Quantitative Tool for the Retrieval of Biogeophysical Parameters

Various efforts in environmental policy on a global level have improved the awareness of the need to monitor geophysical parameters with new techniques. The much referred to Kyoto Protocol to the UN Frame¬ work Convention on Climate Change (UNFCC) proposes a global policy to be applied at international level, based on assessments of carbon emission and Sequestration rates. The aim of the protocol is therefore to stabilize the C02 concentration in the atmos¬ phere in the long run. In particular, the consideration of carbon sinks in the protocol has given a large momentum to the implementation of a scientifically sound accounting and verification System. Key issues still to be resolved are the variability, uncertainty, attribution, non-permanence, leakage and future evolvement of carbon Sequestration in the terrestrial biosphere (Valentini 2000).The estimated carbon up-take of the biosphere needs to correlate with all other evidence at the three levels of Integration of the carbon budget: global, national and local.

Imaging Spectroscopy as a Quantitative Tool for the Retrieval of Biogeophysical Parameters Michael Schaepman, Klaus I. Itten, Zürich Michael Rast, Noordwijk 1 Introduction Various efforts in environmental policy on a global level have improved the awareness of the need to monitor geophysical parameters with new techniques.The much referred to Kyoto Protocol to the UN Frame¬ work Convention on Climate Change (UNFCC) proposes a global policy to be applied at international level, based on assessments of carbon emission and Sequestration rates.The aim of the protocol is therefore to stabilize the C02 concentration in the atmos¬ phere in the long run.In particular, the consideration of carbon sinks in the protocol has given a large momentum to the implementation of a scientifically sound accounting and verification System.Key issues still to be resolved are the variability, uncertainty, attribution, non-permanence, leakage and future evolvement of carbon Sequestration in the terrestrial biosphere (Valentini 2000).The estimated carbon up-take of the biosphere needs to correlate with all other evidence at the three levels of Integration of the carbon budget: global, national and local.
We presently lack the understanding and observations needed to complete the annual carbon budget at the global level.Furthermore, it is not yet possible to deter- mine the spatial distribution of carbon sinks at regional and national scales accurately enough.This Informa¬ tion however, is crucial for policy making.Recent inter¬ national research activities highlighted the need for enhanced experimental and monitoring Systems (e.g.flux measurements, satellite sensors, field and laboratory experiments, global data archives), which allow for a better parameterisation of terrestrial biosphere modeis and therefore contribute towards a better understanding of the global carbon cycle and its likely evolution (Cramer 1999).
Research into the global carbon cycle is organized by many national and international research net- works (e.g. the International Geosphere Biosphere Program (IGBP), the World Climate Research Pro¬ gram (WCRP), the Global Change and Terrestrial Ecosystems and Biological Aspects of the Hydrological Cycle projects (GCTE and BAHC), etc.).For many years, the focus of these activities has been on three main components of the Earth System: the terres¬ trial environment, oceans and atmosphere.The «Great Leap», a Joint IGBP and WCRP pilot study, focused on coupled modeis of physical and biogeochemical Sys¬ tems (land and ocean) in order to understand feedback mechanisms between carbon dioxide and climate.
It was found that changing Vegetation dynamics play a crucial role in climate-vegetation coupling (Cox et al.   2000).
One particular component of the Earth System, the terrestrial environment, has been identified as being critical for the variability of the global carbon cycle.But given the natural diversity of landscapes, the (instrumented) measurement and Validation approach remains challenging.Earth Observation from airborne or spaceborne platforms is the only observational approach capable of providing data at the relevant scales and resolution needed to extrapolate findings of in situ (field) studies to larger areas, to document the heterogeneity of the landscape at the regional scale and to connect these findings into a global view.The close coordination of Earth Observation satellites and airborne platforms is thus essential for the suc- cessful Validation of the contribution of the terrestrial component of the global carbon cycle.Space agen- cies and international organizations have recently established with IGOS-P (Integrated Global Observing Strategy Partnership) a coordination mechanism that facilitates progress in space-based measurements (ESA 2001).
The aim of this contribution is to present new Earth Observation platform strategies that enable the reproducible measurement of the radiance field of the terres¬ trial surface at a local, regional and global scale using well calibrated imaging spectrometers.The derived variables were also used to quantify important pro¬ cesses that are directly relevant for the global carbon cycle.The imaging spectrometers under discussion are APEX (Airborne Prism Experiment) and SPECTRA (Surface Processes and Ecosystem Changes Through Response Analysis), both optimised by design to solve spatial and spectral scaling issues from in situ, regional and global scales.
2 Relevant Processes and Variables The success of an approach using coupled modeis to link the Earth environment to climate variability is based on the global dynamic representation of vege- tation.The understanding of this interaction is based on a suite of modeis that describe the main processes involved on a local scale over a shorter period of time.
