Conditional Simulation of Precipitation Time Series Using Phase Annealing

<p>Precipitation is one of the main inputs for hydrological models. For design purposes observed precipitation at high temporal resolution is often not available. In this case weather generators can be used to simulate realistic precipitation. Synthetic precipitation time series are often produced directly from observed time series using the stochastic methods which are able to reproduce the properties of the observed time series. The main difference and advantage of this research is to generate time series by focusing on the specific properties of the observed time series and trying to obtain these properties indirectly by conducting through investigation on the phases and power spectra and their individual effects using the phase annealing method.</p><p>Phase annealing is mainly based on annealing the phases of precipitation time series which are obtained from Fourier transform in order to meet the desired properties. These are obtained from observed time series and defined in the objective function. The outcome is synthetic time series with altered phases while the power spectrum is kept intact yielding new precipitation time series with properties matching those of the observed time series.</p>

Exploring the uncertainty of Weather Generators’ extreme estimates associated with the length of the input data series

<p>Stochastic Weather Generators (WG) have been extensively used in recent years for hydrologic modeling, among others. Compared to traditional approaches, the main advantage of using WGs is that they can produce synthetic continuous time series of weather data of unlimited length preserving their spatiotemporal distribution. Synthetic simulations are based on the statistical characteristics of the observed weather, thus, relying upon the length and spatial distribution of the input data series. In most cases, and especially in arid/semiarid regions, these are scarce, which makes it difficult for WGs to obtain reliable quantile estimates, particularly those associated with low-frequency events. The present study aims to explore the importance of the input weather data length in the performance of WGs, focusing on the adequate estimation of the higher quantiles, and quantifying their uncertainty.</p><p>An experimental case study consisting of nine rain gauges from the Spain02-v5 network in a 0.11º resolution covering an approximate area of 180 km<sup>2</sup> was implemented. The WG used for the experiment was GWEX, which includes a three-parameter (σ, κ, and ξ) cumulative distribution function (E-GPD) to model de precipitation amounts, being the shape parameter ξ the one directly governing the upper tail of the distribution function. A fictitious climate scenario of 15,000 was simulated fixing the ξ value to 0.11.  From this scenario, 50 realizations of 5,000 years with a different sample length (i.e. 30, 60, 90, 120, 150, 200, 300 years) were simulated for four different particular cases: (1) leaving the ξ value as default (i.e. 0.05); (2) estimating the ξ value from the observations; (3) calibrating the ξ value with the T = 100 years quantile from the 15,000 years; and (4) fixing the ξ value to the fictitious scenario value. Relative root mean square error (RRMSE) and coefficient of variation (CV) were calculated for each set of realizations and compared with the obtained from the fictitious climate scenario.</p><p>Preliminary results showed a clear reduction in the value of both the CV and the RRMSE with the increase of the sample length for the four particular cases, being this reduction more evident for the higher order quantiles and as we move from particular case (1) to (4). Furthermore, it was observed that there was not any significant improvement in the higher quantile estimates between the 200-yrs and the 300-yrs samples, concluding that there is a sample length threshold from which the estimates do not improve. Finally, even observing a clear improvement in all estimates when increasing the sample length, a systematic underestimation of the higher quantiles in all cases was still observed, which remarks the importance of seeking extra sources of information (e.g. regional max. Pd. studies) for a better parameterization of the WG, especially for arid/semiarid climates.</p>

Assessment of meteorological extremes using a synoptic weather generator and a downscaling model based on analogues

Natural risk studies such as flood risk assessments require long series of weather variables. As an alternative to observed series, which have a limited length, these data can be provided by weather generators. Among the large variety of existing ones, resampling methods based on analogues have the advantage of guaranteeing the physical consistency between local weather variables at each time step. However, they cannot generate values of predictands exceeding the range of observed values. Moreover, the length of the simulated series is typically limited to the length of the synoptic meteorological records used to characterize the large-scale atmospheric configuration of the generation day. To overcome these limitations, the stochastic weather generator proposed in this study combines two sampling approaches based on atmospheric analogues: (1) a synoptic weather generator in a first step, which recombines days of the 20th century to generate a 1000-year sequence of new atmospheric trajectories, and (2) a stochastic downscaling model in a second step applied to these atmospheric trajectories, in order to simulate long time series of daily regional precipitation and temperature. The method is applied to daily time series of mean areal precipitation and temperature in Switzerland. It is shown that the climatological characteristics of observed precipitation and temperature are adequately reproduced. It also improves the reproduction of extreme precipitation values, overcoming previous limitations of standard analogue-based weather generators.