VALIDATION OF THE AATSR LST PRODUCT AT THE VALENCIA TEST SITE: CAMPAIGNS

Transcription

1 VALIDATION OF THE AATSR LST PRODUCT AT THE VALENCIA TEST SITE: CAMPAIGNS César Coll, Vicente Caselles, Enric Valor, Raquel Nicl s, Juan M. S nchez and Joan M. Galve Department of Thermodynamics, Faculty of Physics, University of Valencia. 50, Dr. Moliner Burjassot, SPAIN ABSTRACT/RESUME The AATSR Land Surface Temperature (LST) product was validated with ground measurements in a test site close to Valencia, Spain. The test site is located in a large, flat and thermally homogeneous area of rice crops. Experimental campaigns were carried out during the summers of , when the area showed full vegetation cover. Ground LSTs were measured radiometrically along transects covering an area of 1 km 2, concurrently to morning, cloud-free overpasses of the Envisat satellite. A total number of 23 concurrences of ground measurements and AATSR data were obtained. The AATSR LST product is based on the split-window technique ( and μm channels at nadir), with the algorithm coefficients depending on the land cover type (i) and the vegetation cover fraction (f). Currently, values of i and f are assigned to each pixel according to global maps at spatial resolution of 0.5¼ 0.5¼, which appears too coarse to account for the large heterogeneity of land surfaces. For the Valencia test site, the LST product overestimated the ground LSTs by 3.5 ¼C in average, with a standard deviation of 0.7 ¼C. However, it was shown that the AATSR split-window algorithm worked satisfactorily when the characteristics of the area were adequately prescribed through i and f. In this case, the AATSR derived LSTs agreed with the ground LSTs within ±1.0 ¼C for all 23 concurrences, with nearly zero average error and standard deviation of 0.5 ¼C. Similar results were obtained for other emissivity dependent split-window algorithm also checked with the ground database. Therefore, it is recommended to improve the resolution of i and f to 1 km in the AATSR LST product. 1 INTRODUCTION Land surface temperatures (LSTs) are required for the estimation of the energy and water balance between the atmosphere and the land surface, being of prime interest for meteorological and climatological studies. Thermal infrared (TIR) remote sensing is the unique way for the derivation of LST over large portions of the Earth at diverse periodicities. The Advanced Along Track Scanning Radiometer (AATSR), although primarily designed to obtain accurately the sea surface temperature (SST), can also be used for the retrieval of LST. An operational LST product is provided with the AATSR L2 data, which is based on the split-window technique using the and μm channels, nadir view [1]. Due to the heterogeneity of land surfaces, it is difficult to make ground (point) LST measurements comparable to satellite estimates (which represent area-averaged LSTs). In order to validate LSTs, test sites must be much larger than the field of view (FOV) of the satellite sensor and homogeneous in emissivity and temperature, since the major sources of error in the LST validation are the spatial variations of surface temperature and emissivity [2]. The objective of the present work is the evaluation of AATSR derived LSTs with a database of groundmeasured LSTs collected at the Valencia test site, which is part of a network of sites dedicated to the AATSR LST validation [3]. Ground measurements were performed concurrently to morning, cloud-free AATSR overpasses during the summers of A total number of 23 ground-aatsr concurrences or matchups were obtained. A preliminary analysis for five matchups of 2002 can be found in [4] and [5]. In the present paper, the study has been extended to the rest of the data. Results are presented for the validation of the AATSR LST product [1] and for a quadratic split-window algorithm [6] adapted to AATSR. 2. TEST SITE AND EXPERIMENTAL DATA The test site is located in a large marshy plain dedicated to the intensive cultivation of rice in the Mediterranean coast of Spain, close to the city of Valencia. From the end of June to the beginning of September, rice crops are well developed and attain nearly full cover. In these circumstances, the site shows a high thermal homogeneity and is large enough for making ground measurements of LST comparable to satellite estimates Thermal homogeneity of the site Fig.1 presents examples of AATSR images showing the rice field area and the location of the test site, for July, In Fig.1a, a colour composite of the visible channels is shown, where the rice field area appears in red. Fig.1b shows an image of brightness temperatures at μm, nadir view, centred at the rice field area. The thermal homogeneity of the test site can be assessed with the data of Fig.1b. In this image, a polygon enclosing 39 AATSR pixels around the ground measurement sites is shown, for which we obtain a maximum temperature =27.71 ºC, a minimum TM temperature =26.84 ºC, an average temperature Tm =27.19 ºC, and a standard deviation σ=0.20 ºC. Tav Proceedings of the Second Working Meeting on MERIS and AATSR Calibration and Geophysical Validation (MAVT-2006), March 2006, ESRIN, Frascati, Italy (ESA SP-615, July 2006)