The key processes associated with Vegetation are photosynthesis, respiration and water transfer because they are the most important drivers of global dynamic Vege¬ tation modeis.TTiis contribution focuses on these three processes and on their monitoring by means of Earth Observation instruments.This does not mean to say that other processes are not as relevant and discussed internationally, see the UN Convention to Combat Desertification UNCCD, they are simply not dealt with here.Photosynthesis is the process by which canopies fix atmospheric carbon into biomass using energy originating from solar radiation.The rate of carbon fixed by the Vegetation depends therefore on the amount of photosyntheticalfy active radiation (PAR: integrated radiation energy within the 400-700 nm spectral ränge) absorbed by the canopy.The fraction of PAR energy absorbed (fAPAR) depends mainly on the canopy architecture and secondarily on the optical properties of the elements (leaves, stems, trunks, branches, background).The canopy architecture is primarily described using the leaf area index (LAI, leaf area per unit area of soil).Photosynthesis also depends on a large number of other factors, such as carbon concentration in the atmosphere, temperature, mineral deficiencies and nitrogen deficiency.Nitrogen content at leaf level is strongly related to Chlorophyll content.
Respiration occurs both within the Vegetation (autotrophic) and within the soil (heterotrophic).The soil carbon reservoir is large when compared to the above ground biomass.The understanding of the fluxes of carbon to and from this reservoir (i.e. through litter decomposition and respiration, respectively) becomes a major issue when attempting to calculate the carbon balance.The fraction of dead biomass corresponds to a pool of carbon that can rapidly feed the soil reservoir.Heterotrophic respiration is very dependant on tem¬ perature, moisture and availability of nutrients, par- ticularly nitrogen.The nitrogen cycle is therefore intimately linked to the carbon cycle within the soil via the biotic activity.
Water plays a critical role in Vegetation canopies, but the soil constitutes the main reservoir of water.Evaporation and transpiration are combined into evapotran¬ spiration to describe the water flux exiting the sur¬ face.The latent heat flux necessary to transform the plant and soil liquid water into vapor in the atmos¬ phere closely links evapotranspiration to the radiation energy balance.The energy balance itself is mainly driven by the available incident energy Coming from the sun and the sky.Part of this incident energy is reflected towards the sky depending on the albedo of the surface for the shorter wavelengths (fraction of reflected solar energy in the 300-3000 nm spectral domain), and on the emissivity for the longer wave¬ lengths.The remaining energy is either stored in the soil (soil heat flux), re-emitted in the longer wave¬ lengths (thermal infrared domain) according to the surface temperature and emissivity, or dissipated by convection (sensible heat flux) and evapotranspira¬ tion (latent heat flux).The latter two processes are determined by gradients of temperature (sensible flux) or moisture (latent) and by the transfer resistances inherent in the leaf area index and Vegetation type (Menenti & Ritchie 1994).
Finally, many of the processes considered here are influenced by the type of ecosystem or biomes.Ecosystems are built by a collection of species that are organised according to sophisticated rules.Ecosystems tend to adapt themselves to the local pedo-climatic conditions and develop specific strategies for surviving, always submitted to the competition between a large number of species.
From this brief review of the main processes to be found in Vegetation, it is apparent that several key var¬ iables describing canopy structure as well as the biophysical and biochemical composition of the elements either directly govern the processes or are closely linked to them.Therefore, the Earth Observation of these variables should yield pertinent Information on Vegetation State and condition, thereby contributing towards the description and understanding of dynamic Vegetation processes.
3 Interaction of the radiation field with Vegetation The interaction of radiation with canopies and soils depends on the optical thermal or dielectric proper¬ ties of the materials, as well as on their number, area, orientation and position in space, i.e. the primary biophysical variables of the materials.Therefore, remote sensing allows the direct derivation of canopy or soil primary biophysical variables.In addition, secondary variables which are combinations of primary biophysi¬ cal variables, can also be estimated.The main primary and secondary variables accessible by remote sensing in the solar domain are listed in Table 1.
The information contained in the wavelength ränge of the solar reflected electromagnetic spectrum (400-2500 nm) provides estimates of canopy structural variables, such as LAI and the main elements of biochemical composition (Chlorophyll and water con¬ tent).In addition, Fourty & Baret (1997)  Retrieval potential of biophysical variables as a function of the spectral domain used.The accuracy requirements are derived from measurements, and/or model sensitivity analysis.The level of accuracy and robustness of the estimation is indicated by the «+» («++++» for accurate and robust;«-» indicates no estimates possible).Herleitungspotential von biophysikalischen Variablen als Funktion ihrer spektralen Eigenschaften.Die angege¬ benen Genauigkeitsanforderungen für die Prozessbeschreibungen sind von Messungen oder Sensitivitätsanalysen hergeleitet worden.(Die Genauigkeit und Robustheit der Schätzung ist mit «+» und «-» angegeben («++++» sehr genau und robust; «-» keine Schätzung möglich)).Potentiel de recuperation de variables biophysiques en fonction de la gamme spectrale employee.Les conditions d'exactitude sont derivees de mesures, et/ou de Tanalyse de la sensibilite du modele (Texactitude et la stabilite de Tevaluation sont indiquees par «+» et «-» («++++» tres precis et robuste;«-» aucune evaluation possible)).Source: Baret 2001 area) were generally more easily accessible than the leaf Contents (mass of biochemical constituents per unit leaf area).The canopy-integrated content cor- responds to the product of leaf biochemical content and leaf area index.
By summarizing the key variables that can be derived using imaging spectrometer based Earth Observation techniques, a new table was compiled (Table 2).The variables are compared with their retrieval uncertainty using «conventional» satellite Systems (e.g., SPOT HRV, NOAA AVHRR, Landsat ETM+) with two spectral bands and one view angle, and current highresolution spectro-directional sensors (e.g., MERIS, MODIS, MISR, SPECTRA).

