accounting for different inaccuracies of observations coming from physical and
social sensors and is subject for further research.
For static social sensors, α values are higher than for static physical sensors and
are considered to be a random stochastic variable uniformly distributed in time and
space. More details can be found in Mazzoleni et al. [61].
5 Assimilation of Flow Observations from Static
Heterogeneous Sensors
This section aims to explore the benefits of assimilating flow observations from a
network of static heterogeneous sensors in the case of synchronous (Sect. 5.1) or
asynchronous (Sect. 5.2) social observations, depending on the predictability of the
arrival time of the observations. In particular, we assume that social sensors provide
intermittent observations that can lie either in a specific model time step (synchronous) or in between two model time steps (asynchronous). In addition, social sensors
may be distributed within the catchment or be located in a specific point.
5.1 Assimilation of Synchronous Observations
Here, we show the model performance after the assimilation of intermittent synchronous observations, i.e. their arrival time matches the model time step, within the
semi-distributed hydrological models of the Brue catchment. We can divide this
section in two parts: first, the flow observations are assimilated from different social
sensors located within the catchment; second, social sensors are integrated with a
network of physical sensors to evaluate the added value of crowdsourced sensors in
the assimilation process. A straightforward and pragmatic method (based on EnKF)
is used to assimilate the intermittent observations into the hydrological model
updating the model states matrix only when observations are available, while
when there are no observations, it is assumed that the state covariance error does
not change at that time step [60, 96].
5.1.1 Assimilation of Flow Observations Only from Social Sensors
In the first part of this section, we considered MS1 and three different spatial
configurations (SC) of static social sensors within the catchments (called scenarios
SC1, SC2 and SC3). In particular, SC1 refers to social sensors located along the
main river channel, SC2 to sensors located on the upstream part of the main river
channel, while SC3 to sensors located close to the catchment outlet. Model results
obtaining assimilating flow observations from physical sensors are considered as
220
M. Mazzoleni et al.
social sensors and is subject for further research.
For static social sensors, α values are higher than for static physical sensors and
are considered to be a random stochastic variable uniformly distributed in time and
space. More details can be found in Mazzoleni et al. [61].
5 Assimilation of Flow Observations from Static
Heterogeneous Sensors
This section aims to explore the benefits of assimilating flow observations from a
network of static heterogeneous sensors in the case of synchronous (Sect. 5.1) or
asynchronous (Sect. 5.2) social observations, depending on the predictability of the
arrival time of the observations. In particular, we assume that social sensors provide
intermittent observations that can lie either in a specific model time step (synchronous) or in between two model time steps (asynchronous). In addition, social sensors
may be distributed within the catchment or be located in a specific point.
5.1 Assimilation of Synchronous Observations
Here, we show the model performance after the assimilation of intermittent synchronous observations, i.e. their arrival time matches the model time step, within the
semi-distributed hydrological models of the Brue catchment. We can divide this
section in two parts: first, the flow observations are assimilated from different social
sensors located within the catchment; second, social sensors are integrated with a
network of physical sensors to evaluate the added value of crowdsourced sensors in
the assimilation process. A straightforward and pragmatic method (based on EnKF)
is used to assimilate the intermittent observations into the hydrological model
updating the model states matrix only when observations are available, while
when there are no observations, it is assumed that the state covariance error does
not change at that time step [60, 96].
5.1.1 Assimilation of Flow Observations Only from Social Sensors
In the first part of this section, we considered MS1 and three different spatial
configurations (SC) of static social sensors within the catchments (called scenarios
SC1, SC2 and SC3). In particular, SC1 refers to social sensors located along the
main river channel, SC2 to sensors located on the upstream part of the main river
channel, while SC3 to sensors located close to the catchment outlet. Model results
obtaining assimilating flow observations from physical sensors are considered as
220
M. Mazzoleni et al.