Nonlinear estimation of spatial distribution of rainfall — An indicator cokriging approach

Indicator cokriging (Journel 1983) is examined as a tool for real-time estimation of rainfall from rain gage measurements. The approach proposed in this work obviates real-time estimation of real-time statistics of rainfall by using ensemble or climatological statistics exclusively, and reduces computational requirements attendant to indicator cokriging by employing only a few auxiliary cutoffs in estimation of conditional probabilities. Due to unavailability of suitable rain gage measurements, hourly radar rain fall data were used for both indicator covariance estimation and a comparative evaluation. Preliminary results suggest that the indicator cokriging approach is clearly superior to its ordinary kriging counterpart, whereas the indicator kriging approach is not. The improvement is most significant in estimation of light rainfall, but drops off significantly for heavy rainfall. The lack of predictability in spatial estimation of heavy rainfall is borne out in the integral scale of indicator correlation: peaking to its maximum for cutoffs near the median, indicator correlation scale becomes increasingly smaller for larger cutoffs of rainfall depth. A derived-distribution analysis, based on the assumption that radar rainfall is a linear sum of ground-truth and a random error, suggests that, at low cutoffs, indicator correlation scale of ground-truth can significantly differ from that of radar rainfall, and points toward inclusion of rainfall intermittency, for example, within the framework proposed in this work.     

Radar rainfall estimation for flash flood forecasting in small urban watersheds

Radar rainfall estimation for flash flood forecasting in small, urban catchments is examined through analyses of radar, rain gage and discharge observations from the 14.3 km² Dead Run drainage basin in Baltimore County, Maryland. The flash flood forecasting problem pushes the envelope of rainfall estimation to time and space scales that are commensurate with the scales at which the fundamental governing laws of land surface processes are derived. Analyses of radar rainfall estimates are based on volume scan WSR-88D reflectivity observations for 36 storms during the period 2003–2005. Gage-radar analyses show large spatial variability of storm total rainfall over the 14.3 km² basin for flash flood producing storms. The ability to capture the detailed spatial variation of rainfall for flash flood producing storms by WSR-88D rainfall estimates varies markedly from event to event. As spatial scale decreases from the 14.3 km² scale of the Dead Run watershed to 1 km² (and the characteristic time scale of flash flood producing rainfall decreases from 1 h to 15 min) the predictability of flash flood response from WSR-88D rainfall estimates decreases sharply. Storm to storm variability of multiplicative bias in storm total rainfall estimates is a dominant element of the error structure of radar rainfall estimates, and it varies systematically over the warm season and with flood magnitude. Analyses of the 7 July 2004 and 28 June 2005 storms illustrate microphysical and dynamical controls on radar estimation error for extreme flash flood producing storms.     

Hydrological model sensitivity to parameter and radar rainfall estimation uncertainty

Radar estimates of rainfall are being increasingly applied to flood forecasting applications. Errors are inherent both in the process of estimating rainfall from radar and in the modelling of the rainfall–runoff transformation. The study aims at building a framework for the assessment of uncertainty that is consistent with the limitations of the model and data available and that allows a direct quantitative comparison between model predictions obtained by using radar and raingauge rainfall inputs. The study uses radar data from a mountainous region in northern Italy where complex topography amplifies radar errors due to radar beam occlusion and variability of precipitation with height. These errors, together with other error sources, are adjusted by applying a radar rainfall estimation algorithm. Radar rainfall estimates, adjusted and not, are used as an input to TOPMODEL for flood simulation over the Posina catchment (116 km²). Hydrological model parameter uncertainty is explicitly accounted for by use of the GLUE (Generalized Likelihood Uncertainty Estimation). Statistics are proposed to evaluate both the wideness of the uncertainty limits and the percentage of observations which fall within the uncertainty bounds. Results show the critical importance of proper adjustment of radar estimates and the use of radar estimates as close to ground as possible. Uncertainties affecting runoff predictions from adjusted radar data are close to those obtained by using a dense raingauge network, at least for the lowest radar observations available. Copyright © 2004 John Wiley & Sons, Ltd.     

