Search

Reference crop evapotranspiration (ETo) is calculated using a standard formula with temperature, vapor pressure, solar radiation, and wind speed as input variables. ETo forecasts can be produced when forecasts of these input variables from numerical weather prediction (NWP) models are available. As raw ETo forecasts are often subject to systematic errors, statistical calibration is needed for improving forecast quality. The most straightforward and widely used approach is to directly calibrate raw ETo forecasts constructed with the raw forecasts of input variables. However, the predictable signal in ETo forecasts may not be fully implemented by this approach, which does not deal with error propagation from input variables to ETo forecasts. We hypothesize that correcting errors in input variables as a precursor to forecast calibration will lead to more skillful ETo forecasts. To test this hypothesis, we evaluate two calibration strategies that construct raw ETo forecasts with the raw (strategy i) or bias-corrected (strategy ii) input variables in ETo forecast calibration across Australia. Calibrated ETo forecasts based on bias-corrected input variables (strategy ii) demonstrate lower biases, higher correlation coefficients, and higher skills than forecasts produced by the calibration using raw input variables (strategy i). This investigation indicates that improving raw forecasts of input variables could effectively reduce error propagation and enhance ETo forecast calibration. We anticipate that future NWP-based ETo forecasting will benefit from adopting the calibration strategy developed in this study to produce more skillful ETo forecasts.

Monthly Precipitation Forecasts (MPF) play a critical role in drought monitoring, hydrological forecasting and water resources management. In this study, we applied two advanced Machine Learning Models (MLM) and latest General Circulation Models (GCM) to generate deterministic MPFs with a resolution of 0.5° across China. Then the Bayesian Joint Probability (BJP) modeling approach is employed to calibrate and generate corresponding ensemble MPFs. Raw and post-processing MPFs were put against gridded observations over the period of 1981–2015. The results indicated that: (1) for deterministic evaluation, the forecasting performance of MLMs was more inclined to generate random forecasts around the mean value, while the GCMs could reflect the increasing or decreasing trend of precipitation to some degree; (2) for probabilistic evaluation, the four BJP calibrated ensemble MPFs were unbiased and reliable. Compared to climatology, reliability and sharpness were all significantly improved. However, in terms of overall accuracy metric, the ensemble MPFs generated from MLMs were similar to climatology. In contrast, the ensemble MPFs generated from GCMs achieved better forecasting skill and were not dependent on forecasting regions and months. Moreover, the post-processing method is necessary to achieve not only bias-free but also reliable as well as skillful ensemble MPFs. HIGHLIGHTS Two advanced Machine Learning Models are employed to generate monthly precipitation forecasts with a resolution of 0.5 degree.; The latest seasonal forecasts of ECMWF are evaluated with the same forecasting grid cells and lead time.; The BJP modeling approach is used to calibrate above four raw forecasts.; A comprehensive comparison is achieved for the raw and post-processing forecasts.

Deterministic numerical weather prediction (NWP) models and ensemble NWP models are routinely run worldwide to assist weather forecasting. Deterministic forecasts are capable of capturing more detailed spatial features, while ensemble forecasts, often with a coarser resolution, have the ability to predict uncertainty in future conditions. A comparative understanding of the performance of these two types of forecasts is valuable for both users of NWP products and model developers. Past published comparisons tended to be limited in scope, for example, for only specific locations and weather events, and involving only raw forecasts. In this study, we conduct a comprehensive comparison of the performance of a deterministic model and an ensemble model of the Australian Bureau of Meteorology in forecasting daily precipitation across Australia over a period of 3 years. The deterministic model has a horizontal grid spacing of approximately 25 km, and the ensemble model 60 km. Despite the coarser resolution, the ensemble forecasts are found to be superior by a number of measures, including correlation, accuracy and reliability. This finding holds true for both raw forecasts from the NWP models and forecasts post-processed using the recently developed seasonally coherent calibration (SCC) model. Post-processing is shown to greatly improve the forecasts from both models; however, the improvement is greater for the deterministic model, narrowing the performance gap between the two models. This study adds strong evidence to the general notion that coarser-resolution ensemble NWP forecasts perform better than deterministic forecasts.

Concerns have been raised that the use of infrastructure and water efficiency projects to recover water entitlements for the environment in the Murray–Darling Basin (MDB) could be undermined by a r...

A risk assessment of the reduction of streamflow in the Murray-Darling Basin (MDB) from potential increased groundwater extraction has been conducted. This incorporates the uncertainty of future extraction and connectivity between groundwater and surface water. The predicted impact from forty years of growth in extraction is less than 580 Gl/y, and likely to be in range of 100–400 Gl/y. Over 80% of this impact will result from extraction under limits existing before the Basin Plan, with most impact from extraction outside these limits occurring later. Groundwater units with high risk lie within a range of river valleys and hydrogeological domains, and particularly the Goulburn valley and zones of fresher groundwater discharge. Management rules in the new groundwater management plans are designed to reduce use in high impact zones. Monitoring is required to assess effectiveness of these as use increases and conjunctive water management becomes more common.

Calibrated high-temporal-resolution precipitation forecasts are desirable for a range of applications, for example, flood prediction in fast-rising rivers. However, high-temporal-resolution precipitation observations may not be available to support the establishment of calibration methods, particularly in regions with low population density or in developing countries. We present a new method to produce calibrated hourly precipitation ensemble forecasts from daily observations. Precipitation forecasts are taken from a high-resolution convective-scale numerical weather prediction (NWP) model run at the hourly time step. We conduct three experiments to develop the new calibration method: (i) calibrate daily precipitation totals and disaggregate daily forecasts to hourly; (ii) generate pseudohourly observations from daily precipitation observations, and use these to calibrate hourly precipitation forecasts; and (iii) combine aspects of (i) and (ii). In all experiments, we use the existing Bayesian joint probability model to calibrate the forecasts and the well-known Schaake shuffle technique to instill realistic spatial and temporal correlations in the ensembles. As hourly observations are not available, we use hourly patterns from the NWP as the template for the Schaake shuffle. The daily member matching method (DMM), method (iii), produces the best-performing ensemble precipitation forecasts over a range of metrics for forecast accuracy, bias, and reliability. The DMM method performs very similarly to the ideal case where hourly observations are available to calibrate forecasts. Overall, valuable spatial and temporal information from the forecast can be extracted for calibration with daily data, with a slight trade-off between forecast bias and reliability.

