Jaap Schellekens

The Delft-FEWS flow forecasting system

Since its introduction in 2002/2003, the current generation of the Delft-FEWS operational forecasting platform has found application in over forty operational centres. In these it is used to link data and models in real time, producing forecasts on a daily basis. In some cases it forms a building block of a country-wide national forecasting system using distributed client-server technology. In other cases it is applied at a much smaller scale on a simple desktop workstation, providing forecasts for a single basin. The flexibility of the software in open integration of models and data has additionally appealed to the research community.This paper discusses the principles on which the Delft-FEWS system has been developed, as well as a brief background of the architecture of the system and concepts used for storing and handling data. One of the key features of the system is its flexibility in integrating (third-party) models and data, and the available approaches to linking models and accessing data are highlighted. A brief overview of different applications of the system is given to illustrate how the software is used to support differing objectives in the domain of real time environmental modelling. Highlights? A state of the art real time environmental decision support system is presented. ? Open model integration is shown to be key to sustainable application. ? Clear interfaces are shown to be effective in reducing complexity.

Evaporation from a tropical rain forest, Luquillo Experimental Forest, eastern Puerto Rico

Evaporation losses from a watertight 6.34 ha rain forest catchment under wet maritime tropical conditions in the Luquillo Experimental Forest, Puerto Rico, were determined using complementary hydrological and micrometeorological techniques during 1996 and 1997. At 6.6 mm d -1 for 1996 and 6.0 mm d -1 for 1997, the average evapotranspiration (ET) of the forest is exceptionally high. Rainfall interception (Ei), as evaluated from weekly throughfall measurements and an average stemflow fraction of 2.3%, accounted for much (62-74%) of the ET at 4.9 mm d -1 in 1996 and 3.7 mm d -1 in 1997. Average transpiration rates (Et) according to a combination of the temperature fluctuation method and the Penman-Monteith equation were modest at 2.2 mm d -1 and 2.4 mm d -1 in 1996 and 1997, respectively. Both estimates compared reasonably well with the water-budget-based estimates (ET - Ei) of 1.7 mm d -1 and 2.2 mm d -1. Inferred rates of wet canopy evaporation were roughly 4 to 5 times those predicted by the Penman- Monteith equation, with nighttime rates very similar to daytime rates, suggesting radiant energy is not the dominant controlling factor. A combination of advected energy from the nearby Atlantic Ocean, low aerodynamic resistance, plus frequent low-intensity rain is thought to be the most likely explanation of the observed discrepancy between measured and estimated Ei.

Improving Curve Number Based Storm Runoff Estimates Using Soil Moisture Proxies

Advances in data dissemination and the availability of new remote sensing datasets present both opportunities and challenges for hydrologists in improving flood forecasting systems. The current study investigates the improvement in SCS curve number (CN)-based storm runoff estimates obtained after inclusion of various soil moisture proxies based on additional data on precipitation, baseflow, and soil moisture. A dataset (1980-2007) comprising 186 Australian catchments (ranging from 51 to 1979 km 2 in size) was used. In order to investigate the value of a particular proxy, the observed S (potential maximum retention) was compared to values obtained with different soil moisture proxies using linear regression. An antecedent precipitation index (API) based on gauged precipitation using a decay parameter proved most valuable in improving storm runoff estimates, stressing the importance of high quality precipitation data. An antecedent baseflow index (ABFI) also performed well. Proxies based on remote sensing (TRMM and AMSR-E) gave promising results, particularly when considering the expected arrival of higher accuracy data from upcoming satellites. The five-day API performed poorly. The inclusion of soil moisture proxies resulted in mean modeled versus observed correlation coefficients around 0.75 for almost all proxies. The greatest improvement in runoff estimates was observed in drier catchments with low Enhanced Vegetation Index (EVI) and topographical slope (all intercorrelated parameters). The present results suggest the usefulness of incorporating remotely sensed proxies for soil moisture and catchment wetness in flood forecasting systems.

