Landsat Calibration: Interpolation, Extrapolation, and Reflection L DCM Sc ie nc e T e a m Me e ting USGS E ROS Aug ust 16-18, 2011 De nnis He lde r, Da ve Aa ro n And the I P L a b c re w
Outline • I nte rpo la tio n: Wha t ha s b e e n do ne to c a lib ra te the L a ndsa t a rc hive ? • E xtra po la tio n: Ho w is c a lib ra tio n g o ing to e xte nd to the L DCM e ra ? • Re fle c tio n: Ca lib ra tio n, the Sc ie nc e T e a m a nd…
Interpolation • Whe re we re we whe n we sta rte d this da nc e in Ja nua ry 2007? – L a ndsa t 7 E T M+ wa s sta b le with c a lib ra tio n to 5% unc e rta inty – L a ndsa t 5 T M wa s unsta b le b ut c ha ra c te rize d, c ro ss-c a l’ d to E T M+ with 3-5% pre c isio n. No w 27 ye a rs o ld! • Wha t didn’ t we kno w in Ja nua ry 2007? – L a ndsa t 4 T M c a lib ra tio n (a ltho ug h ne a rly do ne ) – L a ndsa t MSS c a lib ra tio n • 5 se nso rs x 4 b a nds x 6 de te c to rs = 120 c ha nne ls • Co nsiste nt with e a c h o the r? Ab so lute ? ? – Use o f Pse udo I nva ria nt Ca l Site s (PI CS) • E xte nd b a c k to 1972? • Da ta a va ila b le ? • E no ug h pre c isio n?
Interpolation (2) • Whe re a re we to da y in Aug ust 2011? – L a ndsa t 4 T M c a lib ra tio n do ne – L a ndsa t 1-5 MSS se nso rs do ne • Co nsiste nt with e a c h o the r • Pla c e d o n a n a b so lute sc a le – Co nfide nt tha t the PI CS a ppro a c h c a n pro vide 3% pre c isio n
Interpolation (3) Table 11. Landsat Sensor Absolute Radiometric Calibration Uncertainties (%) Landsat-7 Landsat-5 Landsat-4 Landsat-5 Landsat-4 Landsat-3 Landsat-2 Landsat-1 ETM+ TM TM MSS MSS MSS MSS MSS Band 1 5 7 9 8 9 9 10 11 Band 2 5 7 9 8 9 9 10 11 Band 3 5 7 9 9 10 10 11 12 Band 4 5 7 9 14 18 18 22 25 Band 5 5 7 9 Band 7 5 7 9 Band 8 5 • F ro m F o rty-Ye a r Ca lib ra te d Re c o rd o f E a rth Re fle c te d Ra dia nc e fro m L a ndsa t: A Re vie w – By Bria n Ma rkha m a nd De nnis He lde r, Re mo te Se nsing o f the E nviro nme nt, Vo l. So me time so o n…
Interpolation (4) • Do e s a ll this c a lib ra tio n e ffo rt, mo stly in the de se rt, a c tua lly impro ve thing s? – A q uic k study in the fo re sts o f Wa shing to n sta te … – L a ndsa t 5: 20 MSS a nd 16 T M sc e ne s fro m 1984 – 1992. 7 sa me da y sc e ne s. – Nine Hype rio n sc e ne s fo r ta rg e t spe c tra
Site Selection • Se le c tio n o f ve g e ta te d site fo r c ro ss c a l is de pe nde nt o n – Na ture o f ve g e ta tio n: no t c ha ng ing fre q ue ntly – Ho mo g e ne ity o f Ve g e ta tio n – Ava ila b ility o f hype rspe c tra l sig na ture o f ta rg e t a re a – Ava ila b le c lo ud-fre e T M a nd MSS sc e ne s • Co nife ro us fo re st site – lo c a te d a t WRS Pa th/ Ro w- 46/ 28 – I n Wa shing to n Sta te
ROI Selection ROI 4: 0.678 ROI 2: 0.414 ROI 1: 0.550 ROI 3: 2.527 km 2 km 2 km 2 km 2 26X 29 Pixe ls 22 X 21 Pixe ls 34 X 18 Pixe ls 52 X 54 Pixe ls
Spectral Signature of Target overlapped with TM and MSS RSR L5 MSS and TM RSR Profile (Band 1-4) with Target Spectral Signature 1 TM MSS 0.9 Average Minimum Maximum 0.8 Minimum - 5/ 31/ 2007 Normalized Response 0.7 Ma ximum – 9/ 7/ 2005 0.6 0.5 0.4 0.3 0.2 0.1 0 400 500 600 700 800 900 1000 1100 1200 Wavelength(nm)
MSS to TM Consistency: Forests No Ca lib ra tio n With Ca lib ra tio n With Ca lib ra tio n No SBAF Co rre c tio n No SBAF With SBAF Δ =3% Δ =7% Δ =23% Δ =3% Δ =8% Δ =16%
MSS to TM Consistency: Forests With Ca lib ra tio n With Ca lib ra tio n No Ca lib ra tio n No SBAF With SBAF No SBAF Co rre c tio n Δ =8% Δ =41% Δ =34% Δ =3% Δ =6% Δ =3%
Interpolation Extrapolation • Se c o nd a re a o f inte re st/ c o nc e rn wa s de te c to r re la tive g a ins, unifo rmity, b a nding , e tc . – No te AL I sc e ne in the b a c kg ro und • T his pro vide s the pe rfe c t se g ue into … 1. Relative SCA-to-SCA 1. Post-Image Bias removal Correction based on 2. SCA based RG the ten detector correction overlap EO12005070130654_SGS_01
Extrapolation (1) • Wha t a re we g e tting with the OL I se nso r? – Co mme nts a lso g e ne ra lly a pply to T I RS – Be tte r dyna mic ra ng e – Be tte r sig na l-to -no ise ra tio – Be tte r ra dio me tric re so lutio n – Be tte r a b so lute c a lib ra tio n – Be tte r sta b ility(? )