Multivariate seasonal climate forecasts are increasingly required for quantitative modeling in support of natural resources management and agriculture. GCM forecasts typically require postprocessing to reduce biases and improve reliability; however, current seasonal postprocessing methods often ignore multivariate dependence. In low-dimensional settings, fully parametric methods may sufficiently model intervariable covariance. On the other hand, empirical ensemble reordering techniques can inject desired multivariate dependence in ensembles from template data after univariate postprocessing. To investigate the best approach for seasonal forecasting, this study develops and tests several strategies for calibrating seasonal GCM forecasts of rainfall, minimum temperature, and maximum temperature with intervariable dependence: 1) simultaneous calibration of multiple climate variables using the Bayesian joint probability modeling approach; 2) univariate BJP calibration coupled with an ensemble reordering method (the Schaake shuffle); and 3) transformation-based quantile mapping, which borrows intervariable dependence from the raw forecasts. Applied to Australian seasonal forecasts from the ECMWF System4 model, univariate calibration paired with empirical ensemble reordering performs best in terms of univariate and multivariate forecast verification metrics, including the energy and variogram scores. However, the performance of empirical ensemble reordering using the Schaake shuffle is influenced by the selection of historical data in constructing a dependence template. Direct multivariate calibration is the second-best method, with its far superior performance in in-sample testing vanishing in cross validation, likely because of insufficient data relative to the number of parameters. The continued development of multivariate forecast calibration methods will support the uptake of seasonal climate forecasts in complex application domains such as agriculture and hydrology.

Statistical postprocessing methods can be used to correct bias and dispersion error in raw ensemble forecasts from numerical weather prediction models. Existing postprocessing models generally perform well when they are assessed on all events, but their performance for extreme events still needs to be investigated. Commonly used joint probability postprocessing models are based on the correlation between forecasts and observations. Because the correlation may be lower for extreme events as a result of larger forecast uncertainty, the dependence between forecasts and observations can be asymmetric with respect to the magnitude of the precipitation. However, the constant correlation coefficient in the traditional joint probability model lacks the flexibility to model asymmetric dependence. In this study, we formulated a new postprocessing model with a decreasing correlation coefficient to characterize asymmetric dependence. We carried out experiments using Global Ensemble Forecast System reforecasts for daily precipitation in the Huai River basin in China. The results show that, although it performs well in terms of continuous ranked probability score or reliability for all events, the traditional joint probability model suffers from overestimation for extreme events defined by the largest 2.5% or 5% of raw forecasts. On the contrary, the proposed variable-correlation model is able to alleviate the overestimation and achieves better reliability for extreme events than the traditional model. The proposed variable-correlation model can be seen as a flexible extension of the traditional joint probability model to improve the performance for extreme events.

Statistical postprocessing models can be used to correct bias and dispersion errors in raw precipitation forecasts from numerical weather prediction models. In this study, we conducted experiments to investigate four factors that influence the performance of regression-based postprocessing models with normalization transformations for short-term precipitation forecasts. The factors are 1) normalization transformations, 2) incorporation of ensemble spread as a predictor in the model, 3) objective function for parameter inference, and 4) two postprocessing schemes, including distributional regression and joint probability models. The experiments on the first three factors are based on variants of a censored regression model with conditional heteroscedasticity (CRCH). For the fourth factor, we compared CRCH as an example of the distributional regression with a joint probability model. The results show that the CRCH with normal quantile transformation (NQT) or power transformation performs better than the CRCH with log–sinh transformation for most of the subbasins in Huai River basin with a subhumid climate. The incorporation of ensemble spread as a predictor in CRCH models can improve forecast skill in our research region at short lead times. The influence of different objective functions (minimum continuous ranked probability score or maximum likelihood) on postprocessed results is limited to a few relatively dry subbasins in the research region. Both the distributional regression and the joint probability models have their advantages, and they are both able to achieve reliable and skillful forecasts.

Calibration and downscaling of ensemble GCM forecasts is becoming increasingly important for hydrological and agricultural modelling in support of the management and protection of valuable natural resources. Moreover, skilful and reliable daily forecast sequences are required to drive decision support models that operate on a daily time step. While downscaling of daily GCM outputs has been developed extensively in climate impacts studies, much less attention has been paid to the downscaling of ensemble GCM forecasts, which has the confronting aspect of low and diminishing skill with increasing lead time. Evidence is building that simple bias‐correction methods that do not model correlation between forecasts and observations, nor attempt to correct spatial, temporal and inter‐variable correlations, produce downscaled forecasts that perform poorly in applications models. Thus downscaling GCM forecasts requires inclusion of a full calibration component to reduce bias, improve reliability, capture skill where it is available and remove negative skill. In this study, we propose a new methodology for coupling a full GCM forecast calibration with empirical methods to (a) instil correct temporal, spatial and inter‐variable correlations in ensembles and (b) perform a multivariate downscaling of ensembles to daily sequences at multiple sites. Through a case study application, the proposed methodology is shown to produce skilful monthly–seasonal forecasts of rainfall, temperature and solar radiation at regional spatial scales. It also produces realistic and coherent multivariate daily sequences at multiple sub‐grid locations. The new methodology can be applied to more effectively integrate climate forecasts into hydrological and crop models and to support proactive decision‐making in agriculture and natural resources management.