Global‐scale regionalization of hydrologic model parameters

Current state-of-the-art models typically applied at continental to global scales (hereafter called macroscale) tend to use a priori parameters, resulting in suboptimal streamflow (Q) simulation. For the first time, a scheme for regionalization of model parameters at the global scale was developed. We used data from a diverse set of 1787 small-to-medium sized catchments ( 10–10,000 km2) and the simple conceptual HBV model to set up and test the scheme. Each catchment was calibrated against observed daily Q, after which 674 catchments with high calibration and validation scores, and thus presumably good-quality observed Q and forcing data, were selected to serve as donor catchments. The calibrated parameter sets for the donors were subsequently transferred to 0.5° grid cells with similar climatic and physiographic characteristics, resulting in parameter maps for HBV with global coverage. For each grid cell, we used the 10 most similar donor catchments, rather than the single most similar donor, and averaged the resulting simulated Q, which enhanced model performance. The 1113 catchments not used as donors were used to independently evaluate the scheme. The regionalized parameters outperformed spatially uniform (i.e., averaged calibrated) parameters for 79% of the evaluation catchments. Substantial improvements were evident for all major Koppen-Geiger climate types and even for evaluation catchments > 5000 km distant from the donors. The median improvement was about half of the performance increase achieved through calibration. HBV with regionalized parameters outperformed nine state-of-the-art macroscale models, suggesting these might also benefit from the new regionalization scheme. The produced HBV parameter maps including ancillary data are available via www.gloh2o.org.

Real-Time Geospatial Data Handling and Forecasting: Examples From Delft-FEWS Forecasting Platform/System

Environmental research that involves assessment of climate impact or extreme events like drought, floods, or other risks often requires the combination of geospatial data, distributed models, and weather forecast products. The Delft-FEWS forecasting platform, used by operational flood warning organizations around the world, offers the opportunity to carry out this type of research within a state-of-the-art forecasting environment. This paper shows opportunities Delft-FEWS offers in dealing with geospatial data like satellite measurements, weather, and climate forecasts. It also shows that Delft-FEWS offers the possibility to include distributed hydrological models using an embedded version of PCRASTER. This is illustrated by two case studies that combine different geospatial data, satellite data, and AOGCM climate data, with two different types of models, a groundwater model and a global hydrological model, within the Delft-FEWS system.

A Methodology to Determine Radio-Frequency Interference in AMSR2 Observations

A study to determine radio-frequency interference (RFI) in low-frequency passive microwave observations of the Advanced Microwave Scanning Radiometer-2 (AMSR2) is performed. RFI detection methods, such as the spectral difference method, have already been applied on microwave satellite sensors. However, these methods may result in false RFI detection, particularly in zones with extreme environmental conditions. To overcome this problem, this paper proposes an approach that uses the additional 7.3-GHz channel of the AMSR2 sensor in a new RFI detection method. This method uses calculated standard errors of estimate to detect RFI contamination in 6.9- and 7.3-GHz observations. It was found that 6.9-GHz observations are mainly contaminated in the USA, India, Japan, and parts of Europe. The 7.3-GHz observations are contaminated in South America, Ukraine, the Middle East, Southeast Asia, and Russia. The fact that these channels are not affected by RFI in exactly the same regions is useful for studies that prefer C-band brightness temperature observations (e.g., soil moisture retrieval algorithms). Therefore, a decision tree approach was set up to determine RFI and to select reliable brightness temperature observations in the lowest frequency free of any man-made contamination. The result is a reduction of the total contaminated pixels in the 6.9-GHz observations of 66% for horizontal observations and even 85% for vertical observations when 7.3 and 10.7 GHz are used. By linking RFI maps with civilization maps, this paper further shows that RFI sources at the C-band frequency are mainly located in urbanized areas.

River gauging at global scale using optical and passive microwave remote sensing

Recent discharge observations are lacking for most rivers globally. Discharge can be estimated from remotely sensed floodplain and channel inundation area, but there is currently no method that can be automatically extended to many rivers. We examined whether automated monitoring is feasible by statistically relating inundation estimates from moderate to coarse (>0.05°) resolution remote sensing to monthly station discharge records. Inundation extents were derived from optical MODIS data and passive microwave sensors, and compared to monthly discharge records from over 8000 gauging stations and satellite altimetry observations for 442 reaches of large rivers. An automated statistical method selected grid cells to construct “satellite gauging reaches” (SGRs). MODIS SGRs were generally more accurate than passive microwave SGRs, but there were complementary strengths. The rivers widely varied in size, regime, and morphology. As expected performance was low (R   0.6. The best results (R > 0.9) were obtained for large unregulated lowland rivers, particularly in tropical and boreal regions. Relatively poor results were obtained in arid regions, where flow pulses are few and recede rapidly, and in temperate regions, where many rivers are modified and contained. Provided discharge variations produce clear changes in inundated area and gauge records are available for part of the satellite record, SGRs can retrieve monthly river discharge values back to around 1998 and up to present.

