be developed. Only fully accomplishing these two steps will push forward the
utilisation of satellite soil moisture in hydrological modelling to a greater extent.
Keywords Flood forecasting, Hydrological modelling, Review, Satellite remote
sensing, Soil moisture, Soil moisture and ocean salinity (SMOS)
1 Introduction
Although soil moisture comprises only 0.01% of the total amount of water on the
Earth [1], the existence of soil moisture is significant for many application areas such
as agriculture, meteorology, climate investigations and natural hazards predictions.
In hydrology, soil moisture is a significant state variable in real-time flood forecasting [2]. Over the past decades, numerous hydrological models have been developed,
representing more or less accurately the main hydrological processes involved at
a catchment scale [3]. The challenge in forecasting floods in a reliable way stems
mainly from the error accumulation of the models, particularly during unusual
hydrological events and after a long period of dryness. Solutions have thus been
introduced to enhance flood forecasting by matching the model with the current
observations prior to its use in forecasting mode – termed as updating or data
assimilation [4]. Since hydrological models are highly sensitive to the state change
of the soil moisture [3], a better soil moisture observation over a catchment should
improve the forecasting performance via correcting the trajectory of the model
[5]. Nevertheless it is very challenging to accurately monitor soil moisture that
varies both spatially and temporally. Conventional in situ networks are expensive
and impractical in large areas, and they are still too sparse to represent the spatial soil
moisture distribution [6–11]. Model-based estimates such as those from land surface
models (LSMs) are another source of soil moisture data, but they are uncertain due
to imperfect parameterisation, meteorological forcing data and time drift problems
(e.g. accumulation of errors) [12–14].
Alternatively, modern satellite remote sensing has shown potential for providing
soil moisture measurements at a large scale [15]. However, satellite soil moisture
products are calibrated mainly by in situ measurements, so they are not directly
relevant to hydrology [8, 16, 17]. Moreover with all orbiting sensors, only the
surface layer soil moisture can be acquired. It has been shown in many studies that
the soil penetration depth is around 0.1–0.2 times the sensor wavelength, where the
longest wavelength is only about 21 cm (L-band, with a penetrating depth ~5 cm)
[18, 19]. Conversely operational hydrological models (most often the conceptual
hydrological models) consider a much deeper surface soil depth (up to 2 m), which
also varies across a catchment.
Clearly there is a mismatch between the satellite-retrieved soil moisture and
the hydrological model-simulated soil moisture, which has caused a commensurate
issue for the full utilisation of remotely sensed soil moisture products in operational
hydrology. Although many studies have been carried out on the evaluation of
260
L. Zhuo
utilisation of satellite soil moisture in hydrological modelling to a greater extent.
Keywords Flood forecasting, Hydrological modelling, Review, Satellite remote
sensing, Soil moisture, Soil moisture and ocean salinity (SMOS)
1 Introduction
Although soil moisture comprises only 0.01% of the total amount of water on the
Earth [1], the existence of soil moisture is significant for many application areas such
as agriculture, meteorology, climate investigations and natural hazards predictions.
In hydrology, soil moisture is a significant state variable in real-time flood forecasting [2]. Over the past decades, numerous hydrological models have been developed,
representing more or less accurately the main hydrological processes involved at
a catchment scale [3]. The challenge in forecasting floods in a reliable way stems
mainly from the error accumulation of the models, particularly during unusual
hydrological events and after a long period of dryness. Solutions have thus been
introduced to enhance flood forecasting by matching the model with the current
observations prior to its use in forecasting mode – termed as updating or data
assimilation [4]. Since hydrological models are highly sensitive to the state change
of the soil moisture [3], a better soil moisture observation over a catchment should
improve the forecasting performance via correcting the trajectory of the model
[5]. Nevertheless it is very challenging to accurately monitor soil moisture that
varies both spatially and temporally. Conventional in situ networks are expensive
and impractical in large areas, and they are still too sparse to represent the spatial soil
moisture distribution [6–11]. Model-based estimates such as those from land surface
models (LSMs) are another source of soil moisture data, but they are uncertain due
to imperfect parameterisation, meteorological forcing data and time drift problems
(e.g. accumulation of errors) [12–14].
Alternatively, modern satellite remote sensing has shown potential for providing
soil moisture measurements at a large scale [15]. However, satellite soil moisture
products are calibrated mainly by in situ measurements, so they are not directly
relevant to hydrology [8, 16, 17]. Moreover with all orbiting sensors, only the
surface layer soil moisture can be acquired. It has been shown in many studies that
the soil penetration depth is around 0.1–0.2 times the sensor wavelength, where the
longest wavelength is only about 21 cm (L-band, with a penetrating depth ~5 cm)
[18, 19]. Conversely operational hydrological models (most often the conceptual
hydrological models) consider a much deeper surface soil depth (up to 2 m), which
also varies across a catchment.
Clearly there is a mismatch between the satellite-retrieved soil moisture and
the hydrological model-simulated soil moisture, which has caused a commensurate
issue for the full utilisation of remotely sensed soil moisture products in operational
hydrology. Although many studies have been carried out on the evaluation of
260
L. Zhuo