Looking at 637 nm VIIRS band, S-NPP

Transcription

1 Looking at 637 nm VIIRS band, S-NPP (Sharpening I1) B. GUENTHER STELLAR SOLUTIONS, INC NOAA-JPSS 1

2 I am looking at houses and have a desire to know how much living area this floor plan has for first floor living. It has a garage that is 21 8 X 20 4 and is built on a 40' X 50' slab Sorry, the picture is very grainy. 2

3 How Would You Do it to Get the Measure of the Living Space? For the house, I'd compute the area of garage and subtract that from 2000ft-sq Another example Installing sheetrock, need piece 8 ft tall by 3ft 3.5in in width Start with full sheet, 4X8 and remove 8.5in (careful which side of line I cut on) Clearly in my mind I am shaving off material to get to final dimension These are everyday questions and are answered easily when the question deals with quantity that is linear 3

4 Pizza sharing, another every day question Now this is easy to split in half, thirds or quarters if the 'measure' is amount of pizza being shared In my house, sharing pepperoni pizza includes 'fair shares' of the pepperoni too Situation is easy when the quantity is properly linearized and not so easy when the quantity is not easily linearized 4

5 VIIRS Imagery Band 1 (I1, 640 nm) Has Large Bandpass Now finally to VIIRS Remember that JPSS started as NPOESS and the Imagery requirements were driven by DoD applications NASA advocated for VIIRS bandpass similar to MODIS Channel 1 at 645 nm to sustain NDVI products DoD advocated for high SNR in high spatial resolution band Also provides easier manufacturing to meet SNR with 375 Km resolution Final specification ended at 80 nm bandpass for I1 Moderate resolution band M5 nested spectrally within I1 Question is how to introduce spectral sharpening to I1, if we may 5

6 RSR for M5 and I1, Actual 6 RSR for M5 and I1, Single Detector Full Spectral Range RSR Expanded RSR for 670 to 695 nm Interval M5 I1 Wavelength in nm RSR Wavelength in nm M5 I1 These bands are clearly not box car, so how do we handle this? Data are for S-NPP VIIRS

7 Process to Sharped I1 I1 must be aggregated to M5 spatial resolution (2 samples in track and 2 samples in scan direction, compute average radiance) Aggregated I1 samples have sharper MTF than M5 Differences in MTF captured in uncertainty 'Align' lw c-o of M5 to I1, called spectral alignment here Compute RSR differences from Chart 7, right side figure (see Chart 9) Need analytic TOA spectral radiance in 675 to 690 nm, includes Rayleigh scatter, aerosol contributions and surface reflectance L surplus is radiance correction needed to align cut-off slopes Integration of L TOA (λ with RSR difference (see Chart 9) Evaluate polarization effects but likely not important Remember need do this in linear space and here that is radiance space Compute in EDR development since surface reflectance(, polarization) and aerosol contributions unknown when SDR computed operationally 7

8 Visualization of Spectral Alignment 1.2 Expanded RSR for 670 to 695 nm Interval RSR difference RSR Wavelength in nm M5 I1 L surplus is adjusting the TOA radiance in long wave cut-off of M5 greater than I1 8

9 Formula, Just to Write it Down I1-sharpened = (I1-M5-L surplus ) Think of L surplus as a 'signed' value L surplus is positive when M5 RSR lw c-o is longer in wavelength than is the I1 RSR lw c-o 9

10 Channel 1 Comparisons 1.2 Relative Spectral Response Wavelength (nm) MODIS M1 VIIRS I1 VIIRS M5 VIIRS I1-sharpened Bands are MODIS M1, VIIRS I1, VIIRS M5 and VIIRS I1-sharpened Shown as Box Car RSR approximations (tails added to VIIRS I1 only for projection clarity Strategy and guidance on how (if?) I1-sparpened may be constructed We already agreed that I1-sharpening may be done with simple subtraction (I1-M5) when the subtraction is done in appropriate linear space, [and adjusted if longwave cut-off (lw c-o) for two bands not congruent] For Box Car RSR the linear space is no more complicated than measured radiance I1-sharpening is no more complicated than presented in earlier examples for floor space and sizing of sheetrock 10

11 I1-sharpened in Application to NDVI for Channel 1 Sensor Band Band Center (nm) Bandpass (nm) AVHRR (Channel 1) MODIS M VIIRS I VIIRS I1-sharpened MODIS M1 used here only for comparisons, and no distinction is made between Terra and Aqua sensors since they were built to a common spectral specification 11