2 In order to show the thermal homogeneity of the test site at a finer spatial resolution, we used a Terra/Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) scene acquired over the area in August 3, Fig.2a shows a km 2 image of brightness temperature in band 13 (10.66 μm) at a spatial resolution of 90 m. We selected the pixels in the rice fields enclosed by the solid-line polygon of Fig.2a (approximately the same area as the polygon in Fig.1b), but excluding the hot pixels within the dashedline square, which is the largest temperature heterogeneity in the rice crop area. For these area (4400 pixels or about 36 km 2 ), T =30.15 ºC, T =25.79 ºC, M m T =26.89 ºC, and σ=0.45 ºC. Fig.2b shows a av histogram of the brightness temperatures for the selected pixels. The largest temperatures ( ºC) were for a few pixels in the southern part of the area (roads). However, about 97 % of the pixels had temperatures between 26.0 ºC and 28.0 ºC. According to the temperatures and size of the aforementioned hot spot, it could increase the 1 km 2 surface temperature by as much as 1 ¼C with regard to the surrounding rice field temperatures. This effect was sometimes noticeable in the AATSR images, so the hot pixel could be discarded. These results show that the experimental area contains a considerable number of 1 km 2 pixels for which the variability in surface temperature can be regarded as ¼C, approximately. Therefore it could be used for the validation of AATSR, whose FOV is 1 km 1 km for the nadir view, and 1.5 km 2 km for the forward view [7] Ground LST measurements Ground LSTs were measured in the test site by means of up to four TIR radiometers distributed over a square of 1

