At site flood frequency analysis (FFA) in arid/semi-arid watersheds poses unique challenges to researchers and practitioners due to the generally limited data records. This study presents a comprehensive evaluation of FFA in arid/semi-arid watersheds in relation to the unique characteristics of these regions, such as the limited number of floods occurring each year and the large variability of the flood peak discharges. Study cases in Israel and the US are examined and compared with non-arid watersheds, characterized by Mediterranean climate, and with synthetic flood records. Results show that the tail of extreme value distributions describing arid/semi-arid watersheds is found to be heavier than the one describing Mediterranean watersheds. The number of yearly floods and the variability of flood peak discharge are shown to have a crucial impact on the accuracy of the quantile estimates with smaller number of events per year and larger coefficient of variation of flood peak discharge being related to larger errors in the estimated quantiles. Partial duration series approach provides a slightly reduced bias in the estimates, but should not be blindly preferred over annual maxima series as it presents comparable estimation uncertainty. In general, the generalized extreme value and the generalized Pareto distribution are found to be non-optimal choices for the examined arid/semi-arid watersheds.
The metastatistical extreme value approach proved promising in the frequency analysis of daily precipitation from ordinary events, outperforming traditional methods based on sampled extremes. However, subdaily applications are currently restrained by two knowledge gaps: It is not known if ordinary events can be consistently examined over durations, and it is not clear to what extent their entire distributions represent extremes. We propose here a unified definition of ordinary events across durations and suggest the simplified metastatistical extreme value formulation for dealing with extremes emerging from the tail, rather than the entire distributions, of ordinary events. This unified framework provides robust estimates of extreme quantiles (\textless10% error on the 100 yr from a 26 yr long record) and allows representations in which ordinary and extreme events share the scaling exponent. Future applications could improve our knowledge of subdaily extreme precipitation and help investigate the impact of local factors and climatic forcing on their frequency.
Abstract Water volume estimates of shallow desert lakes are the basis for water balance calculations, important both for water resource management and paleohydrology/climatology. Water volumes are typically inferred from bathymetry mapping; however, being shallow, ephemeral and remote, bathymetric surveys are scarce in such lakes. We propose a new, remote-sensing based, method to derive the bathymetry of such lakes using the relation between water occurrence, during \textgreater30-yr of optical satellite data, and accurate elevation measurements from the new Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2). We demonstrate our method at three locations where we map bathymetries with \~0.3 m error. This method complements other remotely sensed, bathymetry-mapping methods as it can be applied to: (a) complex lake systems with sub-basins, (b) remote lakes with no in-situ records, and (c) flooded lakes. The proposed method can be easily implemented in other shallow lakes as it builds on publically accessible global data sets.
Flood-fed aquifers along the sandy lower reach of the Kuiseb River sustain a 130-km-long green belt of lush oases across the hyperarid Namib desert. This oasis is a year-round source for water creating dense-tall woodland along the narrow corridor of the ephemeral river valley, which, in turn, supports human activity and fauna including during the long dry austral winters and multi-year droughts. Occasional floods, originating at the river's wetter headwaters, travel ∼280 km downstream, before recharging these aquifers. We analyzed the flood-aquifer-vegetation dynamics at-a-site and along the river, determining the relative impact of floods with diverse magnitude and frequency on downstream reaches. We find that flood discharge that feeds the alluvial aquifers also affects vegetation dynamics along the river. The downstream aquifers are fed only by the largest floods that allow the infrequent germination of plants; mean annual recharge volume is too low to support the aquifers level. These short-term vegetation cycles of green-up and then fast senescence in-between floods are easily detected by satellite-derived vegetation index. This index identifies historical floods and their magnitudes in arid and hyperarid regions; specifically, it determines occurrences of large floods in headwater-fed, ephemeral Namib streams as well as in other hyperarid regions. Our study reveals the importance of flood properties on the oasis life cycle, emphasizing the impact of drought and wet years on the Namib's riparian vegetation.
