Using a remote sensing field experiment to investigate flux-footprint relations and flux sampling distributions for tower and aircraft-based observations

During a remote sensing field experiment conducted in the Southern Great Plains in 1997 (SGP97), tower and aircraft-based flux observations were collected over one of the main study sites in central Oklahoma. This is an agricultural region and contains primarily grassland/pasture and winter wheat, which was recently harvested leaving a significant number of fields either as wheat stubble or plowed bare soil. Multi-spectral data obtained by aircraft provided high-resolution (30 m) spatially-distributed vegetation cover and surface temperature information over the study area. The spatial variations in these surface states strongly affect the partitioning of surface fluxes between sensible and latent heat. These data, together with coarser resolution (5 km) satellite data, are used in a remote sensing-based energy balance modeling system that disaggregates flux estimates from 5 km to 30-m resolution. The resulting high-resolution flux maps provide a means for evaluating whether tower and aircraft-based flux measurements sample a full range in flux conditions for this landscape. In addition, this remote sensing-based modeling system can be used to investigate the influence of variability in these key surface states on tower and aircraft measurements through flux-footprint modeling. Under the light wind and unstable conditions that existed during the observations, highest correlation between aircraft and modeled estimated heat and water vapor fluxes were obtained using different flux-footprint estimates. More specifically, the source area for heat was estimated to be much closer to the aircraft flight line than for water vapor.     

Suitability of satellite-based hydro-climate variables and machine learning for streamflow modeling at various scale watersheds

ABSTRACT

Streamflow modeling is essential to investigate processes in the hydrologic cycle and important for water resource management application. However, in-situ hydrologic data paucity, because of various factors such as economic, political, instrument malfunctioning, and poor spatial distribution, makes the modeling process challenging. To overcome this limitation, we introduced a satellite remote sensing-based machine learning approach – boosted regression tree (BRT) – that integrates spatial land surface and climate variables that describe the sub-units, and applied it in three variable size watersheds in the Upper Mississippi River Basin (UMRB), USA. The model simulation results were tested using an independent dataset and showed Nash–Sutcliffe efficiency values of 0.80, 0.76, and 0.69 for the UMRB, Illinois River Watershed, and Raccoon River Watershed, respectively. In addition, we compared the performance of the machine learning models with existing process-based modeling results. Overall performance is comparable with the process-based approaches, but with significantly less modeling effort and resources.     

Regional crop water use assessment using Landsat-derived evapotranspiration

Reliable information on water use and availability at basin and field scales are important to ensure the optimized constructive uses of available water resources. This study was conducted with the specific objective to estimate Landsat-based actual evapotranspiration (ETa) using the Operational Simplified Surface Energy Balance (SSEBop) model across the state of South Dakota (SD), USA for the 1986–2018 (33-year) period. Validated ETa estimations (r² = 0.91, PBIAS = −4%, and %RMSE = 11.8%) were further used to understand the crop water-use characteristics and existing historic mono-directional (increasing/decreasing) trends over the eastern (ESD) and western (WSD) regions of SD. The crop water-use characteristics indicated that the annual cropland water uses across the ESD and WSD were more or less met by the precipitation amounts in the area. The ample water supply and distribution have led to high rainfed and low percentage of irrigated cropland (~2.5%) in the state. The WSD faced greater crop-water use reductions than the ESD during drought periods. The landscape ETa responses across the state were found to be more sensitive than precipitation for the drought impact assessments. The Mann Kendall trend analysis revealed the absence of a significant trend (p > 0.05) in annual ETa at a regional scale due to the varying weather conditions in the state. However, about 12% and 9% cropland areas in the ESD and WSD, respectively, revealed a significant mono-directional trend at pixel scale ETa. Most of the pixels under significant trend showed an increasing trend that can be explained by the shift in agricultural practices, increased irrigated cropland area, higher productions, moisture regime shifts, and decreased risk of farming in the dry areas. The decreasing trend pixels were clustered in mid-eastern SD and could be the result of dynamic conversion of wetlands to croplands and decreased irrigation practices in the region. This study also demonstrates the tremendous potential and robustness of the SSEBop model, Landsat imagery, and remote sensing-based ETa modelling approaches in estimating consistent spatially distributed evapotranspiration.     