3 km 2 within the rice field area. The instruments were two CIMEL CE 3 radiometers with four bands (1 to 4 at 8-13 μm,.5-.5 μm, μm and μm, respectively), one Everest model 2.2L thermometer with one single band (8-13 μm) and one AGA model 80 thermometer (single band, 8-13 μm). Bands 2 and 3 of CE 3 are similar to channels at and μm of AATSR, respectively. Each radiometer was assigned to one part of the 1 km 2 square in order to cover the area as much as possible. In the 2002 and 2003 campaigns, the 1 km 2 test site was centred at 0¼17 50 W, 39¼14 27 N. For the 2004 campaign, the test site was moved 1 km North (centre at 0¼17 43 W, 39¼15 01 N), while in 2005 it was moved again to the North (center at 0¼18 28 W, 39¼15 54 ) (see Fig. 2a). Radiometers were carried forth and back along transects of m in length, looking at the surface at angles close to nadir. The field of view of the radiometers was 30 cm on the crop surface. Measurements were made at a rate of more than 5 measurements per minute, covering a distance of m per minute. The methodology for the measurement of ground LSTs for comparison with AATSR derived LSTs is described in [5]. Some details are given below: a) Calibration of the field radiometers. The instruments were calibrated with a blackbody source and intercompared in the field along the campaigns. The absolute accuracy, σ(cal), of the CE 3 (CE1 and CE2), was ±0.2 ¼C and ±0.1 ¼C, respectively. For the Everest σ(cal) was between ±0.5 and ±0.7 ¼C, and for the AGA between ±0.7 and ±0.9 ¼C. b) Emissivity correction. Radiometric temperatures were corrected for emissivity effects, including the reflection of the sky radiance. Surface emissivity was measured in the field using the box method [8] for the four channels of the CE 3 radiometers. Measurements showed a high emissivity (ε=0.985) with negligible spectral variation, typical for green vegetation ([8] and [9]). The sky radiance L sky was measured along the transects. For an uncertainty of ±0.010 in ε, the error in temperature due to the emissivity correction, σ(em), was about ±0.3 ¼C. c) Averaging of transect/radiometer ground temperatures. Only the temperatures measured within 3 minutes around the satellite overpass were considered. They were averaged for each transect/radiometer and the standard deviation was calculated. It gives us an estimation of the LST spatial and temporal variability in a part of the test site, σ(var). For the data analyzed here, σ(var) was between ±0.3 ¼C and ±0.5 ¼C. The total uncertainty in LST for each radiometer, σ(t), is given by the combination of the three sources of error (calibration, emissivity correction and variability) according to σ(t) = [σ(cal) 2 +σ(em) 2 +σ(var) 2 ] 1/2 (1) For each day of measurement, the ground LST and uncertainty for each radiometer/transect are given in Table 1 together with the date and time of the AATSR overpass. The most accurate LSTs correspond to the CE1 and CE2 radiometers, for which the largest source of error was σ(var). When the two CE 3 instruments were available, the maximum difference between their LSTs was 0.6 ¼C. In the case of the Everest and AGA instruments, σ(cal) was usually the largest source of error. In order to avoid excessive uncertainty due to the calibration problems of Everest and AGA, we kept only their LST data with σ(t) 1.0 ¼C. In addition, we removed all LSTs measured by these instruments that differed by more than 1.0 ¼C from any of the CE 3 on the same day. d) Average ground LST. The ground LSTs to be compared with AATSR derived LSTs at 1 km 2 resolution were calculated by averaging all the individual ground temperatures within the 3 minute periods for the radiometers available each measurement day. The average LSTs and uncertainties are given in the last column of Table AATSR data For each day of measurements of Table 1, concurrent AATSR scenes were available through the AATSR validation team (University of Leicester) and ESA/ESRIN. Only the nadir view was considered. For each scene, we identified the pixel closest to the centre of the test site in the L1b scenes (geo-referenced, top of the atmosphere data), and we took the 3 3 pixels centred on it. This process was done manually and care was taken in not including any pixel that could be partially out of the rice field area, or the hot pixel mentioned in section 2.1. For the selected pixels, the average brightness temperature and the standard deviation was calculated for the and μm channels (T and T as shown in Table 2. The satellite viewing angle for the centre pixel was also obtained. ),

4 Table 1. Dates and times of AATSR overpasses with ground LSTs. Columns 4-7 give the LST and uncertainty for each radiometer/transect. The last column gives the average LST and uncertainty to be compared with AATSR data. Year Date (day/month) Overpass time(utc) LST ± σ(t) (¼C) CE1 CE2 Everest AGA Average 10/07 10: ± ± ±0.6 13/07 10: ± ± ± ±0.9 26/07 10: ± ± ± ±06 08/08 10: ± ± ±0.7 14/08 10: ± ± ±0.5 17/08 10: ± ± ±0.6 05/09 10: ± ±0.8 08/07 10: ± ± ± ± ±0.7 /07 10: ± ± ± ±0.7 14/07 10: ± ± ± ±0.6 24/07 10: ± ± ± ±0.6 30/07 10: ± ± ± ±0.6 /08 10: ± ± ± ±0.6 28/06 10: ± ± ±0.6 08/07 10: ± ± ±0.6 14/07 10: ± ± ± ±0.7 27/07 10: ± ± ±0.4 30/07 10: ± ± ±0.4 /08 10: ± ± ± ±0.6 //07 10: ± ± ±0.6 21/07 10: ± ± ±0.6 28/07 10: ± ± ± ±0.5 06/08 10: ± ± ± VALIDATION OF SPLIT-WINDOW ALGORITHMS The AATSR LST algorithm [1] expresses the LST as a linear combination of the brightness temperatures T with coefficients determined by regression over and T simulated data-sets and depending on the land cover type (i), the fractional vegetation cover (f), the precipitable water (pw) and the satellite zenith viewing angle (θ): with LST = a + b (T -T ) n + (b + c )T f,i,pw f,i f,i f,i a f,i,pw n = cos(θ/5) 1 = 0.4[sec(θ)-1]pw + f a v,i b = f b f,i v,i c f,i = f c v,i + (1-f) b s,i + (1-f) c s,i + (1-f) a s,i (2) These coefficients are provided for 14 different biomes or land cover classes (i=1 to 14) in [10]. For a given land cover class, two separate sets of coefficients are given for the fully vegetated surface (subscript v) and for the bare surface (subscript s), which are weighted by the fractional vegetation cover f. LST data generated with this algorithm are currently provided with AATSR_L2 data. The algorithm is operationally implemented at the Rutherford Appleton Laboratory (RAL) in the so-called RAL processor. The values of i, f and pw are taken from global classification, fractional vegetation cover maps and global climatology at a spatial resolution of 0.5º 0.5º longitude/latitude. Monthly variability is allowed for f and pw. The LST values given by the AATSR LST operational algorithm for the test site are shown in the fourth column of Table 3. They have been extracted from the L2 data for 3 3 pixels centred on the pixel closest to the test site. The seventh column shows the difference between the ground and the AATSR derived LSTs. According to these results, the RAL processor seems to overestimate the ground LSTs by 3.5 ¼C in average. It appears that the 0.5¼ resolution used by the operational algorithm is too coarse in order to properly assign the land cover type (i) and the fractional cover (f) to specific, relatively small areas such as our test site. For the site, RAL processor assigns class i=6 (broadleaf trees with groundcover) and f= (July August). Particularly the value assigned to f seems very low for the rice crops in summer. Taking the same class (i=6) and f=1 (full vegetation cover), the AATSR LST algorithm was applied to the brightness temperatures of