Estimation of canopy biophysical variables
The increasing use of imaging spectrometer data is gradually enabling more insight into the physical proc¬ esses involved in the radiative transfer.The classical use of empirical relationships between reflectance observed by the sensors in the spectral, directional, temporal and spatial domains may be seen in simple spectral indices, such as the NDVI (Normalized Differ¬ ence Vegetation Index).Nowadays these techniques are increasingly making use of radiative transfer modeis (Sellers 1985), and more sophisticated meth¬ ods, such as inverse techniques (Privette et al. 1996)  and assimilation techniques (Asner et al. 1998), are beoming more common.Retrieval error for key properties of terrestrial Vegetation, based on results of direct and inverse radia¬ tive transfer modeling («not feasible» indicates that the accuracy of the retrieval is not sufficient).
Source: European Space Agency (ESA) 2001 4.1 Models of canopy reflectance Radiative transfer modeis differ mainly by the way the canopy structure is described and the way the radiative transfer equation is approximated (Myneni & Asrar   1993).Some of the modeis are more generalistic and others reflect reality closer, i.e. they are detailed but slow.A combination of the two characteristics appears to not be possible due to the heterogeneous distribution of canopy structure over the Earth's surface.Radiative transfer modeis of canopies exist today in a broad variety and the physical based modeis usually describe continuous scatterers in turbid mediums (cf.SAIL (Verhoef 1984)).The decision of which radiative transfer model to use is therefore a critical point in the process of minimising uncertainties in the subsequent inversion and/or assimilation process.

Inversion techniques
The Interpretation of imaging spectrometer data with the goal to quantify the spatial and temporal variability of canopy characteristics, whether within or between canopies, is basically an inverse problem.Radiative transfer model inversion provides a way to exploit our knowledge of these physical processes.The physi¬ cal processes are represented by a radiative transfer model (RT), which can simulate remote sensing data from canopy characteristics.Inversion techniques are designed to retrieve those canopy characteristics that are of interest from imaging spectrometer data.All techniques of Interpretation of biogeophysical or biogeochemical products are inversion or assimilation Problems requiring the identification of the most rel¬ evant measurements from the limited number availa¬ ble (Pinty et al 2000a, b).Generally, all of these tech¬ niques suffer from having more than one Solution.
Greater measurement accuracy contributes to better constraints on the inversion problem, reducing the number of potential Solutions.By reducing uncertain¬ ties to a level beyond those of the transformation algo- rithms, imaging spectrometer measurements with many contiguous spectral bands will enable decisions on rela¬ tive Performance and allow quantifiable improvements.We will therefore discuss this particular Instrumenta¬ tion approach in the following section.