Statistical calibration of forecasts from numerical weather prediction (NWP) models aims to produce forecasts that are unbiased, reliable in ensemble spread, and as skillful as possible. We suggest that the calibrated forecasts should also be coherent in climatology, including seasonality, consistent with observations. This is especially important when forecasts approach climatology as forecast skill becomes low, such as at long lead times. However, it is challenging to achieve these aims when data available to establish sophisticated calibration models are limited. Many NWP models have only a short period of archived data, typically one year or less, when they become officially operational. In this paper, we introduce a seasonally coherent calibration (SCC) model for working effectively with limited archived NWP data. Detailed rationale and mathematical formulations are presented. In the development of the model, three issues are resolved. These are 1) constructing a calibration model that is sophisticated enough to allow for seasonal variation in the statistical characteristics of raw forecasts and observations, 2) bringing climatology that is representative of long-term statistics into the calibration model, and 3) reducing the number of model parameters through sensible reparameterization to make the model workable with short NWP dataset. A case study is conducted to examine model assumptions and evaluate model performance. We find that the model assumptions are sound, and the developed SCC model produces well-calibrated forecasts.

Evapotranspiration is one of the most important climate variables for agricultural and environmental management. We investigate the characteristics of daily crop evapotranspiration (ETo) forecasts up to 16 days ahead derived from GCM outputs using the Penman-Monteith Equation. Furthermore, we propose a post-processing method to calibrate the raw ETo forecasts to observations and improve overall reliability and skill. The post-processing method comprises: (1) the Yeo-Johnson transformation to handle the non-normal distribution of daily ETo, (2) the bi-variate normal distribution to formulate the relationship between transformed raw forecasts and observations, and (3) the Schaake shuffle to establish realistic temporal behaviour in each ensemble member. For three distinct Australian case-study locations, ETo forecasts derived from raw ACCESS-S1 outputs are positively correlated with observations up to 16 days ahead. However, raw forecasts verify more poorly than naive climatology forecasts because of significant bias and poor ensemble spreads. Post-processing the raw ACCESS-S1 ETo improves skill and reliability markedly. Skilful daily ETo forecasts are achieved in the first week, and the skill of cumulative ETo forecasts extends beyond two weeks into the future. The proposed ETo post-processing method could be applied to support operational, multi-week forecasting of short-term ETo in support of the day-to-day decisions of irrigators in Australia and elsewhere.

Recent research demonstrates that dynamical models sometimes fail to represent observed teleconnection patterns associated with predictable modes of climate variability. As a result, model forecast skill may be reduced. We address this gap in skill through the application of a Bayesian postprocessing technique—the calibration, bridging, and merging (CBaM) method—which previously has been shown to improve probabilistic seasonal forecast skill over Australia. Calibration models developed from dynamical model reforecasts and observations are employed to statistically correct dynamical model forecasts. Bridging models use dynamical model forecasts of relevant climate modes (e.g., ENSO) as predictors of remote temperature and precipitation. Bridging and calibration models are first developed separately using Bayesian joint probability modeling and then merged using Bayesian model averaging to yield an optimal forecast. We apply CBaM to seasonal forecasts of North American 2-m temperature and precipitation from the North American Multimodel Ensemble (NMME) hindcast. Bridging is done using the model-predicted Niño-3.4 index. Overall, the fully merged CBaM forecasts achieve higher Brier skill scores and better reliability compared to raw NMME forecasts. Bridging enhances forecast skill for individual NMME member model forecasts of temperature, but does not result in significant improvements in precipitation forecast skill, possibly because the models of the NMME better represent the ENSO–precipitation teleconnection pattern compared to the ENSO–temperature pattern. These results demonstrate the potential utility of the CBaM method to improve seasonal forecast skill over North America.

Long-range forecasts of climatic variables are generated by climate centres around the world using global climate models (GCMs). This paper investigates ensemble forecasting of reference crop evapotranspiration (ETo) based on GCM outputs. The Penman-Monteith formula is used to calculate raw forecasts of ETo from GCM forecasts of solar radiation, temperature, wind speed, and vapor pressure. The Bayesian joint probability (BJP) modelling approach is applied to post-process raw monthly forecasts, separately for different lead times (month 1, 2 and 3 ahead). The Schaake shuffle is then employed to link the ensemble members of post-processed forecasts for all lead times to give a temporal structure. Forecasts of seasonal ETo total are obtained by aggregating the monthly forecasts. For comparison purposes, seasonal forecasts are also derived directly by post-processing raw seasonal forecasts without going through the monthly steps. Three case studies are presented for post-processing forecasts from the Australian Community Climate and Earth System Simulator-Seasonal (ACCESS-S1). Both raw forecasts and observations of monthly and seasonal ETo are found to be reasonably normally distributed. The post-processed forecasts of monthly and seasonal ETo are skilful in reference to climatology forecasts and statistically reliable in ensemble spread. The indirect and direct ways of generating forecasts of seasonal ETo total show similar skill and reliability, demonstrating the effectiveness of the Schaake shuffle. In this paper, the proposed post-processing method is evaluated through leave-one-out cross validation. The method can be easily adapted for post-processing raw GCM forecasts in real-time to produce ensemble forecasts of monthly and seasonal ETo.

Accurate and reliable seasonal climate forecasts are frequently sought by climate‐sensitive sectors to support decision‐making under climate variability and change. Temperature trend is discernible globally over the past decades, but seasonal forecasts produced by a global climate model (GCM) generally underestimate such trend. Current statistical methods used for calibrating seasonal climate forecasts mostly do not explicitly account for climate trends. Consequently, the calibrated forecasts also fail to capture the observed trend. Solving this problem can enhance user confidence in seasonal climate forecasts. In this study, we extend the capability of the Bayesian joint probability (BJP) modelling approach for statistical calibration of seasonal climate forecasts. A trend component is introduced into the BJP algorithm for embedding the observed trend into calibrated ensemble forecasts. We apply the new model (named BJP‐t) to three test stations in Australia. Seasonal forecasts of daily maximum temperatures from the SEAS5 model, operated by the European Centre for Medium‐Range Weather Forecasts (ECMWF), are calibrated and evaluated. The BJP‐t calibrated ensemble forecasts can reproduce the observed trend, when the raw ensemble forecasts and the BJP calibrated ensemble forecasts both fail to do so. The BJP‐t calibration leads to more skilful, more reliable and sharper forecasts than the BJP calibration.