Cotton canopy reflectance under variable solar zenith angles: Implications of use in evapotranspiration models

Evapotranspiration (ET) is an important parameter in hydrologic processes and modelling. In agricultural watersheds with competing uses of fresh water including irrigated agriculture, estimating crop evapotranspiration (ET_c) accurately is critical for improving irrigation system and basin water management. The use of remote sensing-based basal crop coefficients is becoming a common method for estimating crop evapotranspiration for multiple crops over large areas. The Normalized Difference Vegetation Index (NDVI) and the Soil Adjusted Vegetation Index (SAVI), based on reflectance in the red and near-infrared bands, are commonly used for this purpose. In this paper, we examine the effects of row crop orientation and soil background darkening due to shading and soil surface wetness on these two vegetation indices through modelling, coupled with a field experiment where canopy reflectance of a cotton crop at different solar zenith angles, was measured with a portable radiometer. The results show that the NDVI is significantly more affected than the SAVI by background shading and soil surface wetness, especially in north–south oriented rows at higher latitudes and could lead to a potential overestimation of crop evapotranspiration and irrigation water demand if used for basal crop coefficient estimation. Relationships between the analysed vegetation indices and canopy biophysical parameters such as crop height, fraction of cover and leaf area index also were developed for both indices.     

Modelling the water balance of a mesoscale catchment basin using remotely sensed land cover data

Hydrological modelling of mesoscale catchments is often adversely affected by a lack of adequate information about specific site conditions. In particular, digital land cover data are available from data sets which were acquired on a European or a national scale. These data sets do not only exhibit a restricted spatial resolution but also a differentiation of crops and impervious areas which is not appropriate to the needs of mesoscale hydrological models. In this paper, the impact of remote sensing data on the reliability of a water balance model is investigated and compared to model results determined on the basis of CORINE (Coordination of Information on the Environment) Land Cover as a reference. The aim is to quantify the improved model performance achieved by an enhanced land cover representation and corresponding model modifications. Making use of medium resolution satellite imagery from SPOT, LANDSAT ETM+ and ASTER, detailed information on land cover, especially agricultural crops and impervious surfaces, was extracted over a 5-year period (2000–2004). Crop-specific evapotranspiration coefficients were derived by using remote sensing data to replace grass reference evapotranspiration necessitated by the use of CORINE land cover for rural areas. For regions classified as settlement or industrial areas, degrees of imperviousness were derived. The data were incorporated into the hydrological model GROWA (large-scale water balance model), which uses an empirical approach combining distributed meteorological data with distributed site parameters to calculate the annual runoff components. Using satellite imagery in combination with runoff data from gauging stations for the years 2000–2004, the actual evapotranspiration calculation in GROWA was methodologically extended by including empirical crop coefficients for actual evapotranspiration calculations. While GROWA originally treated agricultural areas as homogeneous, now a consideration and differentiation of the main crops is possible. The accuracy was determined by runoff measurements from gauging stations. Differences in water balances resulting from the use of remote sensing data as opposed to CORINE were analysed in this study using a representative subcatchment. Resulting Nash–Sutcliff model efficiencies improved from 0.372 to 0.775 and indicate that the enhanced model can produce thematically more accurate and spatially more detailed local water balances. However, the proposed model enhancements by satellite imagery have not exhausted the full potential of water balance modelling, for which a higher temporal resolution is required.     

