Considering the popularity of using data-driven non-linear methods for forecasting streamflow, there has been no exploration of how well such models perform in climate regimes with differing hydrological characteristics, nor has the performance of these models, coupled with wavelet transforms, been compared for lead times of less than one month. This study compares the use of four different models, namely artificial neural networks (ANNs), support vector regression (SVR), wavelet-ANN, and wavelet-SVR in a Mediterranean, Oceanic, and Hemiboreal watershed. Model performance was tested for one, two and three day forecasting lead times, measured by fractional standard error, the coefficient of determination, Nash–Sutcliffe model efficiency, multiplicative bias, probability of detection and false alarm rate. SVR based models performed best overall, but no one model outperformed the others in more than one watershed, suggesting that some models may be more suitable for certain types of data. Overall model performance varied greatly between climate regimes, suggesting that higher persistence and slower hydrological processes (i.e. snowmelt, glacial runoff, and subsurface flow) support reliable forecasting using daily and multi-day lead times.

Identification of a geomorphic index to represent lower thresholds for minor flows in ephemeral, alluvial streams in arid environments is an essential step in reliable flash flood hazard estimations and establishing flood warning systems. An index, termed Alluvial wadi Flood Incipient Geomorphologic Index (AFIG), is presented. Analysis of data from an extensive field survey in the arid ephemeral streams in southern and eastern Israel was conducted to investigate the AFIG and the control over its value across the region. During the survey we identified distinguishable flow marks in the lower parts of streams' banks, such as niches, vegetation line, and change in bank material, which are indicative of low flows. The cross-sectional characteristics of the AFIG were studied in relationship with contributing drainage basin characteristics such as lithology, topography, and precipitation. Drainage area and hardness of the exposed lithology (presented as a basin-wide index) are the preferred descriptors to be used in estimating a specific AFIG in un-surveyed sites. Analyses of discharge records from seven hydrometric stations indicate that the recurrence interval of the determined AFIG is equal to or more frequent than 0.5 year.

Runoff and flash flood generation are very sensitive to rainfall's$\backslash$nspatial and temporal variability. The increasing use of radar and$\backslash$nsatellite data in hydrological applications, due to the sparse$\backslash$ndistribution of rain gauges over most catchments worldwide, requires$\backslash$nfurthering our knowledge of the uncertainties of these data. In 2011, a$\backslash$nnew super-dense network of rain gauges containing 14 stations, each with$\backslash$ntwo side-by-side gauges, was installed within a 4 km(2) study area near$\backslash$nKibbutz Galed in northern Israel. This network was established for a$\backslash$ndetailed exploration of the uncertainties and errors regarding rainfall$\backslash$nvariability within a common pixel size of data obtained from remote$\backslash$nsensing systems for timescales of 1 min to daily. In this paper, we$\backslash$npresent the analysis of the first year's record collected from this$\backslash$nnetwork and from the Shacham weather radar, located 63 km from the study$\backslash$narea. The gauge-rainfall spatial correlation and uncertainty were$\backslash$nexamined along with the estimated radar error. The nugget parameter of$\backslash$nthe inter-gauge rainfall correlations was high (0.92 on the 1 min scale)$\backslash$nand increased as the timescale increased. The variance reduction factor$\backslash$n(VRF), representing the uncertainty from averaging a number of rain$\backslash$nstations per pixel, ranged from 1.6% for the 1 min timescale to 0.07%$\backslash$nfor the daily scale. It was also found that at least three rain stations$\backslash$nare needed to adequately represent the rainfall (VRF\textless 5 %) on a typical$\backslash$nradar pixel scale. The difference between radar and rain gauge rainfall$\backslash$nwas mainly attributed to radar estimation errors, while the gauge$\backslash$nsampling error contributed up to 20% to the total difference. The ratio$\backslash$nof radar rainfall to gauge-areal-averaged rainfall, expressed by the$\backslash$nerror distribution scatter parameter, decreased from 5.27 dB for 3 min$\backslash$ntimescale to 3.21 dB for the daily scale. The analysis of the radar$\backslash$nerrors and uncertainties suggest that a temporal scale of at least 10$\backslash$nmin should be used for hydrological applications of the radar data.$\backslash$nRainfall measurements collected with this dense rain gauge network will$\backslash$nbe used for further examination of small-scale rainfall's spatial and$\backslash$ntemporal variability in the coming years.

The spaceborne weather radar onboard the Tropical Rainfall Measuring Mission (TRMM) satellite can be used to adjust Ground-based Radar (GR) echoes, as a function of the range from the GR site. The adjustment is based on the average linear radar reflectivity in circular rings around the GR site, for both the GR and attenuation-corrected NearSurfZ TRMM Precipitation Radar (TPR) images. In previous studies, it was found that in winter, for the lowest elevation of the Cyprus C-band radar, the GR/TPR equivalent rain rate ratio was decreasing, on average, of approximately 8 dB per decade. In this paper, the same analysis has been applied to another C-band radar in the southeastern Mediterranean area. For the lowest elevation of the “Shacham” radar in Israel, the GR/TPR equivalent rain rate ratio is found to decrease of approximately 6 dB per decade. The average departure at the “reference”, intermediate range is related to the calibration of the GR. The negative slope of the range dependence is considered to be mainly caused by an overshooting problem (increasing sampling volume of the GR with range combined with non-homogeneous beam filling and, on average, a decreasing vertical profile of radar reflectivity). To check this hypothesis, we have compared the same NearSurfZ TPR images versus GR data acquired using the second elevation. We expected these data to be affected more by overshooting, especially at distant ranges: the negative slope of the range dependence was in fact found to be more evident than in the case of the lowest GR elevation for both the Cypriot and Israeli radar.