EOCAPHSI FINAL Briefing RIT Technical Activities John Schott

EOCAP-HSI FINAL Briefing RIT Technical Activities John Schott, RIT PI schott@cis. rit. edu (716)475 -5170 Rolando Raqueno, RIT raqueno@cis. rit. edu(716)475 -6907 http: //www. cis. rit. edu/~dirs January 16 -17, 2001 Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Airborne Hyperspectral Imagery Analysis Assessing Near Shore Water Quality Agriculture CDOM Urban phytoplankton bacteria macrophytes particles & algae Bottom Type A Bottom Type B Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Airborne Hyperspectral Model Imagery Analysis Assessing Near Shore. ALGE Water Quality MODTRAN Agriculture Urban bacteria Hydro. Light CDOM macrophytes Modeling Strategy particles & algae Bottom Type A phytoplankton Bottom Type B • Solar Spectrum Model (MODTRAN) • Atmospheric Model (MODTRAN) • Air-Water Interface (DIRSIG/Hydrolight) • In-Water Model (HYDROMOD= Hydrolight/OOPS + MODTRAN) • Bottom Features(HYDROMOD/DIRSIG) Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Long Term Approach: Integrated hybrid physical models validated and fine tuned by real imagery Modtran ALGE: Hydrodynamic Hydrolight DIRSIG difference RMS Real Image Hyperspectral Water Quality Simulated Image

Hyperspectral Imagery Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Overview: Big Picture [ ] Concentrations Model Inherent Optical Properties Reflectance, r(l) Model Atmosphere Digital Counts Hyperspectral Water Quality Radiance, L Digital Imaging and Remote Sensing Laboratory

Signal Sources Atmosphere to Sensor 80% 10% Air/Water Transition Water/Air Transition In Water Hyperspectral Water Quality 10%

Remote Sensing Water Quality Tool: Hydro. Mod Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

absorption IOPs Absorption Water Total suspended material Chlor a DOC Wavelength Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Normalized Scattering Distribution of the Fournier-Forand Phase Function with Parameters (nu, n) Hyperspectral Water Quality

Example LUT Entries [C]=13 [SM]=0 [CDOM]=0 [C]=0 [SM]=0 [CDOM]=50 Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Look Up Table LUT j [C] k [ CDOM ] i [ SM ] Each entry in the LUT [i. e. LUT (i, j, k)] corresponds to a particular output of the Hydrolight code in the form of a spectral vector. These may be in terms of Lλ(h), -Rλ 0 or +Rλ 0. Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Simple Fitting ST truth data λ TRUE min [(ST - SP)2 ] FALSE [C] j [ CDOM ] LUT [CHL] [TSS] [CDOM] k [ SM ] i [CHL] [CDOM] Sp predicted [TSS] Hyperspectral Water Quality SQ Error Final [CHL] [CDOM] [TSS]

Squared Error Interpolated LUT values observation Squared Error = Σ (RLUT - Robs)2 Iterate using a downhill simplex (Amoeba) algorithm to minimize squared error term. Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Trilinear Interpolation SMi, Cj+1, CDOMk SMl, Cm, CDOMk SMi+1, Cj+1, CDOMk SMi, Cm, CDOMk SMi+1, Cm, CDOMk C Smi+1, Cj, CDOMk SMi, Cj, CDOMk C SMl, Cm, CDOMn CDOM SMi, Cj+1, CDOMk+1 SMi+1, Cj+1, CDOMk+1 SMi, Cm, CDOMk+1 SMl, Cm, CDOMk+1 SMi+1, Cm, CDOMk+1 C C SMi, Cj, CDOMk+1 Hyperspectral Water Quality Smi+1, Cj, CDOMk+1 SM

Sample Comparison of Spectral Curve Fit CHL=6. 3, TSS=2. 0, CDOM=4. 8 CHL=0. 0006, TSS=3. 09, CDOM=5. 7 ASD Spectra Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Calibrating AVIRIS Images Figure 1: AVIRIS and Ground Truth Estimates for HYDROMOD Based ELM Low Signal Pixel High Signal Pixel Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

ELM Including Model correction • Assume cloud R » 0. 9 Estimate water constituents in clear water (use ground truth if available) to predict R using Hydro. Mod for the specific conditions under study • Perform Linear transform of Radiance to reflectance, L=m. R+b • NB accounts not only for atmosphere, but for any first order model-atmosphere-sensor mismatch Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

After ELM Calibration 0. 06 AMOEBA FIT 0. 04 0. 02 400 Long Pond 500 600 Wavelength Reflectance Lake Ontario 700 0. 06 0. 04 AMOEBA FIT 0. 02 400 500 600 Wavelength 700 Reflectance Cranberry Pond Reflectance Braddock Bay 0. 06 AMOEBA FIT 0. 04 0. 02 400 0. 06 0. 04 0. 02 400 500 600 Wavelength AMOEBA FIT 500 600 Wavelength 700 Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Long Pond ELM Control Point Reflectance Simulated by Hydro. Mod using Lab Measured Concentrations CHL = 62. 96 microgram/L TSS = 22. 44 milligram/L CDOM = 6. 12 scalar Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