The performances of hydrological models in arid areas are significantly lower than other climates. The reasons are numerous, from the scales involved, to specific processes and the lack of adequate measurements. Effective parameters have been often observed to change between runoff events, limiting the predictive capacity of the models. We look at the problems that can be found in an operational setting and present an analysis to improve the understanding of the errors. Our method characterizes the conditions where the model fails systematically, and the conditions where the parameterization holds between floods. We applied KINEROS2 to 24 years of radar rainfall and streamflow data in 6 arid catchments. A GLUE probabilistic framework is used to characterize model performance, and a method is developed to identify floods with similar calibration. The analysis shows that uninformative conditions are difficult to generalize. A basin-specific analysis can help to identify conditions where the model fails and exclude them from calibration. Despite the large uncertainties, similar catchments display groups of floods with similar parameterization. In some basin we find that it is important to quantify antecedent moisture conditions. Hydrological models show some consistency within limited conditions. These conditions, however, depend on the errors involved, and are site-specific. There is some potential for parameter transfer, but proximity alone might not be enough, and other factors such as mean annual rainfall or storm type, should be taken into account.
Heavy precipitation events (HPEs) can lead to natural hazards (e.g. floods and debris flows) and contribute to water resources. Spatiotemporal rainfall patterns govern the hydrological, geomorphological, and societal effects of HPEs. Thus, a correct characterisation and prediction of rainfall patterns is crucial for coping with these events. Information from rain gauges is generally limited due to the sparseness of the networks, especially in the presence of sharp climatic gradients. Forecasting HPEs depends on the ability of weather models to generate credible rainfall patterns. This paper characterises rainfall patterns during HPEs based on high-resolution weather radar data and evaluates the performance of a high-resolution, convection-permitting Weather Research and Forecasting (WRF) model in simulating these patterns. We identified 41 HPEs in the eastern Mediterranean from a 24-year radar record using local thresholds based on quantiles for different durations, classified these events into two synoptic systems, and ran model simulations for them. For most durations, HPEs near the coastline were characterised by the highest rain intensities; however, for short durations, the highest rain intensities were found for the inland desert. During the rainy season, the rain field's centre of mass progresses from the sea inland. Rainfall during HPEs is highly localised in both space (less than a 10 km decorrelation distance) and time (less than 5 min). WRF model simulations were accurate in generating the structure and location of the rain fields in 39 out of 41 HPEs. However, they showed a positive bias relative to the radar estimates and exhibited errors in the spatial location of the heaviest precipitation. Our results indicate that convection-permitting model outputs can provide reliable climatological analyses of heavy precipitation patterns; conversely, flood forecasting requires the use of ensemble simulations to overcome the spatial location errors.
AbstractWhile CMIP5 models robustly project drying of the subtropics and more precipitation in the tropics and subpolar latitudes by the end of the century, the magnitude of these changes in precipitation varies widely across models: for example, some models simulate no drying in the eastern Mediterranean while others simulate more than a 50% reduction in precipitation relative to the model-simulated present-day value. Furthermore, the factors leading to changes in local subtropical precipitation remain unclear. The importance of zonal-mean changes in atmospheric structure for local precipitation changes is explored in 42 CMIP5 models. It is found that up to half of the local intermodel spread over the Mediterranean, northern Mexico, East Asia, southern Africa, southern Australia, and southern South America is related to the intermodel spread in large-scale processes such as the magnitude of globally averaged surface temperature increases, Hadley cell widening, polar amplification, stabilization of the tropical upper troposphere, or changes in the polar stratosphere. Globally averaged surface temperature increases account for intermodel spread in land subtropical drying in the Southern Hemisphere but are not important for land drying adjacent to the Mediterranean. The factors associated with drying over the eastern Mediterranean and western Mediterranean differ, with stabilization of the tropical upper troposphere being a crucial factor for the former only. Differences in precipitation between the western and eastern Mediterranean are also evident on interannual time scales. In contrast, the global factors examined here are unimportant over most of the United States, and more generally over the interior of continents. Much of the rest of the spread can be explained by variations in local relative humidity, a proxy also for zonally asymmetric circulation and thermodynamic changes.