The high spatio-temporal variability of precipitation is often difficult to characterise due to limited measurements. The available low-resolution global reanalysis datasets inadequately represent the spatio-temporal variability of precipitation relevant to catchment hydrology. The Bureau of Meteorology Atmospheric high-resolution Regional Reanalysis for Australia (BARRA) provides a high-resolution atmospheric reanalysis dataset across the Australasian region. For hydrometeorological applications, however, it is essential to properly evaluate the sub-daily precipitation from this reanalysis. In this regard, this paper evaluates the sub-daily precipitation from BARRA for a period of 6 years (2010–2015) over Australia against point observations and blended radar products. We utilise a range of existing and bespoke metrics for evaluation at point and spatial scales. We examine bias in quantile estimates and spatial displacement of sub-daily rainfall at a point scale. At a spatial scale, we use the Fractions Skill Score as a spatial evaluation metric. The results show that the performance of BARRA precipitation depends on spatial location with poorer performance in tropical relative to temperate regions. A possible spatial displacement during large rainfall is also found at point locations. This displacement, evaluated by comparing the distribution of rainfall within a day, could be quantified by considering the neighbourhood grids. On spatial evaluation, hourly precipitation from BARRA are found to be skilful at a spatial scale of less than 100 km (150 km) for a threshold of 75 % quantile (90 % quantile) at most of the locations. The performance across all the metrics improves significantly at time resolutions higher than 3 h. Our evaluations illustrate that the BARRA precipitation, despite discernible spatial displacements, serves as a useful dataset for Australia, especially at sub-daily resolutions. Users of BARRA are recommended to properly account for possible spatio-temporal displacement errors, especially for applications where the spatial and temporal characteristics of rainfall are deemed very important.

Streamflow modelling results from the GR4H and PDM hydrological models were evaluated in two Australian sub-catchments, using (1) calibration to streamflow and (2) joint-calibration to streamflow and soil moisture. Soil moisture storage in the models was evaluated against soil moisture observations from field measurements. The PDM had the best performance in terms of both streamflow and soil moisture estimations during the calibration period, but was outperformed by GR4H during validation. It was also shown that the soil moisture estimation was improved significantly by joint-calibration for the case where streamflow and soil moisture estimations were poor. In other cases, addition of the soil moisture constraint did not degrade the results. Consequently, it is recommended that GR4H be used, in preference to the PDM, in the foothills of the Murrumbidgee catchment or other Australian catchments with semi-arid to sub-humid climate, and that soil moisture data be used in the calibration process.

Timely and skilful seasonal streamflow forecasts are used by water managers in many regions of the world for seasonal water allocation outlooks for irrigators, reservoir operations, environmental flow management, water markets and drought response strategies. In Australia, the Bayesian joint probability (BJP) statistical approach has been deployed by the Australian Bureau of Meteorology to provide seasonal streamflow forecasts across the country since 2010. Here we assess the BJP approach, using antecedent conditions and climate indices as predictors, to produce Kharif season (April–September) streamflow forecasts for inflow to Pakistan's two largest upper Indus Basin (UIB) water supply dams, Tarbela (on the Indus) and Mangla (on the Jhelum). For Mangla, we compare these BJP forecasts to (i) ensemble streamflow predictions (ESPs) from the snowmelt runoff model (SRM) and (ii) a hybrid approach using the BJP with SRM–ESP forecast means as an additional predictor. For Tarbela, we only assess BJP forecasts using antecedent and climate predictors as we did not have access to SRM for this location. Cross validation of the streamflow forecasts shows that the BJP approach using two predictors (March flow and an El Niño Southern Oscillation, ENSO, climate index) provides skilful probabilistic forecasts that are reliable in uncertainty spread for both Mangla and Tarbela. For Mangla, the SRM approach leads to forecasts that exhibit some bias and are unreliable in uncertainty spread, and the hybrid approach does not result in better forecast skill. Skill levels for Kharif (April–September), early Kharif (April–June) and late Kharif (July–September) BJP forecasts vary between the two locations. Forecasts for Mangla show high skill for early Kharif and moderate skill for all Kharif and late Kharif, whereas forecasts for Tarbela also show moderate skill for all Kharif and late Kharif, but low skill for early Kharif. The BJP approach is simple to apply, with small input data requirements and automated calibration and forecast generation. It offers a tool for rapid deployment at many locations across the UIB to provide probabilistic seasonal streamflow forecasts that can inform Pakistan's basin water management.

Rainfall forecasts are an integral part of hydrological forecasting systems at sub-seasonal to seasonal timescales. In seasonal forecasting, global climate models (GCMs) are now the go-to source for rainfall forecasts. For hydrological applications however, GCM forecasts are often biased and unreliable in uncertainty spread, and calibration is therefore required before use. There are sophisticated statistical techniques for calibrating monthly and seasonal aggregations of the forecasts. However, calibration of seasonal forecasts at the daily time step typically uses very simple statistical methods or climate analogue methods. These methods generally lack the sophistication to achieve unbiased, reliable and coherent forecasts of daily amounts and seasonal accumulated totals. In this study, we propose and evaluate a Rainfall Post-Processing method for Seasonal forecasts (RPP-S), which is based on the Bayesian joint probability modelling approach for calibrating daily forecasts and the Schaake Shuffle for connecting the daily ensemble members of different lead times. We apply the method to post-process ACCESS-S forecasts for 12 perennial and ephemeral catchments across Australia and for 12 initialisation dates. RPP-S significantly reduces bias in raw forecasts and improves both skill and reliability. RPP-S forecasts are also more skilful and reliable than forecasts derived from ACCESS-S forecasts that have been post-processed using quantile mapping, especially for monthly and seasonal accumulations. Several opportunities to improve the robustness and skill of RPP-S are identified. The new RPP-S post-processed forecasts will be used in ensemble sub-seasonal to seasonal streamflow applications.

