The Inframetrics 760 airborne thermal infrared data

Transcription

1 The Inframetrics 76 airborne thermal infrared data on the ReSeDA experiment Quality assessment and first recalibration of brightness temperature maps Table of Contents 7 Quality assessment and first recalibration of brightness temperature maps Field data acquisition Occasional measurements provided by the portable infrared radiometer Everest Interscience AGRITHERM 112 ALCS Occasional measurements provided by the CIMEL infrared radiometer Routine measurements provided by the Heimann KT 1 and KT 17 located on seven fields over the experimental site Comparison between the three field data sets Validation of Inframetrics brightness temperature and in-situ calibration Comparison against field measurements of directional brightness temperature images Comparison of nadir brightness temperature maps to field measurements In-situ Calibration

2 7 Quality assessment and first recalibration of brightness temperature maps Brightness temperature maps (at nadir ) were compared to ground measurements of brightness temperature in order to check map validity. As large differences existed, a recalibration was performed. Differences were due both to radiometric problem observed on the Inframetrics instrument, and to the normalisation procedure presented in Chapter 4 (4-Accounting for temporal instability of the sensor) 7.1 Field data acquisition Occasional measurements provided by the portable infrared radiometer Everest Interscience AGRITHERM 112 ALCS Field data were acquired simultaneously to the airborne observations for days between August, 7th and September, 18th. The instrument used was an Everest Inter-science manual infrared thermometer AGRI-THERM 112ALCS, which measured target brightness temperature from incoming radiance over the [8 m- 14 m] spectral band (close to that of the airborne sensor). Data corresponded to an average of ten instantaneous values over a second period. The FOV of the instrument lens was 4 o. Measurements were performed from nadir at 1 m height, providing a cm diameter circular footprint. Each field was exhaustively sampled using a m grid pattern with measurements acquired in approximately half an hour. Two water surfaces with known temperatures (one cold and one hot) were used as reference targets for calibration that was repeated every 2 measurements to provide a daily calibration (Verbrugghe & Guyot, 1992). The RMSE (Root Mean Square Error) between calibration data and linear regression was about.6 o C. A description of the measurements is given in Table 1. The sampled alfalfa, grassland, sunflower and bare soils field exhibited brightness temperatures ranging from 2 o C to 4 o C. 2

3 Date Crop Field Measurement number features July, 8th Begin blooming 3 Whole canopy alfalfa End blooming 1 Bare soil / vegetation and sunflower shadow / sunlight differentiation July, 9th Begin blooming 3 Whole canopy alfalfa Day / night measurements. End blooming 1 Bare soil / vegetation and sunflower shadow / sunlight differentiation Day / night measurements September, 4th Begin senescent 121 Bare soil / vegetation and sunflower shadow / sunlight differentiation Bare soil 2 Alfalfa before 3 Whole canopy blooming September, 9th Senescent 121 Bare soil / vegetation sunflower shadow / sunlight differentiation Alfalfa before 3 Whole canopy blooming September, 18th Senescent 121 Bare soil / vegetation sunflower shadow / sunlight differentiation Cut and not 1 Whole canopy cut grassland Table 1: Listing and features of the field measurements acquired using the Everest Interscience AGRITHERM 112 ALCS Occasional measurements provided by the CIMEL infrared radiometer Other nadir measurements were performed by CETP using Cimel CE-312 radiometers on fields 2 (July, th, 8th and 9th), 1 (June, 6th and 7th) and 3 (June 3th and 4th). They were acquired during few hours between 8h and 17h (the delay of acquisition differs from one dataset to another) with a. Hz frequency. They corresponded to the 8-14 m spectral range. A brief description is given in table 2. For more information please contact Christophe François or Catherine Ottlé at CETP. 3