Probabilistic radar rainfall nowcasts using empirical and theoretical uncertainty models

Weather radar has a potential to provide accurate short‐term (0–3 h) forecasts of rainfall (i.e. radar nowcasts), which are of great importance in warnings and risk management for hydro‐meteorological events. However, radar nowcasts are affected by large uncertainties, which are not only linked to limitations in the forecast method but also because of errors in the radar rainfall measurement. The probabilistic quantitative precipitation nowcasting approach attempts to quantify these uncertainties by delivering the forecasts in a probabilistic form. This study implements two forms of probabilistic quantitative precipitation nowcasting for a hilly area in the south of Manchester, namely, the theoretically based scheme [ensemble rainfall forecasts (ERF)‐TN] and the empirically based scheme (ERF‐EM), and explores which one exhibits higher predictive skill. The ERF‐TN scheme generates ensemble forecasts of rainfall in which each ensemble member is determined by the stochastic realisation of a theoretical noise component. The so‐called ERF‐EM scheme proposed and applied for the first time in this study, aims to use an empirically based error model to measure and quantify the combined effect of all the error sources in the radar rainfall forecasts. The essence of the error model is formulated into an empirical relation between the radar rainfall forecasts and the corresponding ‘ground truth’ represented by the rainfall field from rain gauges measurements. The ensemble members generated by the two schemes have been compared with the rain gauge rainfall. The hit rate and the false alarm rate statistics have been computed and combined into relative operating characteristic curves. The comparison of the performance scores for the two schemes shows that the ERF‐EM achieves better performance than the ERF‐TN at 1‐h lead time. The predictive skills of both schemes are almost identical when the lead time increases to 2 h. In addition, the relation between uncertainty in the radar rainfall forecasts and lead time is also investigated by computing the dispersion of the generated ensemble members. Copyright © 2013 John Wiley & Sons, Ltd.     

Validation and use of rainfall radar data to simulate water flows in the Rio Escondido basin

This paper presents a combined validation method of radar-sensed rainfall, using rain gauge data and hydrologic closure, with an application to the Rio Escondido basin (North-East of Mexico). The space–time scaling behavior of rainfall between rain gauge and radar scales is compared with the intrinsic variability of rainfall, for a statistical validation of space–time variability. For hydrological validation purposes, the CEQUEAU model is used to perform rainfall-runoff routing. It provides a basin-wide water balance, to be compared with the measured water flow at the Villa de Fuentes hydrometric station, for mean-value gauging closure. A good qualitative agreement in terms of hydrograph shape and timing is obtained between the simulated and the observed water flows, and a multiplicative correction factor of an initially proposed Z–R relationship is adopted for the watershed under study, which agrees approximately with other authors’ findings about that relationship. The results are considered particularly useful as a validation-and-correction methodology of radar rainfall estimates for areas sparsely covered by rain gauges.     

The use of scale recursive estimation for short term quantitative precipitation forecast

Scale recursive estimation (SRE) is adopted for short term quantitative precipitation forecast (QPF). The precipitation field is modelled using a lognormal random cascade, well suited to properly represent the scaling properties of rainfall fields. To estimate the random cascade parameters, scale recursive maximum likelihood estimation (MLE) is carried out by the iterative expectation maximization (EM) algorithm. To illustrate the potentiality of the SRE, forecast of a synthetically generated rainfall time series is shown. Adaptive estimation of the process parameters is carried out and precipitation forecasts are issued. The forecasts from the SRE are compared with those from standard ARMA models, showing a good performance. The SRE is then adopted for forecasting of an observed half hourly precipitation series for a two day storm event in northern Italy. The SRE provides good performance and it can therefore be adopted as a tool for short term QPF.     