Instrumentation Approach
Growing expectations on the use of Earth Observa¬ tion data to support key decisions by governments and industries puts increasing pressure on technology to deliver proven and reliable information.On the one hand, the measurements should be able to document small changes in key terrestrial parameters over many years, and on the other hand, they should be in a form appropriate for pointing to global variability in climate change.Characterisation of these sensor Systems is therefore critical, particularly where the Systems oper- ate in the solar reflected radiation from the Earth's surface (cf.Teillet et al. 1999).
Imaging spectroscopy today enables the geobiophysical and geobiochemical variables of the Continental biosphere processes to be measured with quantifiable accuracy.However, surface characteristics critically govern the involved processes.If we want to take proper actions to mitigate these undesirable effects, surface characteristics must be described accurately to better understand, model and predict the interac¬ tion between surface and atmosphere.The complex global modeis that address these issues use variables and parameters to describe the biophysical and bio¬ chemical processes but, when compounded, the uncertainties in these variables and parameters can only produce moderate to poorly accurate results.Amelioration comes from independent information applied to control or constrain the modeis, but the spatially distributed character of land surface processes -large spatial heterogeneity and dynamic character -require remote-sensing observations to provide this informa¬ tion.ESA's (European Space Agency) Living Planet, NASA's (National Aeronautics and Space Administra¬ tion) EOS programmes and the GMES (Global Moni¬ toring of Environment and Security) initiative recog- nise this, but there are some remaining uncertainties due to the inherent uncertainty of the measurement.

