%0 Journal Article %1 ELMI2023128880 %A Elmi, Omid %A Tourian, Mohammad J. %D 2023 %J Journal of Hydrology %K geoinst imported %P 128880 %R https://doi.org/10.1016/j.jhydrol.2022.128880 %T Retrieving time series of river water extent from global inland water data sets %U https://www.sciencedirect.com/science/article/pii/S0022169422014500 %V 617 %X The accurate monitoring of the surface water storage as an essential component of the global water cycle requires a realistic representation of river networks and channel characteristics. Since such a representation has not been available for many rivers and is becoming less available even for many gauged rivers, crucial questions about the spatio-temporal dynamics of freshwater in river networks cannot be answered properly. The global coverage and fine temporal resolution of satellite imagery provide the opportunity to obtain time series of surface water extent at the global scale for almost all rivers. However, despite recent advances in satellite imaging sensors, water extraction algorithms, and big data processing capabilities, none of the available global water extent data sets can meet the necessary requirements in terms of accuracy and spatio-temporal resolutions. Due to the inherent complexity of monitoring the river surface extent, efforts have been limited to the development of global river extent data sets with a limited number of temporal layers usually obtained from long-term averaged satellite imagery. In this study, we propose a region-based image restoration algorithm to obtain the river surface extent from a pre-existing global inland water data set by incorporating temporal and spatial constraints between pixel labels. We employ our algorithm on the Monthly Water History maps of the Global Surface Water data set (Pekel et al., 2016). We validate the proposed method on 98 river reaches that their average width ranges from approximately 36m to 3400m with in situ discharge measurements in the Mississippi, Amazon, Niger and Po river basins. The obtained river width time series exhibit a strong monotonic relationship with discharge measurements as the Spearman correlation coefficients are predominantly larger than 0.70 (on average 0.74 for Mississippi, 0.84 for Amazon, 0.86 for Niger and 0.77 for Po rivers). Such a performance confirms that the proposed method can facilitate the acquisition of a global dynamic river extent data set, which plays a key role in better understanding the distribution and availability of freshwater across continents.

@article{ELMI2023128880, abstract = {The accurate monitoring of the surface water storage as an essential component of the global water cycle requires a realistic representation of river networks and channel characteristics. Since such a representation has not been available for many rivers and is becoming less available even for many gauged rivers, crucial questions about the spatio-temporal dynamics of freshwater in river networks cannot be answered properly. The global coverage and fine temporal resolution of satellite imagery provide the opportunity to obtain time series of surface water extent at the global scale for almost all rivers. However, despite recent advances in satellite imaging sensors, water extraction algorithms, and big data processing capabilities, none of the available global water extent data sets can meet the necessary requirements in terms of accuracy and spatio-temporal resolutions. Due to the inherent complexity of monitoring the river surface extent, efforts have been limited to the development of global river extent data sets with a limited number of temporal layers usually obtained from long-term averaged satellite imagery. In this study, we propose a region-based image restoration algorithm to obtain the river surface extent from a pre-existing global inland water data set by incorporating temporal and spatial constraints between pixel labels. We employ our algorithm on the Monthly Water History maps of the Global Surface Water data set (Pekel et al., 2016). We validate the proposed method on 98 river reaches that their average width ranges from approximately 36m to 3400m with in situ discharge measurements in the Mississippi, Amazon, Niger and Po river basins. The obtained river width time series exhibit a strong monotonic relationship with discharge measurements as the Spearman correlation coefficients are predominantly larger than 0.70 (on average 0.74 for Mississippi, 0.84 for Amazon, 0.86 for Niger and 0.77 for Po rivers). Such a performance confirms that the proposed method can facilitate the acquisition of a global dynamic river extent data set, which plays a key role in better understanding the distribution and availability of freshwater across continents.}, added-at = {2023-04-03T11:36:03.000+0200}, author = {Elmi, Omid and Tourian, Mohammad J.}, biburl = {https://puma.ub.uni-stuttgart.de/bibtex/2a60cd421c2665bd59daed554427544e2/tourian}, doi = {https://doi.org/10.1016/j.jhydrol.2022.128880}, interhash = {ac9b8e442acbc64e25d3bb9ee93aaf64}, intrahash = {a60cd421c2665bd59daed554427544e2}, issn = {0022-1694}, journal = {Journal of Hydrology}, keywords = {geoinst imported}, pages = 128880, timestamp = {2023-04-03T11:36:03.000+0200}, title = {Retrieving time series of river water extent from global inland water data sets}, url = {https://www.sciencedirect.com/science/article/pii/S0022169422014500}, volume = 617, year = 2023 }