170
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
arid and mountainous regions, where conventional observations are not available at a
sufficient spatial resolution (Mcculloch 2007). Consequently, hydrologic predictions
in response to climate change impacts are difficult to evaluate (Abdulla et al. 2009).
Both observational studies and modeling outputs have suggested changes in
extreme events for future climates significantly in China (Easterling et al. 2000; Sun
et al. 2010). Adaptation to climate variability and change is important both for impact
assessment and for policy development (Smith et al. 2000), especially in some arid
regions due to the vulnerability in adaptation (Kelly and Adger 2000). Facing this
challenge, the International Association of Hydrological Sciences (IAHS) has initiated a plan on predictions in ungauged basins (PUB) to promote hydrologic practices
(Sivapalan et al. 2003; Wagener et al. 2004). According to the “IAHS decade on
predictions in ungauged basins, 2003–2012” (Sivapalan et al. 2003), there is a need
to develop evaluation methods that can operate anywhere, independent of watershed
borders and gauges. At the same time, the increasing weather and climate extremes
call for some simple, pragmatic approaches to estimate runoff in poorly gauged
watersheds. For most cases, the runoff at the watershed scale is controlled by meteorological processes, particularly at a monthly scale; therefore, an approach using
an artificial intelligence model with the aid of remote sensing data may become an
indispensable means to estimate runoff in a poorly gauged watershed.
Remote sensing and image processing can help retrieve spatiotemporal features
from past decades, which is the information that cannot be collected by point measurements at the ground level (Wagner et al. 2003). The aims of such applications
are to (1) measure spatial, spectral, and temporal information and (2) provide data
on the state of the earth’s surface. Previous studies suggested that remotely sensed
data should provide major benefits to hydrologic system analysis and water resources
management; yet, application potential with practical benefits was limited by modeling skill. One reason for this barrier is the lack of mathematical tools to convert
remotely sensed data to the type of information useful to seamlessly fit into the
actual needs in water resource systems (Kite and Pietroniro 1996). Some statistical models played an important role in translating data to information (Chang et al.
2009); however, limited by the inherent modeling structure and the semiempirical
nature, the breakthroughs in applications using statistical models were not salient.
Vastly improved instrumentation for sensing, logging, and transmitting hydrometric
measurements already facilitates retrieval of information from the far corners of
the earth’s surface (Mcculloch 2007). As we move from a data-poor to a data-rich
era due to the advances of remote sensing technologies, knowledge discovery and
management via a plethora of data mining and machine learning techniques become
promising, facing massive data sets (Harvey and Jiawei 2001).
Parameterization of hydrometeorological processes by advanced artificial intelligence methods has been widely used in modern hydrology (Storch and Zwiers 1999).
The empirical orthogonal function (EOF) is one of the decomposition procedures
used to investigate the spatiotemporal behavior of hydrometeorological processes;
thus, the main variance characters of meteorology are expected to be retrieved by
this technique (Puebla et al. 1998). Increasingly, models featured with artificial neural networks (ANNs) have been used in a wide range of disciplinary fields and have
become practical tools in hydrologic analyses for predictions. In many applications,
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
arid and mountainous regions, where conventional observations are not available at a
sufficient spatial resolution (Mcculloch 2007). Consequently, hydrologic predictions
in response to climate change impacts are difficult to evaluate (Abdulla et al. 2009).
Both observational studies and modeling outputs have suggested changes in
extreme events for future climates significantly in China (Easterling et al. 2000; Sun
et al. 2010). Adaptation to climate variability and change is important both for impact
assessment and for policy development (Smith et al. 2000), especially in some arid
regions due to the vulnerability in adaptation (Kelly and Adger 2000). Facing this
challenge, the International Association of Hydrological Sciences (IAHS) has initiated a plan on predictions in ungauged basins (PUB) to promote hydrologic practices
(Sivapalan et al. 2003; Wagener et al. 2004). According to the “IAHS decade on
predictions in ungauged basins, 2003–2012” (Sivapalan et al. 2003), there is a need
to develop evaluation methods that can operate anywhere, independent of watershed
borders and gauges. At the same time, the increasing weather and climate extremes
call for some simple, pragmatic approaches to estimate runoff in poorly gauged
watersheds. For most cases, the runoff at the watershed scale is controlled by meteorological processes, particularly at a monthly scale; therefore, an approach using
an artificial intelligence model with the aid of remote sensing data may become an
indispensable means to estimate runoff in a poorly gauged watershed.
Remote sensing and image processing can help retrieve spatiotemporal features
from past decades, which is the information that cannot be collected by point measurements at the ground level (Wagner et al. 2003). The aims of such applications
are to (1) measure spatial, spectral, and temporal information and (2) provide data
on the state of the earth’s surface. Previous studies suggested that remotely sensed
data should provide major benefits to hydrologic system analysis and water resources
management; yet, application potential with practical benefits was limited by modeling skill. One reason for this barrier is the lack of mathematical tools to convert
remotely sensed data to the type of information useful to seamlessly fit into the
actual needs in water resource systems (Kite and Pietroniro 1996). Some statistical models played an important role in translating data to information (Chang et al.
2009); however, limited by the inherent modeling structure and the semiempirical
nature, the breakthroughs in applications using statistical models were not salient.
Vastly improved instrumentation for sensing, logging, and transmitting hydrometric
measurements already facilitates retrieval of information from the far corners of
the earth’s surface (Mcculloch 2007). As we move from a data-poor to a data-rich
era due to the advances of remote sensing technologies, knowledge discovery and
management via a plethora of data mining and machine learning techniques become
promising, facing massive data sets (Harvey and Jiawei 2001).
Parameterization of hydrometeorological processes by advanced artificial intelligence methods has been widely used in modern hydrology (Storch and Zwiers 1999).
The empirical orthogonal function (EOF) is one of the decomposition procedures
used to investigate the spatiotemporal behavior of hydrometeorological processes;
thus, the main variance characters of meteorology are expected to be retrieved by
this technique (Puebla et al. 1998). Increasingly, models featured with artificial neural networks (ANNs) have been used in a wide range of disciplinary fields and have
become practical tools in hydrologic analyses for predictions. In many applications,