4 Date Crop Field number July, th Growing sunflower 2 July, 8th Growing sunflower 2 July, 9th Growing sunflower 2 June, 3th Alfalfa 3 June, 4th Alfalfa 3 June, 6th Senescent wheat 1 June, 7th Senescent wheat 1 Table 2: Listing and features of the CETP field measurements acquired using the Cimel radiometer Routine measurements provided by the Heimann KT 1 and KT 17 located on seven fields over the experimental site Daily measurements were acquired on seven locations corresponding to alfalfa, wheat and sunflower (see Figure 1 for the locations inside fields 2, 2, 1, 121, 3, 214, 1). The brightness temperatures were obtained from measurements of the incoming radiance over the [8-14] m using Heimann KT 1 and KT 17 radiometers. These radiometers aimed the surfaces inside a solid angle ranging from 2. and 34. o in view zenith angle, which induced a footprint between 2.6 and 8.6 m (see Figure 2). These routine data were acquired with a 1 second frequency and a minutes period averaging. The instruments were calibrated before and after the experiment. It accounted for the sensor internal temperature that was recorded during the data acquisition. The accuracy ranged between.3 and.8 o C from one instrument to another. 4

5 1 met Figure 1: Location of the in-situ measurements inside the 7 seven test fields / 3.1 m 18. o +/ 16 o Figure 2: Scheme of the observation conditions with the HEINMANN KT radiometers. 8.6 / 2.6 m2

6 7.2 Comparison between the three field data sets In order to assess the quality of the surface brightness temperature field measurements, we compare the three available data sets. By accounting for the temporal agreement between these data sets, the comparison was possible in few cases that are presented here. Since the data sets we compared had different temporal and spatial sampling, we compared them in the following ways. The comparison between the Heimann KT and Cimel data was performed using averaged values of the Cimel measurements (acquired with a. Hz frequency) over the acquisition period of the Heimann KT ones (one value every minutes). The comparison of the Heimann KT and Everest measurements was performed using averaged values of the two data sets inter-compared over the period of Everest data acquisition that was about half an hour. In the following discussion, the figures that display the results show also the standard deviation computed simultaneously with the considered averaged values. The Heimann KT and Cimel measurements were significantly different over field 2 (see Figure 3), which was explained by their different locations over a heterogeneous field (see Figures 4 and ). The Cimel and Heimann KT measurements agreed well over field 1 (see Figure 6) that was very homogeneous at this time. The comparison between the Heimann KT and Everest data on field 121 were illustrated using histograms since the two data sets were significantly different: the Heimann KT measurements integrated the whole canopy inside the FOV whereas the Everest ones were performed by distinguishing the soil / vegetation components and the illuminated / shaded ones. Even if the Heimann KT data were included in the range of the Everest ones (see Figure 7), we were not able to draw a general conclusion. For field 3, we observed an underestimation of both the Everest and Cimel measurements from the Heimann KT ones whatever was the day and the time of acquisition (see Figures 8 and 9). Field 3 was uasually homogeneous. However, we had a lot of difficulties of getting right the calibration of the Heimann KT radiometer set on this site. Finally, the comparisons presented here showed that the field measurements were close when being comparable. 6

7 Heimann KT Field 2, RMSE=7.73 C, BIAS= 6.73 C Cimel Figure 3: Comparison between Heimann KT and Cimel surface brightness temperature measurements for the field 2 (Sunflower). The RMSE is the Root Mean Square Error. The BIAS is the mean difference between the Cimel and Heimann KT measurements. The horizontal segment around each point gives the value of the standard deviation corresponding to the Cimel averaged value used for the intercomparison. LAI, Field : 2, Date : 9778, Mean : , Std :.334, StdRel: Figure 4: LAI map over field 2 on July, 8th (derived from PolDER data using the method described by Weiss et al. (1)..8 7

8 Figure : Nadir surface brightness map over field 2 on July, 8th. White bullet corresponds to the CETP measurement location (Cimel radiometer). Black bullet corresponds to the daily measurement location (Heimann KT radiometer). 24 Heimann KT Field 1, RMSE=.467 C, BIAS= Cimel Figure 6: Comparison between Heimann KT and Cimel surface brightness temperature measurements for the field 1 (wheat). The horizontal segment around each point gives the value of the standard deviation corresponding to the Cimel averaged value used for the intercomparison. 8