Flood simulation using the gauge‐adjusted radar rainfall and physics‐based distributed hydrologic model

This paper reports the results of an investigation into flood simulation by areal rainfall estimated from the combination of gauged and radar rainfalls and a rainfall–runoff model on the Anseong‐cheon basin in the southern part of Korea. The spatial and temporal characteristics and behaviour of rainfall are analysed using various approaches combining radar and rain gauges: (1) using kriging of the rain gauge alone; (2) using radar data alone; (3) using mean field bias (MFB) of both radar and rain gauges; and (4) using conditional merging technique (CM) of both radar and rain gauges. To evaluate these methods, statistics and hyetograph for rain gauges and radar rainfalls were compared using hourly radar rainfall data from the Imjin‐river, Gangwha, rainfall radar site, Korea. Then, in order to evaluate the performance of flood estimates using different rainfall estimation methods, rainfall–runoff simulation was conducted using the physics‐based distributed hydrologic model, Vflo^?. The flood runoff hydrograph was used to compare the calculated hydrographs with the observed one. Results show that the rainfall field estimated by CM methods improved flood estimates, because it optimally combines rainfall fields representing actual spatial and temporal characteristics of rainfall. Copyright © 2008 John Wiley & Sons, Ltd.     

Estimation of rainfall from infrared‐microwave satellite data for basin‐scale hydrologic modelling

The infrared‐microwave rainfall algorithm (IMRA) was developed for retrieving spatial rainfall from infrared (IR) brightness temperatures (TBs) of satellite sensors to provide supplementary information to the rainfall field, and to decrease the traditional dependency on limited rain gauge data that are point measurements. In IMRA, a SLOPE technique (ST) was developed for discriminating rain/no‐rain pixels through IR image cloud‐top temperature gradient, and 243K as the IR threshold temperature for minimum detectable rainfall rate. IMRA also allows for the adjustment of rainfall derived from IR‐TB using microwave (MW) TBs. In this study, IMRA rainfall estimates were assessed on hourly and daily basis for different spatial scales (4, 12, 20, and 100 km) using NCEP stage IV gauge‐adjusted radar rainfall data, and daily rain gauge data. IMRA was assessed in terms of the accuracy of the rainfall estimates and the basin streamflow simulated by the hydrologic model, Sacramento soil moisture accounting (SAC‐SMA), driven by the rainfall data. The results show that the ST option of IMRA gave accurate satellite rainfall estimates for both light and heavy rainfall systems while the Hessian technique only gave accurate estimates for the convective systems. At daily time step, there was no improvement in IR‐satellite rainfall estimates adjusted with MW TBs. The basin‐scale streamflow simulated by SAC‐SMA driven by satellite rainfall data was marginally better than when SAC‐SMA was driven by rain gauge data, and was similar to the case using radar data, reflecting the potential applications of satellite rainfall in basin‐scale hydrologic modelling. Copyright © 2010 John Wiley & Sons, Ltd.     

Sampling error of areal average rainfall due to radar partial coverage

Areal average rainfall is important as it is used as an input for most rainfall-runoff analysis in Hydrology and Water Resources. Different from traditional methods of using rain gauge data, the use of radar rainfall for the estimation of areal average rainfall is very straightforward. However, in some cases with severe terrain blockages, the value of the incomplete radar information is of serious concern. This study investigated this problem and derived an equation for estimating the error involved in the areal average rainfall due to partial radar coverage of a basin or sub-basin. When only partial radar information is available, the sampling error decreases with increasing radar coverage and the number of radar bin clusters. As an application example, this study considered the Han River Basin with its rainfall observations using the Ganghwa rain radar. Among a total of 24 mid-sized sub-basins in the Han River Basin evaluated, only five sub-basins were fully covered by the radar and three were totally uncovered. The remaining 16 sub-basins were covered partially by radar leading to incomplete radar information. The results show that the sampling error ranged from several % to tens % of standard deviation of the areal average rainfall depending on the relative areal radar coverage.     

