994
and momentum between the sub-system of interest and its environment, and (ii) to acquire a fundamental
understanding of the physical processes that condition the signal measured in space, to assess the feasibility,
accuracy, and reliability of the remote sensing data and their interpretations. Both objectives will lead to a
better understanding of the functioning of the Earth system, and provide a basis for predicting its evolution.
In this paper, we propose a conceptual framework to discuss whether these two objectives can be at
tained, what is required to achieve them, and the implications of pursuing one or another approach. We hope
that this discussion may shed some light on the setting of priorities for resource allocation in future model and
instrument development. The following sections are framed in the context of the observation of terrestrial
environments using remote sensing methods in the solar spectral range, but most of the arguments could
be extended to other environments ( e.g ., the oceans) or spectral ranges (thermal infrared or microwaves).
Similarly, we will discuss only the measurement of radiant intensities, but additional radiative information,
such as the polarization and phase of the electromagnetic waves, could be considered as additional dimen
sions of the data. In the concluding section, we will review the extent to which various modeling approaches
are able to achieve the two objectives stated above: characterisation of the system under observation and
understanding of the radiative processes that control the measurements.
In order to provide a clear and objective framework to address these issues, we have attempted to express
the principal arguments in a formal way. The theorems that follow are not intended to support a dogmatic
view of remote sensing, but, rather, to support as clearly as possible a logic to the use and development
of these techniques. We hope that these statements will stimulate vigorous discussions, result in a greater
understanding of the opportunities and limitations of this technology, and generate ever better research and
applications.
UNDERSTANDING THE PHYSICS OF THE MEASURED SIGNALS
Most of the users of remote sensing data are ultimately interested in some form of high level integrated
information, such as the distribution of biomass, the fluxes of water, energy and carbon between terrestrial
environments and the atmosphere, etc. These variables of interest will be denoted Y. On the other hand,
because satellite platforms evolve essentially outside our atmosphere, they can only record gravitational
and electromagnetic events. For the purpose of this paper, we will only consider radiation (or reflectance)
measurements in the solar spectral range, and the corresponding data will be denoted Z (bold symbol*
refer to vectors or matrices of numbers). The recourse to remote sensing techniques therefore hinges on the
following theorem:
Theorem 1. Radiative data Z collected on board remote sensing platforms in space can be interpreted
quantitatively in terms of the variables of interest Y.
The primary goal of the remote sensing approach is therefore to establish this quantitative relation,
symbolically depicted in Figure 1, and the extent to which this is actually done will hopefully become dear
by the end of this paper.
I’
Y
Figure 1: Graphical representation of the initial goal of remote sensing.
In fact, it will be seen that there is more than one possible approach to estimate the variables of interest Y on
the basis of the measurements Z. In this section, we focus exclusively on the development of physically-bawd
models. These models are based explicitly on the theory of radiation transfer; they aim at explaining the
mechanisms that affect the measurements and, in many cases, can be used to characterise the state of the
system under observation.
The fundamental reason why this theorem is not obviously true is that, in general, measurement« l
in space are not controlled exclusively or directly by the variables of interest Y at the surface, but, rather,
