Yuan Li

Remote Sensing-Derived Water Extent and Level to Constrain Hydraulic Flood Forecasting Models: Opportunities and Challenges

Accurate, precise and timely forecasts of flood wave arrival time, depth and velocity at each point of the floodplain are essential to reduce damage and save lives. Current computational capabilities support hydraulic models of increasing complexity over extended catchments. Yet a number of sources of uncertainty (e.g., input and boundary conditions, implementation data) may hinder the delivery of accurate predictions. Field gauging data of water levels and discharge have traditionally been used for hydraulic model calibration, validation and real-time constraint. However, the discrete spatial distribution of field data impedes the testing of the model skill at the two-dimensional scale. The increasing availability of spatially distributed remote sensing (RS) observations of flood extent and water level offers the opportunity for a comprehensive analysis of the predictive capability of hydraulic models. The adequate use of the large amount of information offered by RS observations triggers a series of challenging questions on the resolution, accuracy and frequency of acquisition of RS observations; on RS data processing algorithms; and on calibration, validation and data assimilation protocols. This paper presents a review of the availability of RS observations of flood extent and levels, and their use for calibration, validation and real-time constraint of hydraulic flood forecasting models. A number of conclusions and recommendations for future research are drawn with the aim of harmonising the pace of technological developments and their applications.

Assimilation of stream discharge for flood forecasting: Updating a semidistributed model with an integrated data assimilation scheme

Real-time discharge observations can be assimilated into flood models to improve forecast accuracy; however, the presence of time lags in the routing process and a lack of methods to quantitatively represent different sources of uncertainties challenge the implementation of data assimilation techniques for operational flood forecasting. To address these issues, an integrated error parameter estimation and lag-aware data assimilation (IEELA) scheme was recently developed for a lumped model. The scheme combines an ensemble-based maximum a posteriori (MAP) error estimation approach with a lag-aware ensemble Kalman smoother (EnKS). In this study, the IEELA scheme is extended to a semidistributed model to provide for more general application in flood forecasting by including spatial and temporal correlations in model uncertainties between subcatchments. The result reveals that using a semidistributed model leads to more accurate forecasts than a lumped model in an open-loop scenario. The IEELA scheme improves the forecast accuracy significantly in both lumped and semidistributed models, and the superiority of the semidistributed model remains in the data assimilation scenario. However, the improvements resulting from IEELA are confined to the outlet of the catchment where the discharge observations are assimilated. Forecasts at “ungauged” internal locations are not improved, and in some instances, even become less accurate.

Application of Remote Sensing Data to Constrain Operational Rainfall-Driven Flood Forecasting: A Review

Fluvial flooding is one of the most catastrophic natural disasters threatening people’s lives and possessions. Flood forecasting systems, which simulate runoff generation and propagation processes, provide information to support flood warning delivery and emergency response. The forecasting models need to be driven by input data and further constrained by historical and real-time observations using batch calibration and/or data assimilation techniques so as to produce relatively accurate and reliable flow forecasts. Traditionally, flood forecasting models are forced, calibrated and updated using in-situ measurements, e.g., gauged precipitation and discharge. The rapid development of hydrologic remote sensing offers a potential to provide additional/alternative forcing and constraint to facilitate timely and reliable forecasts. This has brought increasing interest to exploring the use of remote sensing data for flood forecasting. This paper reviews the recent advances on integration of remotely sensed precipitation and soil moisture with rainfall-runoff models for rainfall-driven flood forecasting. Scientific and operational challenges on the effective and optimal integration of remote sensing data into forecasting models are discussed.

Estimation of carbon dioxide emission in Beijing city

In order to evaluate the impact of urbanization on carbon cycle, we selected the Beijing city as the study area to estimate CO2 emission both from anthropogenic activity and natural processes. The spatial pattern of annual CO2 emission in 1992 and 2005 in Beijing city was estimated by using the socio-economic statistical data, remote sensing data and GIS-based approach. The main factors include transport, cement production, human respiration, carbon assimilation of green space and cropland. The vehicle density was determined by combining the remote sensing images from Google-Earth with road map and statistical data. The human respiration was estimated from census data and allocated to different districts. In order to estimate the impact of land use/land cover change on carbon absorption, we first extracted the green space and cropland from Landsat TM data in 1992 and 2005 by using maximum likelihood method. The annual carbon assimilation of green space and cropland was estimated finally according to the carbon flux observation in green space and the simulated results for cropland by using Biome-BGC model. The results showed that the vehicles and industry contributed most in Beijing city. The total CO2 emission had increased about 1.7 times from 16,232.99 kiloton in 1992 to 28,091.93 kiloton in 2005. The increase of green space can partly compensate the decrease of cropland caused by urbanization at a certain extent.

