PUB 2011 Posters
Michael Allchin: Mapmatics / Wallingford HydroSolutions
Flow Regime Estimation in UK Ungauged Basins
In the UK, water resources managers often use flow-duration curves as their preferred representation of river flow regimes. Past research has shown that the gradient of a flow-duration curve on a log-normal plot, with flows expressed as a proportion of the mean flow, is similar between catchments of different sizes if their physical characteristics are comparable, and that these may be described in terms of their composition of Hydrology Of Soil Types (HOST), classes. Using this relationship, the annual and monthly natural flow-duration curves at an ungauged site may be estimated by identifying a sub-set of gauged catchments in a reference pool which have the most similar HOST representation (in terms of euclidean offsets for each class, weighted according to their relative importance in controlling runoff) to the target catchment, and then applying appropriate weighting and averaging to their associated flow-duration data. This initial estimate of the curve, in which flows are expressed relative to their mean, is scaled using an annual mean flow for the target catchment derived from a kilometric-resolution runoff grid, itself generated from a regression-based rainfall-runoff model.
If a target-point is located downstream of any gauges for which reliable long-term and essentially natural flow records are available, then these may be employed to reduce the overall uncertainty of the estimate for the ungauged site, by deducting the gauged sub-catchment from the target basin, estimating the flow-regime for the ungauged portion, and adding the gauged flow data to the result. Similar - but reversed - logic is applied to incorporate data from any suitably proximal downstream gauge. Once the natural curve has been estimated, the impact of anthropogenic influences such as abstractions (from surface-water or wells), discharges and impoundments may be incorporated.
This methodology, and a number of associated functions, has been implemented by Mapmatics on behalf of Wallingford HydroSolutions in a software system named LowFlows, now in use by the England & Wales Environment Agency, Scottish Environmental Protection Agency, and others, as a strategic toolset to support water resources regulation and planning.
Dr Matthias Bernhardt: Assistant Professor, LMU Munich, Germany
Zugspitze: Germany’s highest mountain transforms from ungauged to gauged
The vast majority of high mountain areas can be recognized as ungauged and inappropriate for operational in situ monitoring programs. Hence, test sites where model concepts can be tested and improved are sparse. One central aim of the environmental research station Schneefernerhaus at Zugspitze (Germany) is to overcome this lack of information and to make information about the hydrology of an Alpine catchment available for the research community. Zugspitze massif it is built up by limestone with a karst aquifer in the underground. It is therefore representative for wide areas of the European Alps where the rainfall-runoff characteristics are usually complicated because of unknown flow paths. This makes a proper assignment of a runoff measured at a certain gauge to a definable catchment area difficult if not even completely impossible. Zugspitze again has the advantage of a known over- and underground catchment area. Different tracer experiments have indicated that the observed area drains exclusively to a single spring. This fact can be used for a delineation of the specific possibilities and restrictions of modeling tools and modules with respect to the water balance in Alpine karst areas. The available meteorological data at Zugspitze comprises long term measurements from 1900 on and detailed measurements of a denser network from the 1980’s on. The installation of a new gauging station will complete the catchment and can transform it to a possible benchmark area for models which are usually used in similar but ungauged areas.
Mr Jagat K. Bhusal: Member Secretary, International Hydrological Programme, NATCOM, Nepal
Hydrological Prediction in Nepal
Nepal is a mountainous country having several rivers with rapids and falls, ice cold waters and remoteness. Rivers are not monitored as required. Estimates are made mostly by regional approaches. Presently, government has adopted privatising policy in hydropower development. Newly emerging private developers are needing river flows for many ungauged rivers. So far there are not any tested models in Nepalese rivers. However, there are services available but not competent to addess latest knowledge on hydrological sciences.
This paper highlights methodology currently being used in Nepal to assess hydrological parameters for different purposes - hydropower to environmental needs with some illustrations.
Dr Sarah Boon: Assistant Professor, University of Lethbridge
Effect of forest litter on snow surface radiative fluxes
Sub-canopy snow melt is driven by the radiation budget at the snow surface, which is controlled by both canopy cover and snow surface albedo. Snow albedo is affected by snow grain size, snow depth, solar zenith angle, and incident diffuse radiation. Albedo is also affected by the presence of impurities, such as forest litter, within and on the pack. Forest litter is a common impurity that is particularly important in regions impacted by forest disturbance, such as insect infestation, wildfire, or disease, where rates of litter productions are accelerated due to stand mortality. We outline a series of field and experimental approaches to quantify the effects of forest litter on snow surface albedo, and the subsequent implications for snow melt rates and timing. Results suggest that litter may play a significant role in enhancing post-disturbance snow melt, and must be accurately parameterized in numerical models of post-disturbance snow-dominated watersheds.
