Issues in reproducibility of spectral measurements

Issues in reproducibility of spectral measurements

This talk was given at the EUROSPEC Plenary Conference, Milan 2012. It covers a number of issues with the reproducibility of spectral measurements, focussing particularly on atmospheric issues.

B5315e9d0420a5546727a9f4f04010b6?s=128

Robin Wilson

March 26, 2012
Tweet

Transcript

  1. Institute for Complex Systems Simulation Issues in reproducibility of spectral

    measurements Robin Wilson Geography and Environment University of Southampton
  2. Institute for Complex Systems Simulation 2 •  Early in my

    PhD •  Raising questions/points for discussion •  Bring together a variety of projects from Southampton Starting point: We need reproducible measurements or we can’t produce a robust relationship with FLUXNET data This talk
  3. Institute for Complex Systems Simulation How to be reproducible: theory

    •  Perfect sensor –  Stable with temperature –  Small FOV –  Accurate measure of total irradiance (panel or irradiance sensor) •  Black sky with point sun •  Constant sun –  Intensity –  Angle 3
  4. Institute for Complex Systems Simulation How to be reproducible: practice

    •  Imperfect sensor •  Blue sky –  Constantly changing atmosphere –  Sub-visual clouds •  Moving sun 4
  5. Institute for Complex Systems Simulation Best Practice •  We’ve come

    up with some generally-accepted guidance: –  Take simultaneous measurements of irradiance and target –  Measure irradiance from cosine corrected receptor or calibrated reference panel –  Ensure target and reference spectrometers are intercalibrated (preferably in the field – Anderson and Milton, 2006) –  Use careful technique 5
  6. Institute for Complex Systems Simulation Any problems? •  We still

    have limitations in reproducibility – WHY? The Atmosphere •  It is very temporally variable •  It produces an uneven, variable angular distribution of irradiance –  Reference panel is Lambertian; Target IS NOT –  We need to know about the sky irradiance conditions 6
  7. Institute for Complex Systems Simulation 7 Measuring atmospheric properties • 

    Use a Cimel Expensive •  Use a Microtops Not automatic •  How do they do it for satellite image correction? Horizontal Visibility – easy and cheap to measure AOT 1
  8. Institute for Complex Systems Simulation Visibility -> AOT conversion • 

    Very frequently used (built-in to ATCOR, MODTRAN, 6S) •  Accuracy assessed using coincident AOT and Visibility measurements (at Chilbolton, Hampshire, UK) •  High RMSE (Koschmieder, 1925) 6S Simulation ΔRadiance 6-50% ΔNDVI 1-2% Wilson, R. T., Milton, E. J. and Nield, J. M., On the accuracy of visibility-derived estimates of AOT, Remote Sensing of Environment (submitted)
  9. Institute for Complex Systems Simulation 9 What about clouds?

  10. Institute for Complex Systems Simulation •  We measure the Hemispherical

    Conical Reflectance Factor •  Irradiance: Hemispherical; Measurement: Conical •  But this irradiance isn’t uniform – it varies temporally, angularly, spatially… •  Atmospheric properties, clouds, nearby surfaces •  Few models take clouds into account when correcting satellite data or spectral data HCRF and angular incident radiance
  11. Institute for Complex Systems Simulation Moving clouds 11 Total irradiance

    is constant Angular distribution changes Can we produce a simple model to illustrate this?
  12. Institute for Complex Systems Simulation •  Monte Carlo Ray Tracing

    •  2D grid •  Each cell absorbs and may scatter probabilistically (wavelength dependent) •  Pseudo-angular response at sensor •  Exploratory model – results not accurate but representative •  Teaching use? 12 RTWRTM Wilson, R. T., “Can a single cloud spoil the view?”: Modelling the effect of an isolated cumulus cloud on surface solar irradiance, MSc dissertation, University of Southampton
  13. Institute for Complex Systems Simulation •  Add a simple cloud

    (with different aerosol type, particle density, water vapour content etc.) •  Look at how received irradiance differs 13 RTWRTM
  14. Institute for Complex Systems Simulation •  Try and measure these

    effects in real life Next step
  15. Institute for Complex Systems Simulation Tramway System 8 bands 4

    surfaces Anderson, K., Milton, E. J., Odongo, V. and Dungan, J. L., 2010, On the reproducibility of reflectance factors: implications fo EO science, Proceedings of the Ninth International Symposium on Spatial Accuracy Assessment
  16. Institute for Complex Systems Simulation Tramway System •  Good – 