Effective Representation of River Geometry in Hydraulic Flood Forecast Models

Bathymetric data are a critical input to hydraulic models. However, river depth and shape cannot be systematically observed remotely, and field data are both scarce and expensive to collect. In flood modeling, river roughness and geometry compensate for each other, with different parameter sets often being able to map model predictions equally well to the observed data, commonly known as equifinality. This study presents a numerical experiment to investigate an effective yet parsimonious representation of channel geometry that can be used for operational flood forecasting. The LISFLOOD-FP hydraulic model was used to simulate a hypothetical flood event in the Clarence catchment (Australia). A high-resolution model simulation based on accurate bathymetric field data was used to benchmark coarser model simulations based on simplified river geometries. These simplified river geometries were derived from a combination of globally available empirical formulations, remote sensing data, and a limited number of measurements. Model predictive discrepancy between simulations with field data and simplified geometries allowed an assessment of the geometry impact on inundation dynamics. In this study site, the channel geometrical representation for a reliable inundation forecast could be achieved using remote sensing-derived river width values combined with a few measurements of river depth sampled at strategic locations. Furthermore, this study showed that spatially distributed remote sensing-derived inundation levels at the very early stages of a flood event have the potential to support the effective diagnosis of errors in model implementations. Plain Language Summary Floods are among the most frequent and destructive natural disasters worldwide. An accurate and reliable flood forecast can provide vital information for land management and emergency response. Flood forecasts are achieved using numerical models that are able to predict the depth, velocity, and arrival time of the flood wave at each point of the valley. The accuracy of these predictions is strongly related to the quality of the three-dimensional representation of the valley. In particular, information on river geometry (that is cross-section shape, depth, and width) is critical to the application of these numerical models. However, it is impossible to measure river geometry along the entire river length, especially in large basins. This study developed a method to represent river geometry using a limited amount of time and money. Specifically, this objective can be achieved using available satellite imagery complemented with a few measurements. Moreover, this study showed that flood forecast skill can be improved by combining information from satellite and numerical models. Albeit simple, the river geometry representation proposed in this study can support the accurate prediction of floodplain inundation. This method was described and tested using, as example, a hypothetical flood event in an Australian catchment.

Impact of Rain Gauge Quality Control and Interpolation on Streamflow Simulation: An Application to the Warwick Catchment, Australia

Rain gauges are widely used to obtain temporally continuous point rainfall records, which are then interpolated into spatially continuous data to force hydrological models. However, rainfall measurements and interpolation procedure are subject to various uncertainties, which can be reduced by applying quality control and selecting appropriate spatial interpolation approaches. Consequently, the integrated impact of rainfall quality control and interpolation on streamflow simulation has attracted increased attention but not been fully addressed. This study applies a quality control procedure to the hourly rainfall measurements obtained in the Warwick catchment in eastern Australia. The grid-based daily precipitation from the Australian Water Availability Project was used as a reference. The Pearson correlation coefficient between the daily accumulation of gauged rainfall and the reference data was used to eliminate gauges with significant quality issues. The unrealistic outliers were censored based on a comparison between gauged rainfall and the reference. Four interpolation methods, including the inverse distance weighting (IDW), nearest neighbors (NN), linear spline (LN), and ordinary Kriging (OK), were implemented. The four methods were firstly assessed through a cross-validation using the quality-controlled rainfall data. The impacts of the quality control and interpolation on streamflow simulation were then evaluated through a semi-distributed hydrological model. The results showed that the Nash–Sutcliffe model efficiency coefficient (NSE) and Bias of the streamflow simulations were significantly improved after quality control. In the cross-validation, the IDW and OK methods resulted in good interpolation rainfall, while the NN led to the worst result. In term of the impact on hydrological prediction, the IDW led to the most consistent streamflow predictions with the observations, according to the validation at five streamflow-gauged locations. The OK method performed second best according to streamflow predictions at the five gauges in the calibration period (01/01/2007–31/12/2011) and four gauges during the validation period (01/01/2012–30/06/2014). However, NN produced the worst prediction at the outlet of the catchment in the validation period, indicating a low robustness. While the IDW exhibited the best performance in the study catchment in terms of accuracy, robustness and efficiency, more general recommendations on the selection of rainfall interpolation methods need to be further explored.