Following the development of Error Reduction and Representation In Stages (ERRIS) for daily streamflow forecasting, we extend ERRIS to streamflow forecasting at an hourly time step (ERRIS-h). ERRIS applies a staged error model to reduce errors in hydrological simulations and to quantify prediction uncertainty. ERRIS produces probabilistic predictions, and is capable of propagating errors through multiple lead times to generate ensemble traces. In this study, we identify the need to model the residual distribution differently for rising and falling limbs of hydrographs when applying ERRIS to hourly streamflow forecasting. To address this need, ERRIS-h uses different distribution parameters for the two limbs. We evaluate ERRIS-h on eight rivers in Australia. Hourly streamflow simulations are produced by forcing an initialized GR4H hourly rainfall-runoff model with observed rainfall. We apply ERRIS-h to the streamflow simulations to produce ensemble streamflow predictions with lead times up to 48 h. The ensemble streamflow predictions here can be viewed as forecasts when rainfall forecasts are perfect. In this way, we test the ability of ERRIS-h to update forecasts using the most up-to-date streamflow observations and to generate ensemble traces that reflect hydrological uncertainty. As expected, ERRIS-h is highly effective when applied to the zeroth lead time, dramatically reducing errors in the original GR4H simulations and reliably describing forecast uncertainty. We also show that ERRIS-h ensemble forecasts have smaller errors than deterministic simulations at all lead times and are reliable in ensemble spread even at 48 h lead times.

Traditional approaches to inundation modelling are computationally intensive and thus not well suited to assessing the uncertainty involved in estimating flood inundation surfaces for planning, design and forecasting purposes. In this study, a rapid flood inundation modelling framework is developed, consisting of a novel spatial reduction and reconstruction (SRR) approach and a deep learning (DL) modelling component. The SRR approach is developed to reduce computational cost by identifying representative locations of inundation surfaces where water levels are simulated using DL models, and to efficiently reconstruct inundation surfaces based on simulated water level information. The DL model includes a built-in input selection layer to simplify the model development process, and a Long Short-Term Memory layer for time series modelling. The accuracy and efficiency of the SRR-DL framework is assessed by application to a real-world river system where the inundation of over 3 million grid cells can be simulated in 4 s.

Ephemeral rivers pose considerable challenges for hydrological forecasters. A staged error model is devised for hourly streamflow forecasting with emphasis on ephemeral rivers. Apart from key components such as bias-correction, autoregressive updating and mixed Gaussian residual distributions, the model treats both observed and simulated zero flows as censored data. This assumption requires additional novel parameter estimation and ensemble generation methods. When inferring error model parameters by maximum likelihood estimation (MLE), the likelihood is derived to account for simulated and observed streamflow taking zero or non-zero values using data censoring. MLE is trialed to be replaced by least-squares-after-transformation (LST) estimation wherever possible to ease computational burdens. When generating ensemble forecasts, a data augmentation is used approach to recover censored data. The staged error model is applied to 18 ephemeral and 11 perennial rivers in Australia and generate hourly streamflow forecasts up to 10 days. LST leads to similar or even marginally better forecast performance than MLE by checking with various verification metrics, including bias, continuous ranked probability score, the α -index and the average width of confidence intervals. Skillful and reliable ensemble forecasts can be generated in 25 out of 29 catchments even at extended lead times. However, propagation of uncertainty can sometimes be challenging in four highly ephemeral catchments, leading to unreliable forecasts at long (e.g. > 40 h) lead times. Prospects for addressing this issue in future is discussed.

Post-processing outputs from Coupled Global Circulation Models (CGCM) is required to generate skillful and reliable sub-seasonal precipitation forecasts. However, it is not currently known at what spatiotemporal scale can sub-seasonal precipitation forecasts be improved after post-processing. In this study, a Bayesian joint probability (BJP) method is used to post-process European Centre for Medium-Range Weather Forecasts (ECMWF) sub-seasonal precipitation forecasts at various spatiotemporal scales across China during the boreal summer monsoon. The relationships between forecasts and observations are built at daily, weekly, and fortnight temporal scales for each grid cell and each hydroclimatic region after normalization. The results suggest that the skills and reliability of ECMWF raw daily precipitation forecasts are improved after post-processing. The forecast skills are further improved at larger spatial scales or longer temporal scales, especially in the Upper and Middle Yangtze River, Southeast Rivers, and Pearl River. Although the forecast skills decrease rapidly as lead time increases, the uncertainty spread of the BJP calibrated sub-seasonal precipitation forecasts is highly reliable at all lead times and spatiotemporal scales. This work will help to realize the potential of CGCM forecasts at sub-seasonal time scales. Future works will investigate the benefits of post-processing precipitation forecasts for sub-seasonal streamflow predictions.

Study region Australia. Study focus We developed the Seasonally Coherent Calibration (SCC) model to post-process precipitation forecasts from numerical weather prediction (NWP) models. The SCC model is capable of generating high-quality calibrated forecasts which are coherent in climatology, including seasonality, consistent with observations, despite only a short period of forecasts being available. In this study, we post-process deterministic precipitation forecasts from the Australian Bureau of Meteorology’s Australian Community Climate and Earth System Simulator G2 version (ACCESS-G2) NWP model to generate calibrated ensemble forecasts across Australia. We evaluate the effectiveness, robustness, and computational feasibility of forecast post-processing with the SCC model at a high spatial resolution across the continental scale. New hydrological insights for the region Inadequate representation of forecast seasonality in calibration models poses a challenge for the calibration of NWP forecasts. Through this investigation, the statistical assumptions and reparameterization algorithms of the SCC model are confirmed to be effective and robust in addressing the challenge. Calibrated precipitation forecasts could be used to support operational water balance forecasting in Australia. Post-processing of NWP precipitation forecasts with the SCC model will also benefit a broad range of forecast users in Australia by providing accurate, skillful, and reliable ensemble forecasts. Computation strategies adopted in this study will help develop effective procedures for future post-processing of NWP forecasts with sophisticated calibration models at a large spatial scale.