Assessing strategic water availability using remote sensing,GIS and a spatial water budget model: case study of the Upper Ing Basin,Thailand

Abstract

This paper assesses strategic water availability and use under different development pathways on a basin scale using remote sensing (RS), geographical information systems (GIS) and a spatial water budget model (SWBM). The SWBM was applied to the Upper Ing Basin in northern Thailand to investigate the spatial and temporal variations in the location of streams and water yields from different parts of the basin. The base simulation was carried out for the years 1998–2007 using a DEM and actual land-use data at 100-m resolution. The simulated stream network was compared with topographic maps under different flow conditions, which were successfully represented. The 10-year average simulated river flow rate was 1300 L/s, but it more than doubled during periods of heavy rainfall and decreased below 600 L/s in dry seasons. The total length of the streams (based on flow threshold of 25 L/s) on a typical day in the dry season differed by a factor of approx. 1.5. Agricultural water needs and possible extraction were assessed and presented by dividing the basin into 10 different zones based on the stream network. The results show that there is the potential for harvesting significant quantities of water at different spatial gradients with no initial water supply for irrigation. Monthly water yields for each zone were computed; the results varied from less than 50% to over 137% of the per hectare water yield for the entire basin. This variation was due to differences in topography and land cover. The impact of land use and climate change on streamwater availability was also studied. The basin shows very different hydrological responses. The changes in average river flow relative to the base simulation were +27.6%,??32.1%, +94% and +52.9% under deforestation, changing land use from paddy field to orchard, bare soil and increased rainfall scenarios, respectively.

Citation Bahadur KC, K. (2011) Assessing strategic water availability using remote sensing, GIS and a spatial water budget model: case study of the Upper Ing Basin, Thailand. Hydrol. Sci. J. 56(6), 994–1014.     

Operational applications of remote sensing in hydrology: success,prospects and problems

Abstract

The use of remote sensing information in operational hydrology is relatively limited, but specific examples can be cited for determining precipitation, soil moisture, groundwater, snow, surface water and basin characteristics. The application of remote sensing in hydrology can be termed operational if at least one of two conditions are met: (a) the application produces an output on a regular basis, or (b) the remote sensing data are used regularly on a continuing basis as part of a procedure to solve a problem or make decisions. When surveying the various operational applications, simple approaches and simple remote sensing data sets are the most successful. In the data-sparse developing countries, many operational remote sensing approaches exist (out of necessity) that may not be needed in developed countries because of existing data networks. To increase the use of remote sensing in operational hydrology in developed countries, pilot projects need to be increased and information services must be improved. Increased utilization of GIS to combine remote sensing with other information will promote new products and applications. End user training must be improved by focusing on satellite data processing and manipulation. In developing countries the same improvements are needed plus some more basic ones. There is a need for international monetary assistance to establish long-term remote sensing data, improved database systems and image processing capabilities. There is also the need to set up innovative regional training centres throughout the developing world.     

Physical statistical algorithm for precipitable water vapor inversion on land surface based on multi-source remotely sensed data

Water vapor plays a crucial role in atmospheric processes that act over a wide range of temporal and spatial scales, from global climate to micrometeorology. The determination of water vapor distribution in the atmosphere and its changing pattern is very important. Although atmospheric scientists have developed a variety of means to measure precipitable water vapor(PWV) using remote sensing data that have been widely used, there are some limitations in using one kind satellite measurements for PWV retrieval over land. In this paper, a new algorithm is proposed for retrieving PWV over land by combining different kinds of remote sensing data and it would work well under the cloud weather conditions. The PWV retrieval algorithm based on near infrared data is more suitable to clear sky conditions with high precision. The 23.5 GHz microwave remote sensing data is sensitive to water vapor and powerful in cloud-covered areas because of its longer wavelengths that permit viewing into and through the atmosphere. Therefore, the PWV retrieval results from near infrared data and the indices combined by microwave bands remote sensing data which are sensitive to water vapor will be regressed to generate the equation for PWV retrieval under cloud covered areas. The algorithm developed in this paper has the potential to detect PWV under all weather conditions and makes an excellent complement to PWV retrieved by near infrared data. Different types of surface exert different depolarization effects on surface emissions, which would increase the complexity of the algorithm. In this paper, MODIS surface classification data was used to consider this influence. Compared with the GPS results, the root mean square error of our algorithm is 8 mm for cloud covered area. Regional consistency was found between the results from MODIS and our algorithm. Our algorithm can yield reasonable results on the surfaces covered by cloud where MODIS cannot be used to retrieve PWV.     