Inter-station correlation and estimation errors of areal average rain rate

In this study an equation for estimating the error involved in the areal average rain rate considering the inter-station correlation was derived and applied for two cases: the first compared two storm events with different inter-station correlations, and the second evaluated the seasonal variation of estimation error of monthly rainfall. Similar cases, but without considering the rainfall seasonality, were also investigated for the comparison. This study was applied to the Geum River Basin with 28 rain gauge measurements, each having more than 30 years of rainfall data. A summary of the application results follows: (1) When considering the inter-station correlation, the estimation error involved in the areal average rain rate became significantly decreased proportional to the inter-station correlation. (2) The estimation error of monthly areal average rainfall showed strong seasonality with high ones during the wet season and lower ones during the dry season. (3) The estimation error was well proportional to the areal average rain rate as well as to its standard deviation. The ratio of estimation error to the areal average rain rate itself was estimated to be about 0.1 for the case of assuming no inter-station correlations, but decreased to 0.06 for the case of considering the inter-station correlations between measurements. (4) The relation between the standard deviation of areal average rain rate and the estimation error became much stronger than that between the areal average rain rate itself and the estimation error. The ratio of estimation error to the standard deviations of rain rate amount was estimated to be about 0.2 for the case of assuming no inter-station correlations, but decreased to 0.1 for the case of considering the inter-station correlations. This relation was found to be valid for any case of accumulation time such as in daily, monthly, or annual rainfall data.     

Modelling radar-rainfall estimation uncertainties using elliptical and Archimedean copulas with different marginal distributions

Abstract

Given that radar-based rainfall has been broadly applied in hydrological studies, quantitative modelling of its uncertainty is critically important, as the error of input rainfall is the main source of error in hydrological modelling. Using an ensemble of rainfall estimates is an elegant solution to characterize the uncertainty of radar-based rainfall and its spatial and temporal variability. This paper has fully formulated an ensemble generator for radar precipitation estimation based on the copula method. Each ensemble member is a probable realization that represents the unknown true rainfall field based on the distribution of radar rainfall (RR) error and its spatial error structure. An uncertainty model consisting of a deterministic component and a random error factor is presented based on the distribution of gauge rainfall conditioned on the radar rainfall (GR|RR). Two kinds of copulas (elliptical and Archimedean copulas) are introduced to generate random errors, which are imposed by the deterministic component. The elliptical copulas (e.g. Gaussian and t-copula) generate the random errors based on the multivariate distribution, typically of decomposition of the error correlation matrix using the LU decomposition algorithm. The Archimedean copulas (e.g. Clayton and Gumbel) utilize the conditional dependence between different radar pixels to obtain random errors. Based on those, a case application is carried out in the Brue catchment located in southwest England. The results show that the simulated uncertainty bands of rainfall encompass most of the reference raingauge measurements with good agreement between the simulated and observed spatial dependences. This indicates that the proposed scheme is a statistically reliable method in ensemble radar rainfall generation and is a useful tool for describing radar rainfall uncertainty.

Editor D. Koutsoyiannis; Associate editor S. Grimaldi     

Empirically-based modeling of spatial sampling uncertainties associated with rainfall measurements by rain gauges

In the quantitative evaluation of radar-rainfall products (maps), rain gauge data are generally used as a good approximation of the true ground rainfall. However, rain gauges provide accurate measurements for a specific location, while radar estimates represent areal averages. Because these sampling discrepancies could introduce noise into the comparisons between these two sensors, they need to be accounted for. In this study, the spatial sampling error is defined as the ratio between the measurements by a single rain gauge and the true areal rainfall, defined as the value obtained by averaging the measurements by an adequate number of gauges within a pixel. Using a non-parametric scheme, the authors characterize its full statistical distribution for several spatial (4, 16 and 36 km²) and temporal (15 min and hourly) scales.     

Modeling radar-rainfall estimation uncertainties using parametric and non-parametric approaches

There are large uncertainties associated with radar estimates of rainfall, including systematic errors as well as the random effects from several sources. This study focuses on the modeling of the systematic error component, which can be described mathematically in terms of a conditional expectation function. The authors present two different approaches: non-parametric (kernel-based) and parametric (copula-based). A large sample (more than six years) of rain gauge measurements from a dense network located in south-west England is used as an approximation of the true ground rainfall. These data are complemented with rainfall estimates by a C-band weather radar located at Wardon Hill, which is about 40 km from the catchment. The authors compare the results obtained using the parametric and non-parametric schemes for four temporal scales of hydrologic interest (5 and 15 min, hourly and three-hourly) by means of several different performance indices and discuss the strengths and weaknesses of each approach.     