Despite an increasing availability of skilful long-range streamflow forecasts, many water agencies still rely on simple resampled historical inflow sequences (stochastic scenarios) to plan operations over the coming year. We assess a recently developed forecasting system called forecast guided stochastic scenarios (FoGSS) as a skilful alternative to standard stochastic scenarios for the Australian continent. FoGSS uses climate forecasts from a coupled ocean–land–atmosphere prediction system, post-processed with the method of calibration, bridging and merging. Ensemble rainfall forecasts force a monthly rainfall–runoff model, while a staged hydrological error model quantifies and propagates hydrological forecast uncertainty through forecast lead times. FoGSS is able to generate ensemble streamflow forecasts in the form of monthly time series to a 12-month forecast horizon. FoGSS is tested on 63 Australian catchments that cover a wide range of climates, including 21 ephemeral rivers. In all perennial and many ephemeral catchments, FoGSS provides an effective alternative to resampled historical inflow sequences. FoGSS generally produces skilful forecasts at shorter lead times (

A study of spatial and temporal variability of precipitation isotope composition on the southern and north slopes of Wushaoling Mountain was conducted in order to explore the processes influencing its evolution. The analysis indicated that the isotopic composition, the slopes and intercepts of Local Meteroic Water Lines, altitude gradients and temperature effect are higher on the north slope than those on the south slope. The d-excess showed an increase from lower to higher altitudes, and the altitude gradients changed with season. The correlation coefficients between δ18O and d-excess decreased with increasing altitude due to weakening sub-cloud evaporation. Westerly wind principally dominates Wushaoling Mountain, so the relatively negative stable isotope values observed are related to the long distance transportation of water vapor in spring and winter. In summer and autumn, the locally strong sub-cloud evaporation cause relatively higher δ18O and lower d-excess. The results suggested that the sub-cloud evaporation has enriched the δ18O composition by 23%, 23%, 32%, 42% and 29% in May, June, July, August and September, respectively. In some circumstances, δ18O and δD were depleted at the end of multi-days rainfall events due to the rainout process. In addition, monsoonal moisture caused some negative δ18O in summer when an enhanced cyclonic circulation had developed on Tibetan Plateau. The study enhances the knowledge of isotopic evolution of precipitation and provides a basis for further study of isotopic hydrology in arid regions.

In contrast to deterministic forecasts, ensemble forecasts are a multiple forecasts of the same events. The ensemble forecasts are generated by perturbing uncertain factors such as model forcings, initial conditions, and/or model physics.

Few ensemble streamflow forecasting systems are designed to operate for ephemeral rivers. In this study, we revise our error model for generating Forecast Guided Stochastic Scenarios (FoGSS) to produce statistically reliable long-range (12-month) forecasts for ephemeral rivers. FoGSS features an error model with four stages: data transformation, bias-correction, an autoregressive error model and the statistical distribution of residuals. We revise the fourth stage of FoGSS with a parameter estimation method that uses data censoring to account for zero values in both observations and forecasts. This allows FoGSS to produce statistically reliable ensemble forecasts in even highly ephemeral streams (with >50% zero flows). We apply FoGSS to conventional ensemble hydrological prediction (ESP) forecasts for 50 Australian catchments, including 26 ephemeral rivers. We show that FoGSS improves the accuracy of ESP forecasts at short lead times, while at long lead times FoGSS forecasts transition to climatology-like forecasts. FoGSS forecasts are reliable in ensemble spread at individual lead times and for volumes aggregated over lead times, even in highly ephemeral rivers. FoGSS forecasts pave the way for operational long-range forecasts in ephemeral rivers, meeting a key need for improved water management.

Changes in the hydrologic cycle have far reaching impacts on agricultural productivity, water resources availability, riverine ecosystems, and our ability to manage environmental assets, bushfire risk, and flood hazard. For example, declining rainfall in the southeast of Australia has led to a prolonged period of drought, with serious impacts on agriculture, the environment, and water supply to urban and rural towns. Here, using the continental wide Australian Water Resources Assessment Landscape model (AWRA-L), we evaluate historical trends from 1960 to 2017 in rainfall, soil moisture, evapotranspiration, and runoff to explain changing drought and flooding. Northern parts of Australia have experienced increasing annual rainfall totals, resulting in increased water availability in the tropics with increased soil moisture, evapotranspiration, and runoff, particularly during the hot, wet monsoon season. In contrast, the southwest and southeast coast of Australia have experienced declines in rainfall, particularly in the colder months, corresponding with decreasing evapotranspiration, soil moisture, and runoff. Trends in flooding are aligned with runoff trends, and closely follow trends in rainfall, with changes in soil moisture of secondary influence. Streamflow droughts, measured by the standardised runoff index, are increasing across large parts of Australia, with these increases more widespread than changes in rainfall alone. Increases in rainfall in the tropics of northern Australia appear to be related to decreasing drought occurrence and extent, but this trend is not universal, suggesting changes in rainfall alone are not an indicator of changing drought conditions.

Flood inundation models are important tools in flood management. Commonly used flood inundation models, such as hydrodynamic or simplified conceptual models, are either computationally intensive or cannot simulate the temporal behavior of floods. Therefore, emulation models based on data-driven methods, such as artificial neural networks (ANNs), have been developed. However, the performance of ANN models, like any other data-driven models, is limited by available data and will not perform well in data-sparse regions. In this study, we developed an ANN-based hybrid modeling approach to improve model performance in data-sparse regions by leveraging better model performance in data-rich regions. We applied our proposed hybrid modeling approach with three ANN models, including the traditional point-based ANN and two newly proposed block-based ANN models. The results demonstrate that all three ANN models have better performance in data-rich regions compared to data-sparse regions as expected, with the block-based ANN with the most complicated model structure performing better in data-rich regions and the simplest point-based ANN performing better in data-sparse regions. The hybrid modeling approach can significantly improve model performance in data-sparse regions, with the hybrid model based on the most complex block-based ANN performing the best. Our results show the importance of considering the trade-offs between data availability and model complexity in developing data-driven models, and demonstrate the potential for improving performance in data-sparse regions by using a hybrid modeling approach that optimizes model complexity based on data availability.

Moisture recycling by terrestrial evaporation and transpiration has recently been confirmed as an important source of precipitation, but little is known of this contribution in inland river basins of China. This study determines the fractions contributed by terrestrial evaporation and transpiration to precipitation in the Shiyang river basin, located in Gansu province of northwestern China. The basin has an area of 4.16 × 104 km2 and mean annual precipitation of 300 mm/yr. Hundreds of samples of precipitation, surface water, plant stem water and soil water were collected and analyzed for their isotopic compositions. The Craig-Gordon model and a three-end-member mixing model were used to partition precipitation into water sourced from evaporation, transpiration and advection. On average, evaporation, transpiration and advection were responsible for 9%, 14% and 77%, respectively, of precipitation, and the contribution from terrestrial evaporation and transpiration also increased with elavation; they also varied with season, being highest in August. The significant contribution from transpiration highlights the importance of vegetation conservation in this ecologically fragile basin.

