• No se han encontrado resultados

2. LA UNIVERSIDAD ESPAÑOLA

2.1 Una institución en búsqueda de identidad

2.1.2 El convulso siglo XIX

To demonstrate the feasibility of a mission, modelling tools are necessary. Two ways are developed depending on the scene type given as inputs: either a synthetic aquatic ecosystem with an adjacent landscape or a scene acquired from an airborne or spaceborne hyperspectral sensor. The resulting tools are described below.

Several end to end simulators already exist like the EnMAP simulator or the Comanche-Cochise one (Miesch et al., 2005) used for the SPECTRA (Schaepmann et al., 2004) and HYPXIM (Briottet et al., 2011) hyperspectral missions. A general flowchart of such a simulator is proposed in Figure 3.4. As inputs, such a simulator has to consider inputs three types of a water and scenarios of various complexities:

• Hyperspectral airborne or spaceborne acquisition expressed in radiance unit. To have a good estimation of the ranges of upwelling environment radiances, he at sensor radiance needs to be atmospherically corrected to retrieve the surface reflectance Rrs, and then this is used to simulate the corresponding signal at TOA sensor level. Such a procedure allows simulation of multiple environmental water surface state and atmospheric conditions. • Spectral reflectances from existing data bases such as SEABASS or MERMAID for coastal

waters and oceans or LIMNADES for lakes. A dedicated macrophytes, seagrass, macro-algae, coral reef and associated benthic cover types does not yet exist but is advised ( see e.g. Dekker, 2006).

• Synthetic spectral reflectance simulated from radiative transfer tools like Hydrolight (Mobley) or WASI (used for the simulations in chapter 2 and the chapter 2 Appendix. Such tools can simulate the bottom of atmosphere spectral reflectance from an accurate description of the aquatic ecosystem and its corresponding geo-physical variables.

83

We note that for heterogeneous ecosystems like coastal zone or inlands water, the radiative contribution of the environment is crucial The Comanche-MODTRAN tool uses a Mont Carlo kernel to deliver a good estimation of the Earth atmosphere coupling irradiance and the environment upwelling radiance. Although the exo-atmospheric solar irradiance is not polarized, the photons crossing the atmosphere and reflected by the Earth are polarized; it is possible to estimate the polarized TOA radiance by using MODTRAN-P radiative tool or OSOAA (Chami et al., 2015). Further, different atmospheres, aerosols (type and abundance) and acquisitions geometry can be considered using e.g. Multi-Angle Implementation of Atmospheric Correction (MAIAC) products.

Thus, the top of atmosphere (TOA) radiance is estimated from these inputs and a radiative transfer code such as MODTRAN, DISORT or three atmospheric correction codes used by Martin et al., (2017) for Brazilian rivers and floodplains, or Second Simulation of a Satellite Signal in the Solar Spectrum (6SV), ACOLITE and Sen2Cor with the highest spectral resolution. Note that the 6SV code is not suited for hyperspectral simulations, its spectral resolution (2.5 nm) is not sufficient to model the TOA signal for 5 to 8 nm bandwidth. To estimate the resulting signal collected by the hyperspectral camera at the output of the electronic chain (expressed in digital count units) several contributions have to be taken into account such as:

• Spatial module: to simulate the corresponding spatial resolution of the sensor and taking into account its Modulation Transfer Function (MTF) able to simulate various spatial aberrations caused by, e.g. the telescope optics, the double slit and the curved prisms (e.g. EnMap simulator).

• Spectral module: to perform the spectral resampling, taking into account the spectral response functions in all spectral bands, non-uniformities in the spectral domain such as smile, polarization, and an optional spectrometer shift in the spectral dimension (e. g. EnMap)

• Electronic module: to convert the radiance in digital count by taking into account the artefact of the electronic detection chain (noises, temporal shift…)

The resulting digital output signal needs several pre-processing steps to overcome the deterministic artefacts introduced by the instrument itself.

From a radiometric and geometric modelling of the camera, a pre-processing is necessary which depend on its design such as dark current correction, inter-detector sensitivity correction, smile etc. The resulting signal is then converted into output radiance taking into accounts the calibration accuracy. From the output radiance, atmospheric compensation has to be achieved to retrieve the bottom surface reflectance as follows: estimation of the atmospheric state: using the image itself (estimation of the water vapour content, aerosol type and abundance…) or from externally obtained data (atmospheric profiles, AERONET and AERONET-OC data, RAdCalNet data and similar data sources). Using this information, the TOA signal is corrected to retrieve the bottom of atmosphere (BOA) surface reflectance Rrs. The resulting BOA reflectance image can be georeferenced or not. At this level, biophysical parameters can be retrieved following the approach outlined in 4.1.

84 Figure 3.4 General flowchart of an End-to-End simulator

Airborne Acquisition Flight Level At-Sensor Incident Radiance TOA

Level Preprocessing - At Sensor Output

Reference Ground Reflectance Ground

Level

Forward Radiative Transfer

Atmospheric correction

Instrument modeling: spatial and spectral aggregations, noise, MTF, smile, keystone…

Atmospheric correction

Retrieved Ground Reflectance

Dedicated reflectance / Biophysical variables RT

simulator:

• Soil, vegetation: DART, SAIL

Measured spectral reflectance from data base like USGS,

MEMOIRES…

Georeferencing

Biophysical variables retrieval: • Soil, vegetation: Sail… • Water: BOMBER, Lee,

85

4

Aquatic ecosystem earth observation enabling

activities

S

TEEF

P

ETERS

,K

EVIN

R.T

URPIE

,S

INDY

S

TERCKX

,P

ETER

G

EGE

,X

AVIER

B

RIOTTET

,M

ARTIN

B

ERGERON

,

N

ICOLE

P

INNEL

,A

RNOLD

G.D

EKKER

,C

LAUDIA

G

IARDINO

,V

ITTORIO

E.B

RANDO AND

B

RINGFRIED

P

FLUG

.