5 the test site. The resulting LSTs (not shown in Table 3) agreed much better with the ground LSTs: the average overestimation was reduced to 1.4 ¼C, with a standard deviation of 0.5 ¼C and differences within the ±2.5 ¼C limits specified for the LST product for all the days of the comparison. Table 2. Average brightness temperatures for the 3 3 pixels closest to the test site centre, and μm chan-nels, nadir view. The standard deviation of temperatures, σ, and the satellite viewing angle are also given. Year Date (day/month) Viewing angle (º) T (ºC) n σ(t ) (ºC) n (ºC) σ(t ) (ºC) n / / / / / / / / / / / / / / / / / / / // / / / The best results for the AATSR LST algorithm were obtained for class i=8 (broadleaf shrubs with groundcover) with f=1, which is appropriate for the fully The precipitable water was taken pw=2.5 cm. (The impact of pw on LST is small: less than 0.04 ¼C for a variation of 1 cm in pw.) Applying Eq. (3) to the AATSR brightness temperatures for the test area, we obtained the LSTs shown in the fifth column of Table 3, together with the differences with the ground LSTs (column 8). Differences ranged between 1.1 and 1.0 ¼C, the average difference was 0.1 ¼C and the standard deviation was 0.5 ¼C. We also validated the LST split-window algorithm For the full cover rice crops of the site, we can take ε=0.983 and Δε=0.005 based on field emissivity measurements. The LST values derived with Eq. (4) for the AATSR data of the test site are shown in the sixth column of Table 3. The differences with the ground data (column 9) ranged from 1.0 to 1.0 ¼C, with zero T n developed rice crops in summer. Using the split-window coefficients of [10] in Eq. (2), the AATSR LST equation locally tuned to our study area is, LST = 0.4[sec(θ)-1]pw (T -T ) n T with LST, T and T in ¼C. of [6], which has a quadratic dependence on the brightness temperature difference (T -T ), and depends explicitly on surface emissivity through the mean channel emissivity, ε=(ε +ε emissivity difference, Δε=ε (3) )/2, and the channel -ε. The algorithm coefficients were calculated from a regression analysis over a database of simulated top-of-the-atmosphere AATSR radiances covering global atmospheric conditions. The algorithm can be written as LST = T (T T ) (T T ) (1 ε) 55Δε (4) average bias and standard deviation of 0.5 ¼C. Finally, Fig.3 shows graphically the comparison of the ground measured LSTs against the AATSR derived LSTs for the three split-window formulations studied (RAL processor, Eq. (3) and Eq. (4)). The overestimation of the LSTs provided by the RAL