Promise
Imaging spectroscopy -the quantitative and qual¬ itative characterization of both a surface and the atmosphere using geometrically coherent spectro-radiometric measurements -measures the continuous, upwelling spectral radiance in each spatial pixel.The results can be used for: 1. unambiguous direct and indirect identification of surface materials and atmospheric trace gases, 2. measurement of their relative concentrations, 3. assignment of the proportional contribution of mixed pixel Signals (spectral un-mixing problem), 4. derivation of their spatial distribution (mapping problem), and 5. study over time (multi-temporal analysis).
Although biases between sensors, even of the same design like NOAA AVHRR, may limit the detection of subtle temporal changes, «conventional multispectral» remote sensing usually Covers aspects 4-5 with sufficient accuracy.New instruments will cover aspects 1-3, but conventional calibration techniques will still restrict uncertainties to a few percent at bestthereby being only marginally adequate for cur- rent user demands.Even so, imaging spectroscopy has already supported or been used for the following appli¬ cations and the corresponding variables have been directly or indirectly quantified using inversion tech¬ niques (Green et al. 1998 In addition to these applications, Table 3 lists those varia¬ bles that have been derived using imaging spectrometers. As imaging spectroscopy matures, governments, agencies and industry see potential in the utilization of the data and demands for higher spectral and spatial resolution become greater.Climate change, a field that requires long-time base measurements for small changes in specific key parameters to be detected, is an example of one of those fields that rely on higher spec¬ tral resolution to be meaningful.TTie «near-fingerprinting sensors» or «ultraspectral sensors», yet to become fully operational, are an example of the improvements being made in the area of spatial resolution.Preliminary results using Fourier Transform spectro-meters (Yarbrough et al. 2001) demonstrate the feasibility of these sensors to detect fluorescence phenomena in Veg¬ etation, which usually occur at spectral bandwidths well below the assumed atmospherically induced effects that exist around 10 nm bandwidth in the 400-700 nm ränge of the electromagnetic spectrum. 5The Imaging Spectrometer APEX Two major Instrumentation approaches are currently being developed, enabling the assessment of quantifi¬ able measurements in the near future.On the local, regional and national scale this is the imaging spectrome¬ ter APEX, and on the regional to global scale, the instru¬ ment under consideration is SPECTRA (ESA 2001).
Within the framework of the European Space Agen-  4'000 -lO'OOO m.a.s.l. (7'500 m Standard) Spectral Channels VNIR: approx. 140; SWIR: approx. 145  Spectral ränge 400 -2500 nm Spectral sampling interval 400 -1050 nm: < 5 nm 1050 -2500 nm: < 10 nm Spectral sampling width < 1.5 * Spectral sampling interval  local, regional and national campaign oriented data acquisition for the quantification of important proc¬ esses that are directly relevant for the global carbon cycle, and further, for fostering the use of spectroscopy throughout the scientific Community.SPECTRA, as opposed to APEX, is a satellite based spectro-direc- tional imager, that will be able to sample a number of test sites world wide within three days at seven differ¬ ent viewing angles and up to 60 programmable spec¬ tral bands.In addition to the APEX approach, SPEC¬ TRA also makes use of directional thermal infrared (10.3 -12.3 pm).
Technically, APEX is designed to be a pushbroom imager with approx. 300spectral Channels in the 400 - 2500 nm wavelength region, having 1000 pixels across track and a swath width of 2.5 -5 km depending on flight altitude.The APEX hardware consists of an airborne imaging spectrometer with an optimised spectrometer sensor designed for the detection of land surface proc¬ esses, a flexible aircraft Integration scheme, an internal calibration facility, a laboratory calibration home base, and a Processing and Archiving Facility (PAF) for the generation of calibrated radiance data.A summary of all relevant speeifications for APEX is listed in the Table 4, and its schematic design is illustrated in Figure 1.
The APEX instrument is designed to cover most of the relevant land applications.However, before APEX can do this, a detailed scientific analysis has to be per¬ formed and the requirements have to be calculated in terms of SNR (Schaepman et al. 2001b; Schläpfer   & Schaepman 2002).Figure 2 lists the dynamic ränge (minimum and maximum radiance) for the applica¬ tions investigated.Figure 3 shows the SNR require¬ ments derived for these applications.
Diagrammes schematiques d'APEX: Vue de face avec trois niveaux: banc optique (fond), composants electroni- ques (milieu), unite de calibrage (dessus) (gauche); Talignement de cloison et projection du faisceau (milieu); con- ception optique en deux prismes (droite).Figure 3: Calculated Signal to Noise Ratios (SNR) for APEX using a forward spectral model Berechnetes Signal-Rausch-Verhalten für AP EX basierend auf einem spektralen Vorwärtsmodell Rapports signal-bruit calcules pour VAPEX d'apres un modele spectral direct Source: Schläpfer & Schaepman 2002   is scheduled for 2005.The operational phase of APEX is foreseen for a duration of 5 years from VITO in Bel- gium and should be completed in time for the planned launching of the SPECTRA Earth Explorer Core Mis¬ sion in 2008.The industrial team responsible for the construction of APEX is composed of the Swiss compa¬ nies HTS AG (industrial prime, System integration) and Netcetera AG (mission Software, electronics), the Belgian Company OIP Sensor Systems (spectrometer), the French Company Sofradir (detectors), and the German Company DLR (calibration home base, airworthiness certification).ESA is responsible for the technical project management and contractual issues.The scien¬ tific team and the Principal Investigator are situated at the University of Zürich, Switzerland.
Summary: Imaging Spectroscopy as a Quantitative Tool for the Retrieval of Biogeophysical Parameters Representing the dynamic character of Vegetation in global climate modelling is a challenging task.We present a new Instrumentation approach in the form of APEX, an airborne imaging spectrometer to address the key variables and processes relevant for monitoring the biogeophysical and biochemical parameters at the local, regional and national scale.We are able to demonstrate that a large number of relevant proc¬ esses and variables have been monitored successfully using imaging spectrometers.In particular, the accu¬ racy of the variable retrieval using this observational approach has much improved and the quantification of the interaction of the radiation field with Vegeta¬ tion has been made possible.The airborne instrument APEX will be made available to the scientific Commu¬ nity in early 2005, whereas it's space successor SPEC¬ TRA is scheduled for launch in 2008.The inclusion of the spectro-directional and thermal components will contribute toward minimizing retrieval uncertain¬ ties.Until these instruments are introduced into the market, there is enough time to develop the necessary products and to inform the scientific Community about the new possibilities.
Teaching of Geography -pertinent questions -Which are the six central issues that have to be dealt with in order to solve the C02 problem? -Which three key processes are connected to the monitoring of Vegetation?What are those variables called that can be monitored by remote sensing?-Which five central issues does imaging spectroscopy guarantee to solve?Which of these five could todate not be solved using classical remote sensing?-What is the reason for retrieving hydrocarbon values?-What are the advatages expected of the deployment of the new APEX spectrometer and when is the mis-

Figure 2 :
Figure 2: Minimum and maximum radiance levels for six land applications in the solar reflected wavelength ränge Minimale und maximale Strahlung für sechs terrestrische Anwendungen im solar reflektiven Bereich des Spek¬ trums Minima et maxima de radiance pour six applications terrestres dans la gamme de longueurs d'onde refletees par le soleil Source: Schaepman et al. 2001b

Table 1 .
dem-  onstrated, that canopy integrated biochemical Con¬ tents (mass of biochemical constituents per unit soil Biophysical variables

Table 2 .
VariableSampling of angular and spectral dimensions 1 view angle, 2 spectral bands