Pedro Luiz Borges Chaffe: PhD Student, School of Engineering, Kyoto University, Japan
Snowmelt observation and modeling in the Ane river basin, Japan
The Ane River Basin is located in the Northeast region of Shiga Prefecture, Japan, and it is the largest contributing basin to Lake Biwa. The objective of the present work was to develop and compare simple snowmelt model formulations using snow and meteorological data from the Ane River Basin. This type of simple temperature index point studies might be helpful for further implementation of spatially distributed models or for validation of their accuracy. Three different approaches for simulating snowmelt were adopted. One based on a previously developed and tested degree-hour method for the Surumi site using 2001-2002 data. The other methods were the temperature-index method and an enhanced temperature-index method (which considers radiation). The models were calibrated considering different time resolution (10 minutes, hourly and daily). The results show that the empirically derived base model using hourly time steps had a good agreement with the data. The Enhanced Temperature-index compared to the other models showed little to no improvement.
Cornelia Barth, Douglas P. Boyle, Scott Bassett, and Chris Garner: Department of Geography, University of Nevada
Towards Improved Hydrologic Model Predictions in Ungauged Snow-Dominated Watersheds Utilizing a Multi-Criteria Approach and SNODAS Estimates of SWE
In many of the mountainous watersheds in the western United States, a majority of the annual streamflow runoff originates as melt water from snow. Hydrologic models of these regions typically have components that represent the snow water equivalent (SWE) throughout the accumulation and depletion processes of the snowpack; however, obtaining accurate estimates of the spatial and temporal distribution of SWE is a challenge due to the limited number of point observations of SWE. Thus, the calibration of these models generally focuses on fitting simulated streamflow to observed streamflow data. In this study, we are investigating the use of SWE estimates obtained from the Snow Data Assimilation System (SNODAS) product as a surrogate for observation data to perform a multi-objective, multi-behavior sensitivity analysis and calibration of the Precipitation-Runoff Modeling System (PRMS). One of the more interesting results, highlighted in this poster presentation, is that when the PRMS model is calibrated using only the SNODAS SWE estimates (i.e., the streamflow observations are not used in the calibration process) the resulting simulation of streamflow compares reasonably well with the observed streamflow values. This suggests that the SNODAS estimates of SWE may contain information that, in the absences of any streamflow observations, may be very useful for improving model predictions in ungauged basins.
Chris DeBeer: PhD Student, Centre for Hydrology, University of Saskatchewan
Improving model representation of alpine snowmelt runoff genereration
Snowmelt runoff from headwater basins in the Rocky Mountains represents a valuable water resource in western Canada, providing a significant amount of the flow to major river systems in the prairies. Realistically simulating the timing and magnitude of meltwater generation and runoff using hydrological models applied to these basins is difficult however, due to the large spatial variability of energy inputs and pre-melt snowcover. This variability controls the rate of areal snowcover depletion (SCD) and the contributing area of snowmelt runoff, thereby affecting the rate, magnitude, location, and duration of meltwater generation. An approach to represent this combined variability was developed and applied to a small (~1.2 km2) alpine basin in the Front Ranges of the Canadian Rocky Mountains. Binned SWE distributions, stratified by terrain units of relatively homogenous slope and aspect, were used to define the initial pre-melt snowcover variability. Snowmelt was simulated over these distributions using a point scale model (Snobal energy balance model) within the Cold Regions Hydrological Model platform, after correcting for the effects of slope, aspect, skyview, and elevation on incident shortwave and longwave radiation and air temperature. This allowed the differential evolution of the internal energetics and snowpack state, and snowmelt timing and rate to be effectively represented over the landscape and over the cold and highly redistributed snowcover. A simple hydrological model was developed, which accounted for infiltration to frozen soils, soil moisture balance, snow interception and radiation attenuation by the forest canopy, groundwater recharge, and routing of the meltwater through the basin. The results showed that hydrograph simulations based on a single melt rate, spatially uniform SWE, or both failed to adequately represent the timing and magnitude of flow, while the approach developed here was capable of reasonably reproducing the observed hydrograph. This is useful because the approach retains model simplicity and physical integrity, whilst avoiding the need for complex, highly parameterized, fully distributed simulations.