9 18 Repartition (%) Temperature ( C) Figure 7: Comparison between Heimann KT and Everest surface brightness temperature measurements for the field 121 (sunflower). Bars correspond to the histogram of the Everest data. The two lines correspond to the two Heimann KT data collected over the Everest data acquisition period. Heimann KT Field 3; a=.9816; b= 1.613, RMSE=2.1 o C, BIAS= o C Everest Figure 8: Comparison between Heimann KT and Everest surface brightness temperature measurements for the field 3 (alfalfa). The horizontal (respectively vertical) segment around each point gives the value of the standard deviation corresponding to the Everest (respectively Heimann KT) averaged value used for the inter-comparison. 9

10 Heimann KT Field 3, RMSE=1.192 C, BIAS=.9726 C Cimel Figure 9: Comparison between Heimann KT and Cimel surface brightness temperature measurements for the field 3 (alfalfa). The horizontal segment around each point gives the value of the standard deviation corresponding to the Cimel averaged value used for the intercomparison.

11 7.3 Validation of Inframetrics brightness temperature and in-situ calibration Comparison against field measurements of directional brightness temperature images. The validation of the INFRAMETRICS 76 surface brightness temperature estimates by using the Everest field measurements was possible for 4 cases corresponding to spatial and temporal agreements between the data sets. This validation showed a good agreement between the two data sets, with differences ranging from.4 to 1. o C when the footprint of the data sets was similar (Figures to 12). We were not able to draw a conclusion about the results obtained over fied 121 (Figure 13) since the Everest field data set distinguished the both soil/vegetation and shaded/lighted components whereas the airborne data integrated the patchwork inside the FOV. 11

12 1 Date : 4/9/97. Bare soil. Sensor altitude : m. Raw airborne data Mean value :.3 C Standard deviation :.6 C 1 Instrumentally corrected airborne data Mean value :.9 C Standard deviation :.7 C Distribution (%) 1 Instrumentally and atmospherically corrected airborne data Mean value :.3 C Standard deviation : 1.2 C 1 Field data Mean value : 38.9 C Standard deviation : 1.2 C Temperature ( C) Figure : Comparison of the histograms corresponding to both the airborne data at three processing level and the Everest field meaasurement for the field 2 on the September, 4th. 12

13 2 1 Date : 9/9/97. Crop of alfalfa. Sensor Altitude : m. Raw airborne data Mean value : 2. C Standard deviation :.8 C Distribution (%) Instrumentally and atmospherically corrected airborne data Mean value : 28.2 C Standard deviation : 1.1 C Instrumentally corrected airborne data Mean value : 26.2 C Standard deviation :.8 C 2 1 Field data Mean value : 27.7 C Standard deviation : 1.6 C Temperature ( C) Figure 11: Comparison of the histograms corresponding to both the airborne data at three processing level and the Everest field meaasurement for the field 3 on the September, 9th. 13

14 Date : 18/9/97. Grassland. Sensor altitude : m. 1 Raw airborne data Mean value : 2.3 C Standard deviation : 1. C 1 Instrumentally corrected airborne data Mean value : 26.8 C Standard deviation : 1.6 C Distribution (%) 1 Instrumentally and atmospherically corrected airborne data Mean value : 29.7 C Standard deviation : 2.1 C 1 Field data Mean value : 28.7 C Standard deviation : 2 C Temperature ( C) Figure 12: Comparison of the histograms corresponding to both the airborne data at three processing level and the Everest field meaasurement for the field 1 on the September, 18th. 14

15 Date : 4/9/97. Crop of sunflower. Sensor altitude : m. Raw airborne data Mean value : 29.6 C Standard deviation :.7 C Instrumentally corrected airborne data Mean value :.8 C Standard deviation :.8 C Distribution (%) Instrumentally and atmospherically corrected airborne data Mean value : 39.4 C Standard deviation : 1.2 C Field data Mean value : 31.9 C Standard deviation :.6 C Temperature ( C) Figure 13: Comparison of the histograms corresponding to both the airborne data at three processing level and the Everest field meaasurement for the field 121 on the September, 4th. 1