Improvement of TOPLATS‐based discharge predictions through assimilation of ERS‐based remotely sensed soil moisture values

In this paper, we investigate the possibility to improve discharge predictions from a lumped hydrological model through assimilation of remotely sensed soil moisture values. Therefore, an algorithm to estimate surface soil moisture values through active microwave remote sensing is developed, bypassing the need to collect in situ ground parameters. The algorithm to estimate soil moisture by use of radar data combines a physically based and an empirical back‐scatter model. This method estimates effective soil roughness parameters, and good estimates of surface soil moisture are provided for bare soils. These remotely sensed soil moisture values over bare soils are then assimilated into a hydrological model using the statistical correction method. The results suggest that it is possible to determine soil moisture values over bare soils from remote sensing observations without the need to collect ground truth data, and that there is potential to improve model‐based discharge predictions through assimilation of these remotely sensed soil moisture values. Copyright © 2002 John Wiley & Sons, Ltd.     

Estimation of the Budyko model parameter for small basins in China

At the mean annual scale, water availability of a basin is substantially determined by how much precipitation will be partitioned into evapotranspiration and run-off. The Budyko framework provides a simple but efficient tool to estimate precipitation partitioning at the basin scale. As one form of the Budyko framework, Fu's equation has been widely used to model long-term basin-scale water balance. The major difficulty in applications of Fu's equation is determining how to estimate the curve shape parameter ω efficiently. Previous studies have suggested that the parameter ω is closely related to the long-term vegetation coverage on large river basins globally. However, on small basins, the parameter ω is difficult to estimate due to the diversity of controlling factors. Here, we focused on the estimation of ω for small basins in China. We identified the major factors controlling the basin-specific (calibrated) ω from nine catchment attributes based on a dataset from 206 small basins (≤50,000 km²) across China. Next, we related the calibrated ω to the major factors controlling ω using two statistical models, that is, the multiple linear regression (MLR) model and artificial neural network (ANN) model. We compared and validated the two statistical models using an independent dataset of 80 small basins. The results indicated that in addition to vegetation, other landscape factors (e.g., topography and human activity) need to be considered to capture the variability of ω on small basins better. Contrary to previous findings reached on large basins worldwide, the basin-specific ω and remote sensing-based vegetation greenness index exhibit a significant negative correlation. Compared with the default ω value of 2.6 used in the Budyko curve method, the two statistical models significantly improved the mean annual ET simulations on validation basins by reducing the root mean square error from 114 mm/year to 74.5 mm/year for the MLR model and 70 mm/year for the ANN model. In comparison, the ANN model can provide a better ω estimation than the MLR model.     

A conceptual–numerical model to simulate hydraulic head in aquifers that are hydraulically connected to surface water bodies