    Automatic and unattended –  Data from a variety of atmospheric conditions and previous weather conditions •  Bad –  Time gap between measurement of different surfaces – atmosphere will have changed –  Reference panel imperfect 16
  17. Institute for Complex Systems Simulation Tramway Results •  Measurements were

    significantly affected by time difference (1 minute) between reference and target measurements •  Antecedent weather had a large impact (dry -> wet surfaces) •  Different surfaces responded differently to changes in sky conditions –  Gravel: Overcast -> Increased reflectance –  Tiles: Overcast -> Reduced reflectance 17
  18. Institute for Complex Systems Simulation Sky Radiance Mapper (SKRAM) 18

    Andrew McGonigle, University of Sheffield
  19. Institute for Complex Systems Simulation Sky Radiance Mapper (SKRAM) • 

    Developed for NCAVEO field experiment (www.ncaveo.ac.uk) •  Irradiance of the sky at many angles (zeniths/azimuths) •  Produces a map of irradiance –  Essential for BRDF work –  Could be used to correct spectral measurements •  Takes a long time to take the measurements –  Atmosphere will have changed over the time! •  Not automatic 19
  20. Institute for Complex Systems Simulation •  We want: Lots of

    angles, hyperspectral, automatic Fast 20 Can we do better than this? Skydome/HemiSpec Choi, K.Y. and Milton, E. J., 2011, Development of Airborne Hemispheric Spectrometer, QA4EO
  21. Institute for Complex Systems Simulation SkyDome •  NERC Technology Proof

    of Concept instrument •  Calibration/Characterisation by NERC FSF •  No moving parts •  143 Narrow Angle Probes 1 Wide Angle Probe •  Spectrometer equivalent to 144 USB2000’s 21 Choi, K.Y. and Milton, E. J., 2011, Airborne Hyperspectral Skydome: The new dimension of hyperspectral sensing, EARSeL
  22. Institute for Complex Systems Simulation What can we do with

    this? •  Atmospheric irradiance measurements at all angles •  Ground surface measurements at all angles •  Well calibrated data •  Use on ground, platform, aircraft… •  Measurements in a few seconds Tells us all that we need to know! 22
  23. Institute for Complex Systems Simulation SkyDome for FluxNet 23 90∘

    120∘ Flux tower Surface reflected radiance WAP Incident Solar radiance •  Measures both surface and sky •  Same spectrometer for all measurements –  No inter-calibration issues •  Perfect? –  Expensive! (£70,000) Choi, K.Y. and Milton, E. J., 2011, Development of Airborne Hemispheric Spectrometer, QA4EO
  24. Institute for Complex Systems Simulation Is there a cheaper option?

    •  Use data from SkyDome to choose a few essential angles/ wavelengths to measure •  Create a cheap instrument to just measure these •  Set up a wireless sensor network using these cheap instruments to measure: –  Different locations –  Different surfaces –  Frequent measurements 24
  25. Institute for Complex Systems Simulation •  LED spectroscopy – cheap

    therefore can make many sensors •  Logging to an Android mobile phone – easy to send data back Cheap wireless spectral sensor network? 25 Possible way forward…?
  26. Institute for Complex Systems Simulation Summary/Conclusions •  We can’t ignore

    the atmosphere •  Reference panels are not enough –  Even when coincident measurements are taken! •  Must try and characterise atmosphere & irradiance •  Not easy to do this – often expensive Various options – what should we focus on? 26
  27. Institute for Complex Systems Simulation Relevant references •  Anderson, K.,

    Milton, E. J., Odongo, V. and Dungan, J. L., 2010, On the reproducibility of reflectance factors: implications for EO science, Proceedings of the Ninth International Symposium on Spatial Accuracy Assessment •  Anderson, K. and Milton, E. J., 2006, On the temporal stability of ground calibration targets: implications for the reproducibility of remote sensing methodologies, IJRS •  Anderson, K, Milton, E. J. and Rollin, E. M., 2006, Calibration of dual-beam spectroradiometric data, IJRS •  Brooks, D. R. and Mims, F. M., 2001, Development of an inexpensive handheld LED-based sun photometer for the GLOBE program, Journal of Geophysical Research •  Choi, K.Y. and Milton, E. J., 2011, Development of Airborne Hemispheric Spectrometer, QA4EO •  Choi, K.Y. and Milton, E. J., 2011, Airborne Hyperspectral Skydome: The new dimension of hyperspectral sensing, EARSeL •  Milton, E. J., Schaepman, M. E., Anderson, K., Kneubuhler, M. and Fox, N., 2009, Progress in field spectroscopy, RSE •  Wilson, R. T., Milton, E. J. and Nield, J. M., On the accuracy of visibility-derived estimates of AOT, Remote Sensing of Environment (submitted) •  Wilson, R. T., “Can a single cloud spoil the view?”: Modelling the effect of an isolated cumulus cloud on surface solar irradiance, MSc dissertation, University of Southampton 27