ELM Including Model correction • Assume cloud R » 0. 9 Estimate water constituents in clear water (use ground truth if available) to predict R using Hydro. Mod for the specific conditions under study • Perform Linear transform of Radiance to reflectance, L=m. R+b • NB accounts not only for atmosphere, but for any first order model-atmosphere-sensor mismatch Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Atmospheric Compensation Improvement with Addition of Ground Truth Data Point Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Weighted Fitting ST truth data Final [CHL] [CDOM] [TSS] Weighting function MIN [(ST - SP)2 ] SQ Error [C] j [ CDOM ] LUT k [ SM ] i [CHL] [CDOM] Sp predicted [TSS] Hyperspectral Water Quality FALSE TRUE

Northwest Ponds of Rochester Embayment Lake Ontario Braddock Bay Cranberry Pond Long Pond Buck Pond Round Pond Russell Station Lake Ontario Bathymetry (feet) AVIRIS (Color Infrared) May 20, 1999 Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Hyperspectral data: to quantify multiple water quality parameters (chlorophyll, suspended solids, & yellowing organics). solar glint AVIRIS Flightlines May 20, 1999 11: 45 AM Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality Digital Imaging and Remote Sensing Labor

May 20, 1999 AVIRIS-MISI Flight AVIRIS Study Area Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Phenomenology/Ground Truth MISI underflight image of Ginna Power Plant in-water optical properties spectral measurements Reference: field support • Schott, Barsi, de Alwis, Raqueno. “Application of LANDSAT 7 to Great Lakes Water Resource Assessment, ” presented at the International Association for Great Lakes Research 43 rd Conference on Great Lakes and St. Lawrence River Research, Cornwall, Ontario, May, 2000. • Schott, Gallagher, Nordgren, Sanders, Barsi. “Radiometric calibration procedures and performance for the Modular Imaging Spectrometer Instrument (MISI). ” Proceedings of the Earth Intl. Airborne Remote Sensing Conference, ERIM, 1999. • Schott, Nordgren, Miller, Barsi. “Improved mapping of thermal bar phenomena using remote sensing, ” presented at the International Association for Great Lakes Research (IAGLR) Annual Conference, Mc. Master University, Hamilton, Hyperspectral Ontario, May 1998. Water Quality

Aviris GT Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

CHL Ground Truth Comparison RMS = 11. 6 mg/m 3 18% of [CHL] range Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

TSS Ground Truth Comparison Glint Area RMS = 4. 0 g/m 3 17. 8% of [TSS] range Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

CDOM Ground Truth Comparison Glint Area RMS = 2. 2 [scalar] 17. 2% of [CDOM] range Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Evidence of solar glint slicks AVIRIS Rochester Embayment May 20, 1999 Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Scalar Concentration of CDOM(350 nm)=5. 0 CDOM(350 nm)=1. 0 CDOM(350 nm)=0. 2 Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

CHL Model Prediction Means vs. Ground Truth Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

CDOM Model Prediction Means vs. Ground Truth Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

TSS Model Prediction Means vs. Ground Truth Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Lake Bottom at Different Spatial Resolutions AVIRIS: 20 meter pixels Rochester Embayment May 20, Water 1999 Quality Hyperspectral

Lake Bottom at Different Spatial Resolutions Region: Lake Ontario North of Irondequoit Bay AVIRIS with 20 m pixels MISI with 9 ft pixels Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Hyperspectral Imaging for Bottom Type Classification and Water Depth Determination M. S. Thesis Defense Nikole Wilson 10 Aug 2000 Hyperspectral Water Quality

Depth Varies Linearly Case 1 constant bottom Philpot’s synthetic data a|| has a parallel relationship with direction of changing depth Depth varies linearly X at 650 nm X at 550 nm Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Case 2 : Varied depth, bottom type Data form separate but parallel clusters in linearized space Clusters separated in linearized space by a distance relating to differences in bottom reflectances X at 650 nm X at 550 nm Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Data Collection Ginna Bottoms Gray rock 1 Redrock with algae Red rock Light gray rock Yellow rock Gray rock 2 Digital Imaging and Remote Sensing Laboratory Hyperspectral Water Quality

Ontario Beach Qualitative Results Depth Bottom 1 2 Rock 2 1. 6 Sand 3 2. 4 Rock 4 2. 2 Sand 4 3 2 1 Picking up different bottom type Hyperspectral Water Quality Depth

Lake Bottom at Different Spatial Resolutions Lake Ontario at Russell Station solar glint MISI with 2 ft pixels Hyperspectral Water Quality Lake Ontario at Cranberry Pond MISI with 4 ft pixels

Lake Ontario Bathymetry Hyperspectral Water Quality