16 7.3.2 Comparison of nadir brightness temperature maps to field measurements. We noticed that this processing induced the removal of the unsystematic error we observed when comparing the airborne measurements against the daily field ones. Figure 14 to display the results of the daily validation. The title gives the date, the flight altitude, the RMSE between field and airborne data. The RMS given inside the figure is the unsystematic RMSE. The field label nomenclature is: 1=field 1; 2=2; 3=1; 4=121; =3; 6=214; 7=1. The point corresponds to the brightness temperature water surface close to the meteorological site in X-abscissa and to the air temperature in Y-abscissa. This point has not been taken into account for the computation of the linear regression and is just presented as a verification. We included also the field measurement provided by the Everest Interscience when they could give more information, i.e. when they corresponded to fields without routine data. They are indicated by the number of the field (2 or 1 for example). We observed that the airborne estimates provided a systematic error of the field estimates with a low discrepancy around the linear regression. Moreover, the results obtained for the July, 28th and 29th, July and the 4th September were unsatisfactory (see Figure 33, Figure 34, Figure 3, Figure 36, Figure 37). Indeed, the validation point corresponding to the field 121 induced a low slope value for the regression linear. Since this field was significantly heterogeneous, we checked the accuracy of the field measurement location. A new location provided an airborne estimate higher about 4 o C, but this difference was not enough large to provide a significant increasing of the slope of the linear regression. At the present time, we think that the troubles should be induced by the sub-pixel variability (footprint of the field measurements about 8 m as compared to the / m of the INFRAMETRICS 76 pixel size). 16

17 Inframetrics 76 SBT ( o C) Daily Validation ;. Rmse : o C T(Field)=1.629*T(I76) Rms= Figure 14: Validation temperature maps against Heimann KT field measurements. Title gives: the date, the flight altitude, the RMSE between field and airborne data. The RMS given inside the figure is the unsystematic RMSE. Field label nomenclature: 1=field 1; 2=2; 3=1; 4=121; =3; 6=214; 7=1. Point : brightness temperature of the water surface close to the meteorological site in X-abscissa and air temperature in Y-abscissa. 17

18 Inframetrics 76 SBT ( o C) Daily Validation ;. Rmse :.342 o C T(Field)=1.312*T(I76) Rms= Figure 1: Validation temperature maps against Heimann KT field measurements. Title gives: the date, the flight altitude, the RMSE between field and airborne data. The RMS given inside the figure is the unsystematic RMSE. Field label nomenclature: 1=field 1; 2=2; 3=1; 4=121; =3; 6=214; 7=1. Point : brightness temperature of the water surface close to the meteorological site in X-abscissa and air temperature in Y-abscissa. Daily Validation ;. Rmse :.9372 o C 8 Inframetrics 76 SBT ( o C) T(Field)=1.1723*T(I76) Rms= Figure 16: Validation 18

19 8 Daily Validation. 974;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=1.243*T(I76) Rms= Figure 17: Validation 8 Daily Validation. 974;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=1.1711*T(I76) Rms= Figure 18: Validation 19

20 Daily Validation ;. Rmse :.4841 o C 8 Inframetrics 76 SBT ( o C) T(Field)=1.1714*T(I76) Rms= Figure 19: Validation Daily Validation ;. Rmse :.3263 o C 8 Inframetrics 76 SBT ( o C) T(Field)=1.268*T(I76) Rms= Figure : Validation

21 8 Daily Validation ;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=1.148*T(I76) Rms= Figure 21: Validation 21

22 8 Daily Validation. 971;. Rmse : 6.4 o C Inframetrics 76 SBT ( o C) T(Field)=1.317*T(I76) Rms= Figure 22: Validation 22

23 8 Daily Validation. 972;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=1.31*T(I76) Rms= Figure 23: Validation 23