OLI Radiometric Performance (Slide courtesy Brian Markham) SNR – SNR significantly exceeds Median SNR requirements and heritage Calibration – Absolute uncertainty ~4% Extensive round robin for validation Transfer-to-Orbit uncertainties included – Stability over 60 seconds (2 standard scenes) <0.02% 2 σ 16 Day Stability – Stability over 16 days (time between Solar Diffuser Cals) Change in Response, <0.54% 2 σ for all but Cirrus Green band, w/ hot Band which is <1.19% cycle in middle SPIE Earth Observing Systems XVI NASA GSFC / USGS EROS
Extrapolation • Co mpa riso n o f ETM+ High Gain OLI ETM+/OLI ra dio me tric re so lutio n o f E T M+ a nd OL I Band Min Sat Level Rad./DN Min Sat Level Rad./DN Res. Ratio – E 5.2 T M+ = 8 b its Blue 190 0.742 581 0.142 5.7 Green 194 0.758 544 0.133 – OL I = 12 b its 5.2 Red 150 0.586 462 0.113 • Ba se d o n pub lishe d 8.5 NIR 150 0.586 281 0.069 do c ume nts 7.1 SWIR 1 31.5 0.123 71 0.017 – L a ndsa t 7 Sc ie nc e Da ta 7.4 SWIR 2 11.1 0.043 24 0.006 Use rs Ha ndb o o k 4.8 PAN 156 0.609 515 0.126 – L DCM OL I Re q uire me nts Do c ume nt • 5—8 time s impro ve d ra dio me tric re so lutio n with the SNR to suppo rt it! 15
Excerpts from OLI Requirements 5.6.2.3 Pixe l- to- Pixe l Unifor mity • ie ld o f Vie w 5.6.2.3.1 F ull F – T he sta nda rd de via tio n o f a ll pixe l c o lumn a ve ra g e ra dia nc e s a c ro ss the F OV within a b a nd sha ll no t e xc e e d 0.5% o f the a ve ra g e ra dia nc e . • 5.6.2.3.2 Banding – T he ro o t me a n sq ua re o f the de via tio n fro m the a ve ra g e ra dia nc e a c ro ss the full F OV fo r a ny 100 c o ntig uo us pixe l c o lumn a ve ra g e s o f ra dio me tric a lly c o rre c te d OL I ima g e da ta within a b a nd sha ll no t e xc e e d 1.0% o f tha t a ve ra g e ra dia nc e . • 5.6.2.3.3 Str e aking – T he ma ximum va lue o f the stre a king pa ra me te r within a line o f ra dio me tric a lly c o rre c te d OL I ima g e da ta sha ll no t e xc e e d 0.005 fo r b a nds 1-7 a nd 9, a nd 0.01 fo r the pa nc hro ma tic b a nd. T he se r e quir e me nts allow the pr e se nc e of str iping and banding…
OLI Scene Simulation • L a ke T a ho e simula te d OL I ima g e b e fo re g a in/ b ia s c o rre c tio n • Co ur te sy Jo hn Sc ho tt/ RIT via DIRSIG – F ully synthe tic sc e ne • OL I Simula tio n – 14 a rra ys – 60 de te c to rs e a c h; a c tua l va lue s – 12 b it q ua ntiza tio n – Ac tua l OL I no ise le ve ls – Ac tua l spe c tra l re spo nse – Ac tua l de te c to r g a ins/ b ia se s – Sa mple d o b se rve d no n-line a rity func tio n – No ra dio me tric c o rre c tio ns a pplie d—ra w da ta – Pe rfe c t g e o me try 17
OLI Scene Simulation (2) • L a ke T a ho e I ma g e a fte r g a in a nd b ia s c o rre c tio n – No no n-line a rity c o rre c tio n a pplie d • Be a utiful! 18
OLI Scene Simulation (3) • Ga in/ b ia s c o rre c te d ima g e with la nd stre tc h – Sq ua re ro o t stre tc h • Be a utiful! 19
OLI Scene Simulation (4) • Wa te r stre tc h o n L a ke T a ho e Simula te d I ma g e – L ine a r 2% • Striping • Ba nding • No ise • OL I (and T IRS) will be be tte r than anything you’ve se e n , b ut the y will ha ve ‘ a dditio na l fe a ture s’ 20
Extrapolation • OL I a nd T I RS will b e sub sta ntia lly b e tte r tha n a ny pre vio us L a ndsa t se nso r with re spe c t to ra dio me tric pe rfo rma nc e • Sub sta ntia l inc re a se in ra dio me tric re so lutio n a nd SNR will a llo w use rs to de te c t the sig na ture o f the instrume nt in ho mo g e ne o us re g io ns with se ve re stre tc he s • Stro ng ly sug g e st use rs a c c e pt this a s a n a dditio na l b e ne fit o f using hig h pe rfo rma nc e se nso rs ra the r tha n vie wing it a s a dra wb a c k
Reflections • Wha t a g re a t jo b ! – Nic e to wo rk with so me re a lly sma rt pe o ple fo r a c ha ng e ! • Push the c a lib ra tio n in yo ur a pplic a tio ns – Wha t a re the limits? – Whe re do e s it e xc e e d yo ur ne e ds? – Whe re do e s the c a l fa ll sho rt? • Wha t’ s the va lue pro po sitio n? • Ho w do yo u se ll a 40 ye a r pro g ra m to a 2 ye a r g o ve rnme nt?
Recommend
More recommend