Spatial reconstruction of rainfall fields from rain gauge and radar data

Rainfall is a phenomenon difficult to model and predict, for the strong spatial and temporal heterogeneity and the presence of many zero values. We deal with hourly rainfall data provided by rain gauges, sparsely distributed on the ground, and radar data available on a fine grid of pixels. Radar data overcome the problem of sparseness of the rain gauge network, but are not reliable for the assessment of rain amounts. In this work we investigate how to calibrate radar measurements via rain gauge data and make spatial predictions for hourly rainfall, by means of Monte Carlo Markov Chain algorithms in a Bayesian hierarchical framework. We use zero-inflated distributions for taking zero-measurements into account. Several models are compared both in terms of data fitting and predictive performances on a set of validation sites. Finally, rainfall fields are reconstructed and standard error estimates at each prediction site are shown via easy-to-read spatial maps.     

A multiplier-based method of generating stochastic areal rainfall from point rainfalls

Catchment modelling for water resources assessment is still mainly based on rain gauge measurements as these are more easily available and cover longer periods than radar and satellite-based measurements. Rain gauges however measure the rain falling on an extremely small proportion of the catchment and the areal rainfall obtained from these point measurements are consequently substantially uncertain. These uncertainties in areal rainfall estimation are generally ignored and the need to assess their impact on catchment modelling and water resources assessment is therefore imperative. A method that stochastically generates daily areal rainfall from point rainfall using multiplicative perturbations as a means of dealing with these uncertainties is developed and tested on the Berg catchment in the Western Cape of South Africa. The differences in areal rainfall obtained by alternately omitting some of the rain gauges are used to obtain a population of plausible multiplicative perturbations. Upper bounds on the applicable perturbations are set to prevent the generation of unrealistically large rainfall and to obtain unbiased stochastic rainfall. The perturbations within the set bounds are then fitted into probability density functions to stochastically generate the perturbations to impose on areal rainfall. By using 100 randomly-initialized calibrations of the AWBM catchment model and Sequent Peak Analysis, the effects of incorporating areal rainfall uncertainties on storage-yield-reliability analysis are assessed. Incorporating rainfall uncertainty is found to reduce the required storage by up to 20%. Rainfall uncertainty also increases flow-duration variability considerably and reduces the median flow-duration values by an average of about 20%.     

Scale-recursive assimilation of precipitation data

This paper presents a new statistical method for assimilating precipitation data from different sensors operating over a range of scales. The technique is based on a scale-recursive estimation algorithm which is computationally efficient and able to account for the nested spatial structure of precipitation fields. The version of the algorithm described here relies on a static multiplicative cascade model which relates rainrates at different scales. Bayesian estimation techniques are used to condition rainrate estimates on measurements. The conditioning process is carried out recursively in two sweeps: first from fine to coarse scales and then from coarse to fine scales. The complete estimation algorithm is similar to a fixed interval smoother although it processes data over scale rather than time. We use this algorithm to assimilate radar and satellite microwave data collected during the tropical ocean–global atmosphere coupled ocean–atmosphere response experiment (TOGA-COARE). The resulting rainrate estimates reproduce withheld radar measurements to within the level of accuracy predicted by the assimilation algorithm.     

A non-parametric automatic blending methodology to estimate rainfall fields from rain gauge and radar data

Quantitative estimation of rainfall fields has been a crucial objective from early studies of the hydrological applications of weather radar. Previous studies have suggested that flow estimations are improved when radar and rain gauge data are combined to estimate input rainfall fields. This paper reports new research carried out in this field. Classical approaches for the selection and fitting of a theoretical correlogram (or semivariogram) model (needed to apply geostatistical estimators) are avoided in this study. Instead, a non-parametric technique based on FFT is used to obtain two-dimensional positive-definite correlograms directly from radar observations, dealing with both the natural anisotropy and the temporal variation of the spatial structure of the rainfall in the estimated fields. Because these correlation maps can be automatically obtained at each time step of a given rainfall event, this technique might easily be used in operational (real-time) applications. This paper describes the development of the non-parametric estimator exploiting the advantages of FFT for the automatic computation of correlograms and provides examples of its application on a case study using six rainfall events. This methodology is applied to three different alternatives to incorporate the radar information (as a secondary variable), and a comparison of performances is provided. In particular, their ability to reproduce in estimated rainfall fields (i) the rain gauge observations (in a cross-validation analysis) and (ii) the spatial patterns of radar fields are analyzed. Results seem to indicate that the methodology of kriging with external drift [KED], in combination with the technique of automatically computing 2-D spatial correlograms, provides merged rainfall fields with good agreement with rain gauges and with the most accurate approach to the spatial tendencies observed in the radar rainfall fields, when compared with other alternatives analyzed.     