24 8 Daily Validation. 971;. Rmse :. o C Inframetrics 76 SBT ( o C) T(Field)=1.78*T(I76) Rms= Figure 24: Validation 24

25 Daily Validation. 9722;. Rmse :.817 o C 8 Inframetrics 76 SBT ( o C) T(Field)=1.491*T(I76) Rms= Figure 2: Validation Daily Validation. 9722;. Rmse :.4811 o C 8 Inframetrics 76 SBT ( o C) T(Field)=1.347*T(I76) Rms= Figure 26: Validation 2

26 8 Daily Validation. 9769;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=1.2246*T(I76) Rms= Figure 27: Validation 8 Daily Validation. 9769;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=1.2874*T(I76) Rms= Figure 28: Validation 26

27 8 Daily Validation ;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=1.366*T(I76) Rms= Figure 29: Validation 27

28 8 Daily Validation ;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=.89234*T(I76) Rms= Figure : Validation 28

29 Daily Validation. 9778;. Rmse : o C 8 Inframetrics 76 SBT ( o C) T(Field)=1.4363*T(I76) Rms= Figure 31: Validation Daily Validation. 9778;. Rmse :.26 o C 8 Inframetrics 76 SBT ( o C) T(Field)=1.6832*T(I76) Rms= Figure 32: Validation 29

30 8 Daily Validation ;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=2.7922*T(I76) Rms= Figure 33: Validation

31 8 Daily Validation ;. Rmse : 4.8 o C Inframetrics 76 SBT ( o C) T(Field)=2.469*T(I76) Rms= Figure 34: Validation 8 Daily Validation ;. Rmse :.729 o C Inframetrics 76 SBT ( o C) T(Field)=2.933*T(I76) Rms= Figure 3: Validation 31

32 8 Daily Validation. 9794;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=1.62*T(I76) Rms= Figure 36: Validation 8 Daily Validation. 9794;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=2.112*T(I76) Rms= Figure 37: Validation 32

33 8 Daily Validation. 9799;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=1.79*T(I76) Rms= Figure 38: Validation 33

34 8 Daily Validation ;. Rmse : 2.39 o C Inframetrics 76 SBT ( o C) T(Field)=.92961*T(I76) Rms= Figure 39: Validation 8 Daily Validation ;. Rmse : o C Inframetrics 76 SBT ( o C) T(Field)=1.684*T(I76) Rms= Figure : Validation 34

35 7.3.3 In-situ Calibration The regression lines were used to generate new image products. The evolution of the coefficients of the regression line is shown on Figure 41. However, it seemed that the first generation of that products was not adequate and that wrong maps were obtained. As a matter of fact maps for DOE between and 613 contained values larger than o C. A new product, with an improved recalibration scheme, was provided latter (see Chapter 7-bis). References Verbrugghe, M. & Guyot, G. (1992). Notes sur l étalonnage des radiomètres infrarouge portables. Agronomie, 12, Weiss, M., Jacob, F., Baret, F., Pragnère, A., Leroy, M., Hautecoeur, O., Prévot, L., & Bruguier, N. (1). Evaluation of kernel-driven BRDF models for the normalization of Alpilles/ReSeDA PolDER data. In: Physics and Chemistry of the Earth, EGS symposium, special ReSeDA session. 3

36 Day 3/12 3/26 3/26 4/ 4/ 4/16 4/16 4/18 /1 /2 /1 /22 /22 6/9 6/9 6/12 6/24 7/8 7/8 7/28 7/29 7/29 9/4 9/4 9/9 9/18 9/18 8 Offset b ( o C) 6 Day 1 3/12 3/26 3/26 4/ 4/ 4/16 4/16 4/18 /1 /2 /1 /22 /22 6/9 6/9 6/12 6/24 7/8 7/8 7/28 7/29 7/29 9/4 9/4 9/9 9/18 9/ Slope a 2. x: flight altitude = m +: flight altitude = m 3 3. Slope and Offset of the calibration T Field = a.t I76 + b Figure 41: Validation Heimann KT field measurements. 36