Statistical seasonal forecasts of 3-month streamflow totals are released in Australia by the Bureau of Meteorology and updated on a monthly basis. The forecasts are often released in the second week of the forecast period, due to the onerous forecast production process. The current service relies on models built using data for complete calendar months, meaning the forecast production process cannot begin until the first day of the forecast period. Somehow, the bureau needs to transition to a service that provides forecasts before the beginning of the forecast period; timelier forecast release will become critical as sub-seasonal (monthly) forecasts are developed. Increasing the forecast lead time to one month ahead is not considered a viable option for Australian catchments that typically lack any predictability associated with snowmelt. The bureau's forecasts are built around Bayesian joint probability models that have antecedent streamflow, rainfall and climate indices as predictors. In this study, we adapt the modelling approach so that forecasts have any number of days of lead time. Daily streamflow and sea surface temperatures are used to develop predictors based on 28-day sliding windows. Forecasts are produced for 23 forecast locations with 0–14- and 21-day lead time. The forecasts are assessed in terms of continuous ranked probability score (CRPS) skill score and reliability metrics. CRPS skill scores, on average, reduce monotonically with increase in days of lead time, although both positive and negative differences are observed. Considering only skilful forecast locations, CRPS skill scores at 7-day lead time are reduced on average by 4 percentage points, with differences largely contained within +5 to −15 percentage points. A flexible forecasting system that allows for any number of days of lead time could benefit Australian seasonal streamflow forecast users by allowing more time for forecasts to be disseminated, comprehended and made use of prior to the commencement of a forecast season. The system would allow for forecasts to be updated if necessary.

Under a warming climate, the cryosphere shrinking is accelerating and the evopotranspiration is strengthening, which have caused the spatial and temporal changes of water resources and water cycle in inland river basins. With a vast area of 1.0 × 104 km2 and an annual average evopotranspiration of 515 mm, the influence from recycling moisture to precipitation and the contribution from cryosphere meltwater to runoff have been quantified in source region of Heihe river basin at the central Qilian Mountains, where 365 glaciers locate within an area of 77.22 km2, whiles frozen soil accounts for 80% of the region. Results indicated that frozen soil meltwater and glacier snow meltwater have contributed by 28% and 7%, on average, to the outlet river water in the basin, respectively. It was founded that evaporation and transpiration moisture were responsible for 10% and 17%, on average, of local precipitation, respectively. These findings provide new progresses on isotopic hydrology of cold basin, which will strengthen further understanding on the role of frozen soil meltwater and local moisture recycling in the water cycle for inland river basins.

Assessment of forecast precipitation is required before it can be used as input to hydrological models. Using radar observations in southeastern Australia, forecast rainfall from the Australian Community Climate Earth-System Simulator (ACCESS) was evaluated for 2010 and 2011. Radar rain intensities were first calibrated to gauge rainfall data from four research rainfall stations at hourly time steps. It is shown that the Australian ACCESS model (ACCESS-A) overestimated rainfall in low precipitation areas and underestimated elevated accumulations in high rainfall areas. The forecast errors were found to be dependent on the rainfall magnitude. Since the cumulative rainfall observations varied across the area and through the year, the relative error (RE) in the forecasts varied considerably with space and time, such that there was no consistent bias across the study area. Moreover, further analysis indicated that both location and magnitude errors were the main sources of forecast uncertainties on ho...

There are a number of challenges that must be overcome if GCM forecasts are to be widely adopted in climate-sensitive industries such as agriculture and water management. GCM outputs are frequently biased relative to observations and their ensembles are unreliable in conveying uncertainty through appropriate spread. The calibration, bridging, and merging (CBaM) method has been shown to be an effective tool for postprocessing GCM rainfall forecasts to improve ensemble forecast attributes. In this study, CBaM is modified and extended to postprocess seasonal minimum and maximum temperature forecasts from the POAMA GCM in Australia. Calibration is postprocessing GCM forecasts using a statistical model. Bridging is producing additional forecasts using statistical models that have other GCM output variables (e.g., SST) as predictors. It is demonstrated that merging calibration and bridging forecasts through CBaM effectively improves the skill of POAMA seasonal minimum and maximum temperature forecasts for Australia. It is demonstrated that CBaM produces bias-corrected forecasts that are reliable in ensemble spread and reduces forecasts to climatology when there is no evidence of forecasting skill. This work will help enable the adoption of GCM forecasts by climate-sensitive industries for quantitative modeling and decision-making.

The Schaake shuffle is a simple and effective method for re-ordering calibrated ensemble forecasts. It is widely used in forecast calibration methods where realistic spatial and temporal sequences are important. We illustrate a previously unidentified problem with the application of the Schaake shuffle. When the autocorrelation of uncalibrated forecasts is markedly different from observations, the Schaake shuffle cannot guarantee that the calibrated ensemble is reliable when ensemble members are accumulated through time. Accumulations in time and space are particularly important for applications that integrate rainfall over these dimensions, notably hydrological modelling. We demonstrate that ensemble means of uncalibrated forecasts tend to be more autocorrelated than observations. This can cause poor reliability if variables are accumulated across lead times under certain conditions, even if forecasts are perfectly reliable at individual lead times. Specifically, if the following conditions occur, ensemble predictions of accumulated variables tend to be too wide: 1) Forecasts are more autocorrelated than observations. 2) Forecasts are skillful; i.e., cross-correlations between forecasts and observations are high. We show with a real-world case study that these conditions can readily occur. We discuss potential solutions to this issue.