In this paper, we present a conceptual‐numerical model that can be deduced from a calibrated finite difference groundwater‐flow model, which provides a parsimonious approach to simulate and analyze hydraulic heads and surface water body–aquifer interaction for linear aquifers (linear response of head to stresses). The solution of linear groundwater‐flow problems using eigenvalue techniques can be formulated with a simple explicit state equation whose structure shows that the surface water body–aquifer interaction phenomenon can be approached as the drainage of a number of independent linear reservoirs. The hydraulic head field could be also approached by the summation of the head fields, estimated for those reservoirs, defined over the same domain set by the aquifer limits, where the hydraulic head field in each reservoir is proportional to a specific surface (an eigenfunction of an eigenproblem, or an eigenvector in discrete cases). All the parameters and initial conditions of each linear reservoir can be mathematically defined in a univocal way from the calibrated finite difference model, preserving its characteristics (geometry, boundary conditions, hydrodynamic parameters (heterogeneity), and spatial distribution of the stresses). We also demonstrated that, in practical cases, an accurate solution can be obtained with a reduced number of linear reservoirs. The reduced computational cost of these solutions can help to integrate the groundwater component within conjunctive use management models. Conceptual approximation also facilitates understanding of the physical phenomenon and analysis of the factors that influence it. A simple synthetic aquifer has been employed to show how the conceptual model can be built for different spatial discretizations, the parameters required, and their influence on the simulation of hydraulic head fields and stream–aquifer flow exchange variables. A real‐world case was also solved to test the accuracy of the proposed approaches, by comparing its solution with that obtained using finite‐difference MODFLOW code. Copyright © 2011 John Wiley & Sons, Ltd.     

Towards hydrological model calibration and validation: simulation of stable water isotopes using the isoWATFLOOD model

Calibration and validation of hydrological models is a challenge, particularly in remote regions that are minimally gauged. This paper develops a novel methodology for large‐scale (>1000 km²) hydrological model calibration and validation using stable water isotopes founded on the rigorous constraints imposed by the need to conserve both water mass and stable isotopes simultaneously. The isoWATFLOOD model is applied to five basins within the Fort Simpson, Northwest Territories region of northern Canada to simulate stream discharge and oxygen‐18 signals over a 3‐year period. The isotopic variation of river discharge, runoff components, and evaporative fractionation are successfully simulated on both a seasonal and continual basis over the watershed domain to demonstrate the application of isotope tracers to regional hydrologic calibration. The intended application of this research is to remote, large‐scale basins, showing promise for improving predictions in minimally gauged basins and climate change research where traditional, rigorous approaches to constraining parameter uncertainty may be impractical. This coupled isotope‐hydrological (i.e. iso‐hydrological) approach to modelling reduces the number of possible parameterizations, resulting in potentially more physically‐based hydrological predictions. isoWATFLOOD provides a tool for water resource managers and utilities to use operationally for water use, allocation, and runoff generation estimations. Copyright © 2012 John Wiley & Sons, Ltd.     

Recharge estimation for transient ground water modeling

Reliable ground water models require both an accurate physical representation of the system and appropriate boundary conditions. While physical attributes are generally considered static, boundary conditions, such as ground water recharge rates, can be highly variable in both space and time. A practical methodology incorporating the hydrologic model HELP3 in conjunction with a geographic information system was developed to generate a physically based and highly detailed recharge boundary condition for ground water modeling. The approach uses daily precipitation and temperature records in addition to land use/land cover and soils data. The importance of the method in transient ground water modeling is demonstrated by applying it to a MODFLOW modeling study in New Jersey. In addition to improved model calibration, the results from the study clearly indicate the importance of using a physically based and highly detailed recharge boundary condition in ground water quality modeling, where the detailed knowledge of the evolution of the ground water flowpaths is imperative. The simulated water table is within 0.5 m of the observed values using the method, while the water levels can differ by as much as 2 m using uniform recharge conditions. The results also show that the combination of temperature and precipitation plays an important role in the amount and timing of recharge in cooler climates. A sensitivity analysis further reveals that increasing the leaf area index, the evaporative zone depth, or the curve number in the model will result in decreased recharge rates over time, with the curve number having the greatest impact.     

