mixture of 75% gravel and 25% dry grass determined for Gobabeb the emissivity for the KT15.85 IIP is slightly lower (0.940 ±0.015) than the respective emissivity for SEVIRI ch10.8 (0.944 ±0.015) (Göttsche and Hulley, 2012). Furthermore, Figure 3 shows that – in contrast to the sand and gravel – the emissivity of dry grass (‘E01grass’) decreases for wavelengths larger than about 9 µm. 4.2 Dahra, Senegal
its LSE to become similar to that of the sand. Furthermore, for meaningful comparisons between emissivities determined for different sensors, e.g. a KT15.85 IIP in-situ radiometer versus MSG/SEVIRI
Dahra station is located in so called ‘tiger bush’ and is covered by strongly seasonal grass (95%) and sparse, evergreen trees (dominantly acacia trees) with a background of reddish sand (Rasmussen et al., 2011). The strong seasonality is caused by a pronounced rainy season (about Jul/Aug to Sep/Oct), during which
n.4
LST retrieval is highly challenging. Furthermore, the strong vegetation cycle, from nearly bare soil to full vegetation cover was shown to cause errors of up to 7% in some LSE products (Xu et al., 2014). Using the emissivity box method as described in sections 2 and 3, on the 15th of January 2013 the following KT15.85 IIP emissivities were determined at Dahra: 0.960 ±0.004 over a ‘fire strip’, i.e. an area cleared of remaining dry vegetation for fire protection, 0.953 ±0.004 over areas of dry soil cleared just before the measurements, and 0.941 ±0.005 for the unaltered validation site (mainly bare soil with some debris of desiccated vegetation lying on the surface). Additional measurements over dry grass yielded a KT15.85 IIP emissivity of 0.971 ±0.011. Since the vegetation at Dahra is highly seasonal, the value of 0.941 given above is only valid for the 15th of January 2013; however, LSE is expected to be similar until the start of the next rainy season. Figure 4 shows emissivity spectra of soil samples from Dahra and a dry grass spectrum from the ASTER spectral library (Baldridge et al., 2009). The spectrum labelled ‘tree mast’ is for a sample taken from the surface near the LST validation station, whereas the spectrum labelled ‘1m depth’ is for a sample (mainly sand) from 1 m depth, which clearly contains a substantial amount of SiO2.
5 SUMMARY AND CONCLUSIONS Gobabeb and Dahra (the latter during the dry season) both have generally favourable conditions for employing the one-lid emissivity box method:
Frequently clear-sky with surface BT - sky BT > 80K give the radiance measurements a considerably higher SNR than usually achieved with the two-lid box method.
The improved field technique gives accurate and stable results with fewer measurements and the fast sampling allows identifying and rejecting erroneous data.
Furthermore, in combination with the emissivity box the KT15.85 IIP has a sufficiently large field of view (12 cm diameter) to integrate over a larger number of small surface components, e.g. small pieces of gravel: for many surface types, gravel and grass, this yields more representative emissivities. At Gobabeb, the determined in-situ LSE showed that several of the current LSE satellite products had errors of 3% or more (Göttsche and Hulley, 2012), whereas at Dahra the strong seasonal cycle (monsoon; from bare soil to full vegetation cover) caused errors of up to 7% in some LSE products. Therefore, the results presented here underline the importance of in-situ LSE for validating LST retrieval algorithms. ACKNOWLEDGMENTS The in-situ validation data for the Gobabeb and Dahra sites were obtained within the context of the Land Surface Analysis – Satellite Application Facility (LSA-SAF), a project funded by the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT). REFERENCES Baldridge, A., Hook, S., Grove, C., and Rivera, G., 2009. The ASTER spectral library version 2.0. Remote Sensing of Environment, 113(4), 711–715.
Figure 4 Emissivity spectra for samples from Dahra, a dry grass spectrum from the ASTER spectral library, and the KT15.85 IIP and SEVIRI ch10.8 response functions.
Combs, A. C., Weickmann, H. K., Mader, C., and Tebo, A., 1965. Application of infrared radiometers to meteorology. Journal of Applied Meteorology, 4(2), 253–262. Göttsche, F.-M. and Hulley, G. C., 2012. Validation of six satellite-retrieved land surface emissivity products over two land cover types in a hyper-arid region. Remote Sensing of Environment, 124, 149–158.
n.5
Göttsche, F.-M., Olesen, F.-S., and Bork-Unkelbach, A., 2013. Validation of land surface temperature derived from MSG/SEVIRI with in situ measurements at Gobabeb, Namibia. International Journal of Remote Sensing, 34(9-10), 3069–3083. Special Issue. Jimenez-Munoz, J. C., Sobrino, J. A., Mattar, C., Hulley, G., and Gottsche, F.-M., 2014. Temperature and emissivity separation from MSG/SEVIRI data. IEEE Transactions on Geoscience and Remote Sensing, 52(9), 5937–5951. Rasmussen, M. O., Göttsche, F.-M., Diop, D., Mbow, C., Olesen, F.-S., Fensholt, R., and Sandholt, I., 2011. Tree survey and allometric models for tiger bush in northern Senegal and comparison with tree parameters derived from high resolution satellite data. International Journal of Applied Earth Observation and Geoinformation, 13(4), 517–527. Rubio, E., Caselles, V., and Badenas, C., 1997. Emissivity measurements of several soils and vegetation types in the 8–14 µm wave band: analysis of two field methods. Remote Sensing of Environment, 59, 490–521. Rubio, E., Caselles, V., Coll, C., Valour, E., and Sospedra, F., 2003. Thermal-infrared emissivities of natural surfaces: Improvements on the experimental set-up and new measurements. International Journal of Remote Sensing, 24(24), 5379–5390. Theocharous, E., Usadi, E., and Fox, N. P., 2010. CEOS comparison of IR brightness temperature measurements in support of satellite validation. Part I: Laboratory and ocean surface temperature comparison of radiation thermometers. NPL REPORT OP3. Technical Report ISSN: 1754-2944, National Physical Laboratory, Teddington, UK. Xu, H., Yu, Y., Tarpley, D., Gottsche, F.-M., and Olesen, F.-S., 2014. Evaluation of GOES-R Land Surface Temperature Algorithm Using SEVIRI Satellite Retrievals With In Situ Measurements. IEEE Transactions on Geoscience and Remote Sensing, 52(7), 3812–3822.
n.6