Optimal design of rain gauge network in the Middle Yarra River catchment,Australia

Rainfall data are a fundamental input for effective planning, designing and operating of water resources projects. A well‐designed rain gauge network is capable of providing accurate estimates of necessary areal average and/or point rainfall estimates at any desired ungauged location in a catchment. Increasing network density with additional rain gauge stations has been the main underlying criterion in the past to reduce error and uncertainty in rainfall estimates. However, installing and operation of additional stations in a network involves large cost and manpower. Hence, the objective of this study is to design an optimal rain gauge network in the Middle Yarra River catchment in Victoria, Australia. The optimal positioning of additional stations as well as optimally relocating of existing redundant stations using the kriging‐based geostatistical approach was undertaken in this study. Reduction of kriging error was considered as an indicator for optimal spatial positioning of the stations. Daily rainfall records of 1997 (an El Niño year) and 2010 (a La Niña year) were used for the analysis. Ordinary kriging was applied for rainfall data interpolation to estimate the kriging error for the network. The results indicate that significant reduction in the kriging error can be achieved by the optimal spatial positioning of the additional as well as redundant stations. Thus, the obtained optimal rain gauge network is expected to be appropriate for providing high quality rainfall estimates over the catchment. The concept proposed in this study for optimal rain gauge network design through combined use of additional and redundant stations together is equally applicable to any other catchment. © 2014 The Authors. Hydrological Processes published by John Wiley & Sons Ltd.     

Rainfall spatial variability observed by X-band weather radar and its implication for the accuracy of rainfall estimates

The main objective of this paper is to estimate the error in the rainfall derived from a polarimetric X-band radar, by comparison with the corresponding estimate of a rain gauge network. However the present analysis also considers the errors inherent to rain gauge, in particular instrumental and representativeness errors. A special emphasis is addressed to the spatial variability of the rainfall in order to appreciate the representativeness error of the rain gauge with respect to the 1 km square average, typical of the radar derived estimate. For this purpose the spatial correlation function of the rainfall is analyzed.     

Detection of mean-field bias of the radar rain rate using rain gauges available within a small portion of radar umbrella: a case study of the Donghae (East Sea) radar in Korea

Observation of a storm approaching from the ocean to the in-land area is very important for the flood forecasting. Radar is generally used for this purpose. However, as rain gauges are mostly located within the in-land area, detection of the mean-field bias of radar rain rate cannot be easily made. This problem is obviously different from that with evenly-spaced rain gauges over the radar umbrella. This study investigated the detection problem of mean-field bias of radar rain rate when rain gauges are available within a small portion of radar umbrella. To exactly determine the mean-field bias, i.e. the difference between the radar rain rate and the rain gauge rain rate, the variance of the difference between two observations must be small; thus, a sufficient number of observations are indispensable. Therefore, the problem becomes determining the number of rain gauges that will satisfy the given accuracy, that being the variance of the difference between two observations. The dimensionless error variance derived by dividing the expected value of the error variance by the variance of the areal average rain rate was introduced as a criteria to effectively detect the mean field bias. Here, the variance of the areal average rain rate was assumed to be the climatological one and the expectation for the error variance could be changed depending one the sampling characteristics. As an example, this study evaluated the rainfall observation over the East Sea by the Donghae radar. About 6.8 % of the entire radar umbrella covered in-land areas, where the rain gauges were available. It was found that, to limit the dimensionless error variance to 2 %, a total of 26 rain gauges are required for the entire radar umbrella; whereas, a total of 24 rain gauges would be required within the in-land area with available for the rain gauge data.