The hydrology of sand rivers in Zimbabwe and the use of remote sensing to assess their level of saturation

Sand rivers are ephemeral watercourses containing sand that are occasionally flooded with rainwater runoff during the rainy season. Although the riverbed appears dry for most of the year, there is perennial groundwater flow within the sand. This water flowing beneath the surface is a valuable resource for local communities; nonetheless our understanding of such river systems is limited. Hence, this paper aims to improve our understanding of the hydrology of sand rivers and to examine the potential use of remote sensing to detect the presence of water in the sand. The relationship between rainfall events and changes in the water level of two sand rivers in the Matabeleland South Province of Zimbabwe was investigated. A lagged relationship was observed for the Manzamnyama River but for the Shashani River the relationship was seen only when considering cumulative rainfall events. The comparison of the modelled flow as simulated by a water balance model with observations revealed the important influence of the effective sediment depth on the recharge and recession of the alluvial channels in addition to the length of the channel. The possibility of detecting water in the alluvial sands was investigated using remote sensing. During the wet season, optical images showed that the presence of water on the riverbed was associated with a smooth signal, as it tends to reflect the incident radiation. A chronological analysis of radar images for different months of the year demonstrates that it is possible to detect the presence of water in the sand rivers. These results are a first step towards the development of a methodology that would aim to use remote sensing to help reducing survey costs by guiding exploratory activities to areas showing signs of water abstraction potential.     

The Representation of Tropospheric Water Vapor Over Low-Latitude Oceans in (Re-)analysis: Errors,Impacts, and the Ability to Exploit Current and Prospective Observations

This paper addresses the representation of lower tropospheric water vapor in the meteorological analyses—fully detailed estimates of atmospheric state—providing the wide temporal and spatial coverage used in many process studies. Analyses are produced in a cycle combining short forecasts from initial conditions with data assimilation that optimally estimates the state of the atmosphere from the previous forecasts and new observations, providing initial conditions for the next set of forecasts. Estimates of water vapor are among the less certain aspects of the state because the quantity poses special challenges for data assimilation while being particularly sensitive to the details of model parameterizations. Over remote tropical oceans observations of water vapor come from two sources: passive observations at microwave or infrared wavelengths that provide relatively strong constraints over large areas on column-integrated moisture but relatively coarse vertical resolution, and occultations of Global Positioning System provide much higher accuracy and vertical resolution but are relatively spatially coarse. Over low-latitude oceans, experiences with two systems suggest that current analyses reproduce much of the large-scale variability in integrated water vapor but have systematic errors in the representation of the boundary layer with compensating errors in the free troposphere; these errors introduce errors of order 10% in radiative heating rates through the free troposphere. New observations, such as might be obtained by future observing systems, improve the estimates of water vapor but this improvement is lost relatively quickly, suggesting that exploiting better observations will require targeted improvements to global forecast models.     

A new saturated/unsaturated model for stormwater infiltration systems

Infiltration systems are widely used as an effective urban stormwater control measure. Most design methods and models roughly approximate the complex physical flow processes in these systems using empirical equations and fixed infiltration rates to calculate emptying times from full. Sophisticated variably saturated flow models are available, but rarely applied owing to their complexity. This paper describes the development and testing of an integrated one‐dimensional model of flow through the porous storage of a typical infiltration system and surrounding soils. The model accounts for the depth in the storage, surrounding soil moisture conditions and the interaction between the storage and surrounding soil. It is a front‐tracking model that innovatively combines a soil‐moisture‐based solution of Richard's equation for unsaturated flow with piston flow through a saturated zone as well as a reservoir equation for flow through a porous storage. This allows the use of a simple non‐iterative numerical solution that can handle ponded infiltration into dry soils. The model is more rigorous than approximate stormwater infiltration system models and could therefore be valuable in everyday practice. A range of test cases commonly used to test soil water flow models for infiltration in unsaturated conditions, drainage from saturation and infiltration under ponded conditions were used to test the model along with an experiment with variable depth in a porous storage over saturated conditions. Results show that the model produces a good fit to the observed data, analytical solutions and Hydrus. Copyright © 2008 John Wiley & Sons, Ltd.     