A 30 m Resolution Surface Water Mask Including Estimation of Positional and Thematic Differences Using Landsat 8, SRTM and OpenStreetMap: A Case Study in the Murray-Darling Basin, Australia

Accurate maps of surface water are essential for many environmental applications. Surface water maps can be generated by combining measurements from multiple sources. Precise estimation of surface water using satellite imagery remains a challenging task due to the sensor limitations, complex land cover, topography, and atmospheric conditions. As a complementary dataset, in the case of hilly landscapes, a drainage network can be extracted from high-resolution digital elevation models. Additionally, Volunteered Geographic Information (VGI) initiatives, such as OpenStreetMap, can also be used to produce high-resolution surface water masks. In this study, we derive a high-resolution water mask using Landsat 8 imagery and OpenStreetMap as well as (potential) a drainage network using 30 m SRTM. Our approach to derive a surface water mask from Landsat 8 imagery comprises the use of a lower 15% percentile of Landsat 8 Top of Atmosphere (TOA) reflectance from 2013 to 2015. We introduce a new non-parametric unsupervised method based on the Canny edge filter and Otsu thresholding to detect water in flat areas. For hilly areas, the method is extended with an additional supervised classification step used to refine the water mask. We applied the method across the Murray-Darling basin, Australia. Differences between our new Landsat-based water mask and the OpenStreetMap water mask regarding positional differences along the rivers and overall coverage were analyzed. Our results show that about 50% of the OpenStreetMap linear water features can be confirmed using the water mask extracted from Landsat 8 imagery and the drainage network derived from SRTM. We also show that the observed distances between river features derived from OpenStreetMap and Landsat 8 are mostly smaller than 60 m. The differences between the new water mask and SRTM-based linear features and hilly areas are slightly larger (110 m). The overall agreement between OpenStreetMap and Landsat 8 water masks is about 30%.

Rapid setup of hydrological and hydraulic models using OpenStreetMap and the SRTM derived digital elevation model

A stepwise procedure has been developed in Python to extract information from OpenStreetMap (OSM) for hydrological and hydraulic models using existing and newly developed tools. The procedure focuses on the extraction of paved areas and water bodies. Road density is used to fill in gaps in OSM polygon coverage. Furthermore, it includes automatic downloading of Shuttle Radar Topography Mission (SRTM) elevation data and improving the elevation model with man-made landscape features such as elevated roads that are sampled from OSM. This is useful for hydraulic modelling in data scarce flood plain areas, where sharp elevation differences are dominated by man-made elevated elements. Test cases in Europe, South East Asia and East Africa demonstrate the potential of the procedure, although large differences in completeness of OSM coverage suggest it is best used in combination with other data sources. A tool is presented that extracts data for hydrological modelling from OpenStreetMap.Extracted impervious area shows good correlation with existing sources in Europe.Extracted road density is a good proxy to detect urbanized areas.

Spatial Downscaling of Satellite-Based Precipitation and Its Impact on Discharge Simulations in the Magdalena River Basin in Colombia

Precipitation is one of the most important components of the water cycle and its accurate spatial and temporal representation is fundamental for hydrological modeling. In the present study, we investigated the impact of spatial resolution of various precipitation datasets on discharge estimates. First, a new precipitation spatial downscaling procedure was developed and applied to four gridded global precipitation datasets based on (i) solely satellite observations: CMORPH and PERSIANN, (ii) satellite and in situ observations: TRMM and (iii) satellite and in situ observations and reanalysis data: MSWEP. The here presented downscaling methodology blended global precipitation datasets with data on vegetation and topography to improve the representation of precipitation spatial variability. Second, interpolated in situ, non-downscaled (25 km) and downscaled (1 km) precipitation data were used to force a grid-distributed version of the HBV-96 rainfall-runoff model for the Magdalena River basin in Colombia. Results showed that MSWEP and TRMM outperformed CMORPH and PERSIANN precipitation datasets. The downscaling procedure resulted in considerable improvements in coefficient of determination, root mean square error and bias in comparison with in situ precipitation observations. Discharge model estimates were also in better agreement with the observations when the model was forced with the downscaled precipitation. Model performance was improved with Kling Gupta efficiency increases in the order of 0.1 to 0.5. Moreover, better discharge simulations were obtained using downscaled precipitation compared to using only in situ precipitation data when using less than 100 stations.