12 Grass reflectance (black), I1 (red) and I1-sharpeded (Green) reflectance 0r BANDPASS WAVELENGTH (NM) Reflectance curve is data shown by Larry Leigh, SDSU, yesterday in Workshop 2 Data are an SDSU grass site test hyperspectral reflectance reference curve (unpublished) 12

13 13

14 Wait, Before You Start Using This MUST validate the ideas first May seem crazy, may not VIIRS S-NPP can be used for concept validation Ground truth for the I5-sharpened can be obtained from overpass data with HICO I1-sharpened is the item to be validated L surplus is the item that may be tuned to improve uncertainty in I1- sharpened Process also will allow estimate of uncertainty for I1-sharpened radiometry Uncertainty is relative calibration scales for I1 and M5 will complicate this computation 14

15 Lessons for New Instrument Builders Higher spatial resolution frequently dictates broader spectral measurements to maintain SNR for manufacturability More careful alignment of band edges between the two bands will facilitate band sharpening with lower uncertainty of the resulting band Provides strategy to distinguish physical processes at the larger FOV measurements than can be done using individual bands Please never again provide multiple spatial resolution measurements with common spectral resolution filter Only up-side is to verify that 1+1=2, OK, actually 2+2 = 4 VIIRS down-side: 2 of 16 moderate resolution bands are nearly useless and waste precious internal sensor data rate 15

16 Summary and Conclusions Logic is presented to provide spectral sharpening of the VIIRS I1 bands, at moderate spatial resolution I1 (centered at 640 nm, 80 nm bandpass) presents as I1 sharp (centered at 630 nm, 60 nm bandpass) Accuracy of sharpened band depends on how closely aligned the band cut-offs match and the extent And extent that scene physics is "benign" to allow accurate bandpass alignment Computation is simple mathematical subtraction after adjustments for cut-off alignment Approach broadly applicable within constraint of known scene physics for cut-off alignment in L surplus 16

17 Parting Remarks - I Idea occurs due to fortuitous similarity in VIIRS S-NPP spectral shape for I1 and M5 Also driven by concepts to tune data content for Direct Broadcast data Usefulness of spectral stratification depends entirely on uncertainty involved (and maybe complexity by scene physics in computing L surplus Concept does not seem useful for analysis of band-pairs that are not tightly nested in spectral dimension VIIRS J1 filters within 10 nm in lw c-o differences but L surplus will be of opposite sign compared to S-NPP in computing I-sharpened Filter suite for J3 and J4 will be similar to J2 filter suite Work in getting this concept applied will be useful for all VIIRS sensors through ~2035, OK at least as useful as it is for S-NPP 17

18 Parting Remarks - II Spectral sharpening is not something ever discussed for VIIRS within my earshot, so I think it is new idea Idea is simple Presume you did not enter room today with these ideas clear in your mind Hope you leave room today with clarity for approaches for spectral sharpening Please send me the paper that you or your grad student comes up with in taking this sketchy talk into actual and useful remote sensing applications Thanks for listening, I had fun putting these ideas together and I hope you enjoyed listening, and maybe even have some better ideas for remote sensing science applications 18

19 Acronyms Acronym Meaning Acronym Meaning AVHRR Advanced Very High Resolution MGS Mid Gain Stage (DNB used in Radiometer transition from day to night observations) Bandpass Full Width at Half Maximum MODIS Moderate Resolution Imaging Spectroradiometer Box Car Spectral shape that is flat top MTF Modulation Transfer Function with steep spectral turn-on DoD Department of Defense NASA National Aeronautics and Space Administration DNB Day Night Band NDVI Normalized Difference Vegetation Index EDR Environmental Data Record (Level 2 Data Product) NPOESS National Polar-orbiting Operational Environmental Satellite System F1 VIIRS S-NPP sensor RSB Reflected Solar Band HICO Hyperspectral Imager for RSR Relative Spectral Response Coastal Ocean I (band) Imagery Band SDR Sensor Data Record (Level 1 Data Product) J1 JPSS 1 Mission S-NPP Suomi-National Polar Partnership JPSS Joint Polar Satellite System SNR Signal Noise Ratio LGS Low Gain Stage (DNB used in TOA Top Atmosphere Radiance day observations) lw c-o Longwave cut-off VIIRS Visible/Infrared Imaging Radiometer Suite M (band) Moderate Resolution Band 19