6 processor is clearly observed, as well as the good agreement of both the AATSR LST algorithm optimised for the test site and the quadratic, emissivity dependent algorithm with the ground LSTs. Table 3. Comparison of ground LSTs with AATSR derived LSTs given by the RAL processor, Eq. (3), and Eq. (4). The last three columns show the difference between the ground LST and the AATSR LST for each algorithm. Year Date (day/ month) Ground LST (¼C) AATSR LST (¼C) Ground AATSR LST (¼C) RAL Eq. (5) Eq. (7) RAL Eq. (5) Eq. (7) 10/ / / / / / / / / / / / / / / / / / / // / / / Average difference ( ¼C) Standard deviation ( ¼C) CONCLUSIONS The results shown in this study stress the need for a good specification of the split-window coefficients for LST retrieval, based either on land cover classification and vegetation cover fraction estimates, or with the surface emissivity. In these cases (Eqs. (3) or (4)), the split-window technique can provide the LST well within the accuracy specified for the AATSR LST product (±2.5 ¼C). The 0.5¼ 0.5¼ longitude/latitude resolution used by the AATSR LST operational algorithm in order to assign values for i and f is too coarse to account for the high heterogeneity of land surfaces. This could lead to large LST errors in particular areas, as shown for the Valencia test site. A much finer spatial resolution in land cover classification and vegetation fraction (ideally the same as for the LST product, i. e., 1 km 2 ), should be recommended in order to improve the LST accuracy.

7 ACKNOWLEDGMENTS This work was financed by the Ministerio de Ciencia y Tecnolog a (Acci n Especial REN E/CLI; Project REN /CLI, and Ram n y Cajal Research Contract of Dr. E. Valor), the Ministerio de Educaci n y Ciencia (Project CGL C03-01, co-financed with European Union FEDER funds, Acci n Complementaria CGL E and Research Contract of Dr. R. Nicl s), and the University of Valencia (V Segles Research Grant of Mr. J. M. S nchez). We thank the AATSR Validation Team, University of Leicester, and the European Space Agency (under CAT-1 project 3466) for providing the AATSR data. REFERENCES 1. Prata, A. J., Land surface temperature measurement from space: AATSR Algorithm Theoretical Basis Document, Technical Report, CSIRO, 27 pp., Wan, Z., Zhang, Y., Zhang, Q. and Li, Z.-L., Validation of the land-surface temperature products retrieved from Terra Moderate Resolution Imaging Spectroradiometer data. Remote Sens. Environ., 83, , Prata, A. J., Land surface temperature measurement from space: Validation of the AATSR Land Surface Temperature product, Technical Report, CSIRO, 40 pp., Coll, C., Valor, E., Caselles, V., Nicl s, R., Rivas, R., S nchez, J. M. and Galve, J. M., Evaluation of the Envisat-AATSR land surface temperature algorithm with ground measurements in the Valencia test site. Proc. Envisat Symposium, 6-10 Sept. 2004, Salzburg, Austria, ESA SP-572, Coll, C., Caselles, V., Galve, J. M., Valor, E., Nicl s, R., S nchez, J. M. and Rivas, R., Ground measurements for the validation of land surface temperatures derived from AATSR and MODIS data. Remote Sens. Environ., 97, , Coll, C. and Caselles, V., A global split-window algorithm for land surface temperature from AVHRR data: Validation and algorithm comparison, J. Geophys. Res., 102D: 16,697-16,713, Llewellyn-Jones, D., Edwards, M. C., Mutlow, C. T., Birks, A. R., Barton, I. J. and Tait, H., AATSR: Global-change and Surface-Temperature measurements from ENVISAT. ESA Bulletin, February 2001, -21, Rubio, E., Caselles, V., Coll, C., Valor, E. and Sospedra, F., Thermal-infrared emissivities of natural surfaces: Improvements on the experimental set-up and new measurements, Int. J. Remote Sensing, 24 (24), , Salisbury, J. W. and D Aria, D. M., Emissivity of terrestrial materials in the 8-14 μm atmospheric window. Remote Sens. Environ., 42, , Prata, A. J., Land surface temperature measurement from space: Global surface temperature simulations for the AATSR. Technical Report, CSIRO, 15 pp, 2002.

8