Errors from hydrological simulations have substantial influence on hydrological applications. There are increasing interests to incorporate statistical models of the errors (error models) into hydrological applications to improve simulations. However, conventional error models are usually temporally unvaried and may be inadequate to handle model simulation errors when the underlying error statistics are strongly seasonal. To overcome this problem, the use of temporally varied error models in both prediction and forecasting applications are investigated. We analyze different levels of temporal granularity to identify the optimal temporally varied error model to best improve simulations and to quantify uncertainty. The Error Reduction and Representation In Stages (ERRIS) model is adapted in this study to improve streamflow simulations produced by the Variable Infiltration Capacity model for the Yarlung Tsangbo River basin. Well-marked seasonal and sub-seasonal variations in error statistics are found. Accordingly, three temporally varied ERRIS models are constructed at semi-annual (ERRIS-H), seasonal (ERRIS-S) and monthly (ERRIS-M) temporal granularity and compared with a benchmark model, the temporally unvaried model (ERRIS-A). Results show that the temporally varied ERRIS models are considerably more effective than the temporally unvaried one, with 34% reduction in continuous ranked probability score (CRPS) and 23% increase in Nash-Sutcliffe Efficiency (NSE) for prediction applications. With respect to forecasting applications, improvements of about 7% in CRPS are achieved by the temporally varied models. The performance of different temporally varied error models roughly follows the same order as the level of temporal granularity. Generally, ERRIS-S and ERRIS-M are similarly effective, with ERRIS-M providing additional improvement by more than 15% in CRPS for the spring season and by at least 30% for the autumn season. It is concluded that the consistency of temporal granularity between error models and variations of error statistics is the key to effective error reduction and uncertainty quantification. When complicated hydrological processes exist, hydrological model tends to produce seasonally and even sub-seasonally varied error statistics. Consequently, temporally finer error models are expected to lead to higher accuracy and reliability.

Available high quality precipitation datasets generally cover only recent periods. To extend these datasets back in time is challenging due to historically lower rain gauge network density or poorer remote sensing. In this study, a Bayesian joint probability method is presented to extend the temporal coverage of high quality precipitation datasets by calibrating reanalysis estimates. Relationships between precipitation estimates from a high quality dataset and precipitation estimates from a reanalysis dataset are established by using data from the overlapping period of the two datasets. The relationships are then used to calibrate reanalysis estimates for the period when only reanalysis data is available. Rain gauge observations are also used to enhance the calibration. The method brings the reanalysis dataset to the same spatial resolution as the high quality dataset, corrects bias, and makes the reanalysis data more consistent with the high quality dataset. The calibrated estimates generated are presented in the form of ensembles to represent uncertainty. The method is applied to the Han River basin in China, using the CMPA high quality dataset and the ERA-interim reanalysis dataset. The method is shown to be highly effective in improving the quality of ERA-Interim.

Hydrodynamic models are commonly used to understand flood risk and inform flood management decisions. However, their high computational cost can impose practical limits on real-time flood forecasting and uncertainty analysis which require fast modelling response or many model runs. Emulation models have the potential to reduce simulation times while still maintaining acceptable accuracy of the estimates. In this study, we propose an artificial neural networks (ANNs) based emulation modelling framework for flood inundation modelling. We investigate the suitability of ANNs as flood inundation models using a river segment in Queensland, Australia. Our results show that ANNs can model the time series behaviour of flood inundation and significantly reduce the simulation times required, which facilitates their use in applications requiring fast model response or a large number of model runs. Based the model development process and results, the major challenges and future research directions are discussed.

Rainfall forecasts are an integral part of hydrological forecasting systems at sub-seasonal to seasonal time scales. In seasonal forecasting, global climate models (GCMs) are now the go-to source for rainfall forecasts. However, for hydrological applications, GCM forecasts are often biased and unreliable in uncertainty spread, and therefore calibration is required before use. There are sophisticated statistical techniques for calibrating monthly and seasonal aggregations of the forecasts. However, calibration of seasonal forecasts at the daily time step typically uses very simple statistical methods or climate analogue methods. These methods generally lack the sophistication to achieve unbiased, reliable and coherent forecasts of daily amounts and seasonal accumulated totals. In this study, we propose and evaluate a Rainfall Post-Processing method for Seasonal forecasts (RPP-S) based on the Bayesian joint probability approach for calibrating daily forecasts and the Schaake Shuffle approach for connecting the daily ensemble members of different lead times. We apply the method to post-process ACCESS-S forecasts for 12 perennial and ephemeral catchments across Australia and for 12 initialisation dates. RPP-S significantly reduces bias in raw forecasts and improves both skill and reliability. RPP-S forecasts are more skilful and reliable than forecasts derived from ACCESS-S forecasts that have been post-processed using quantile mapping, especially for monthly and seasonal accumulations. Several opportunities to improve the robustness and skill of RPP-S are identified. The new RPP-S post-processed forecasts will be used in ensemble sub-seasonal to seasonal streamflow applications.

International audience; GCMs are used by many national weather services to produce seasonal outlooks of atmospheric and oceanic conditions and fluxes. Postprocessing is often a necessary step before GCM forecasts can be applied in practice. Quantile mapping (QM) is rapidly becoming the method of choice by operational agencies to postprocess raw GCM outputs. The authors investigate whether QM is appropriate for this task. Ensemble forecast postprocessing methods should aim to 1) correct bias, 2) ensure forecasts are reliable in ensemble spread, and 3) guarantee forecasts are at least as skillful as climatology, a property called ‘‘coherence.’’ This study evaluates the effectiveness of QM in achieving these aims by applying it to precipitation forecasts from the POAMA model. It is shown that while QM is highly effective in correcting bias, it cannot ensure reliability in forecast ensemble spread or guarantee coherence. This is because QM ignores the correlation between raw ensemble forecasts and observations. When raw forecasts are not significantly positively correlated with observations, QM tends to produce negatively skillful forecasts. Even when there is significant positive correlation, QM cannot ensure reliability and coherence for postprocessed forecasts. Therefore, QM is not a fully satisfactory method for post- processing forecasts where the issues of bias, reliability, and coherence pre-exist. Alternative post- processing methods based on ensemble model output statistics (EMOS) are available that achieve not only unbiased but also reliable and coherent forecasts. This is shown with one such alternative, the Bayesian joint probability modeling approach