基于遥感震害指数的尼泊尔地震综合烈度评估研究

2015年4月25日在尼泊尔廓尔喀县发生的8.1级地震及后续强烈余震,造成尼泊尔北部严重的人员伤亡和财产损失,灾区建筑物倒塌损失严重。本文利用现场震害调查资料和高分卫星遥感影像,开展建筑物震害遥感解译,得到各个遥感解译点的遥感震害指数,结合现场调查点评估的烈度拟合了遥感震害指数-实际震害指数转换关系,再根据遥感震害指数估计了全部解译点的震害指数及地震烈度。估计的烈度与现场调查结果对比显示出较好的一致性,研究结果为该地区今后发生地震提供了可借鉴的遥感评估震害指数转换模型。     

Integrating the LISFLOOD‐FP 2D hydrodynamic model with the CAESAR model: implications for modelling landscape evolution

Landscape evolution models (LEMs) simulate the geomorphic development of river basins over long time periods and large space scales (100s–1000s of years, 100s of km²). Due to these scales they have been developed with simple steady flow models that enable long time steps (e.g. years) to be modelled, but not shorter term hydrodynamic effects (e.g. the passage of a flood wave). Nonsteady flow models that incorporate these hydrodynamic effects typically require far shorter time steps (seconds or less) and use more expensive numerical solutions hindering their inclusion in LEMs. The recently developed LISFLOOD‐FP simplified 2D flow model addresses this issue by solving a reduced form of the shallow water equations using a very simple numerical scheme, thus generating a significant increase in computational efficiency over previous hydrodynamic methods. This leads to potential convergence of computational cost between LEMs and hydrodynamic models, and presents an opportunity to combine such schemes. This paper outlines how two such models (the LEM CAESAR and the hydrodynamic model LISFLOOD‐FP) were merged to create the new CAESAR‐Lisflood model, and through a series of preliminary tests shows that using a hydrodynamic model to route flow in an LEM affords many advantages. The new model is fast, computationally efficient and has a stronger physical basis than a previous version of the CAESAR model. For the first time it allows hydrodynamic effects (tidal flows, lake filling, alluvial fans blocking valley floor) to be represented in an LEM, as well as producing noticeably different results to steady flow models. This suggests that the simplification of using steady flow in existing LEMs may bias their findings significantly. Copyright © 2013 John Wiley & Sons, Ltd.     

Towards dynamic catchment modelling: a Bayesian hierarchical mixtures of experts framework

Despite the abundance of existing hydrological models, there is no single model that has been identified as performing consistently over the range of possible catchment types and catchment conditions. An attractive alternative to selecting a single model is to combine the results from several different hydrological models, thereby providing a more appropriate representation of model uncertainty than is the case otherwise. Methods based on Bayesian statistical techniques provide an ideal means to compare and combine competing models, as they explicitly account for model uncertainty. Bayesian model averaging is one such alternative that combines individual models by weighting models proportional to their respective posterior probability of selection. However, the necessity of having fixed weights for each model over the entire length of the simulation period means that the relative usefulness of different models at different times is not considered. The hierarchical mixtures of experts (HME) framework is an appealing extension of the model averaging framework that allows the individual model weights to be estimated dynamically. Consequently, a model more capable at simulating low flow characteristics attains a higher weight (probability) when such conditions are likely, switching over to a lower weight when catchment storage increases. In this way, different models apply in different hydrological states, with the probability of selecting each model being allowed to depend on relevant antecedent condition characteristics. HME models provide additional flexibility compared with simple combinations of models, by allowing the way that model predictions are combined to depend on predictor variables. Thus, for hydrological models, the ‘switch’ from one model to another can depend on the existing catchment condition. This new modelling framework is applied using a simple conceptual model to 10 selected Australian catchments. The study regions are chosen to vary considerably in terms of size, yield and location. Results from this application are compared with the alternative where a single fixed model structure is applied. Comparison of the model simulations using the maximum log‐likelihood and the Nash‐Sutcliffe coefficient of efficiency show that more variance in streamflow was explained by the HME model, compared with the conceptual model alone for each of the catchments investigated. Copyright © 2006 John Wiley & Sons, Ltd.