Dr Chad Ellis: Centre for Hydrology, University of Saskatchewan
Forest harvesting impacts on mountain snowmelt
Utilising a physically-based modelling approach built upon extensive field observations of radiation dynamics and snow processes in cold regions forest environments, the impacts of prescribed forest harvesting treatments on spring snowmelt were investigated in a headwater basin of the eastern Canadian Rocky Mountains. Both field observations and model simulations show that relative to open sites, canopy sublimation losses are capable of reducing forest snow accumulation by over half, which acts to substantially increase total spring snowmelt from forested mountain landscapes patterned with small clear-cuts. However, in terms of snowmelt timing, forest-cover impacts were clearly dependent on slope orientation, acting to delay snowmelt on south-facing landscapes through reduced shortwave radiation melt energy, while advancing melt on north-facing landscapes via longwave radiation enhancements from canopy emissions. Consequently, the removal of forest-cover across opposing north-facing and south-facing mountain landscapes acted to greatly extend the spring melt period, and illustrates the important hydrological control of forest-cover through its synchronisation of snowmelt across complex terrain. Such results demonstrate a considerable potential for altering the magnitude and timing of spring snowmelt runoff through targeted forest harvesting practises across these mountain regions.
Dr Anil Gupta, Mr Werner Herrera: Regional Hydrologists, Alberta Environment
Prediction of Near Real Time Natural Flows in Gauged and Ungauged Watershed
Knowledge of natural water supplies (i.e. how much water is naturally available, the time when it is available and its variability) is important for planning and operational purposes including regulatory aspects such as licensing and compliance. Quantification of natural water supplies is often hampered by lack of hydrological, climatic and other relevant data. High costs of installation and maintenance of hydrometric stations, particularly in remote regions, contribute to the scarcity of observed flow time series required for hydrological characterization of the basin. Hydrological models and tools are therefore employed to generate these time series and estimate flow characteristics at ungauged sites. Continuous natural daily flow time-series at gauged or ungauged sites can be generated either by deterministic rainfall-runoff models or by making use of spatial interpolation and regionalization. These two methods may significantly enhance the capabilities of quantifying water availability and can be added to the quantitative toolbox to support Cumulative Effects Management System and Alberta Water for Life initiatives. Hydrological model that operates at a spatially-distributed level using physical properties that are both land-use based and non-land-use based is expected to produce more detailed and potentially more accurate results compared to a model that operates at lumped level. However, such models require adequate quantification of multiple model parameters, reliable climate inputs, and normally represent a time consuming, labor-intensive approach. The inclusion of such models as predictive tools has been identified as part of the long-term research objectives. Applied research in many parts of the world suggest that spatial interpolation and regionalization methods using Flow Duration Curves (called Pragmatic Approach) would offer an initial, parsimonious approach for simulating natural flow regimes at gauged and ungauged sites. However, this approach may be limited for predicting flows in intermittent/ephemeral streams; therefore, the Province of Alberta is pursuing a pilot project to test the applicability of a fully distributed hydrologic model capable of coupling surface-ground water interaction to enhance the capability of predicting flows in intermittent/ephemeral streams. This poster describes the development and application of two very different hydrologic modeling approaches for prediction of natural flows in gauged and ungauged streams representing two extreme ends of hydrologic modeling spectrum – from a simple pragmatic approach based hydrological model to a very complex spatially-distributed, physically-based hydrological model.
Prof Denis Hughes: Director, Institute for Water Research, Rhodes University, South Africa
Water Resources Uncertainty Assessment in South Africa
A project on including uncertainty in water resources modelling and assessment has recently been completed. The outcomes include revised methods of rainfall-runoff parameter estimation (including uncertainty) for gauged and ungauged basins, regional hydrological indices to constrain the ensemble outputs and an assessment of links to water resources yield models. Part of the project focused on the practical application of uncertainty in standard water resources assessment methodology used within South Africa coupled with the effects on decision-making risk analysis. The poster will summarise the uncertainty framework that has been developed and illustrate the results.
Annelen Kahl: PhD Student, University of California at Santa Barbara
Reconstruction of SWE at subpixel resolution
The spatial heterogeneity of snow water equivalent (SWE) affects the timing and magnitude of daily and annual melt and consequently a basin’s stream flow. Heterogeneous snow cover smoothes daily runoff, and areas of greater accumulation produce quasi- riparian zones of increased soil moisture that persist well into the summer. A grid cell with spatial heterogeneity in SWE will develop patchy snow coverage during the melt season. On one hand this lowers the surface to volume ratio of the snowpack and decreases melt, on the other hand exposed ground increases sensible heat flux to the remaining snow and dust lowers the snow albedo to further increase melt from solar radiation.
We combine an energy balance snowmelt model, Isnobal, with fractional snow-covered area (fSCA) derived from MODIS imagery, to estimate SWE heterogeneity retrospectively throughout the melt season for a 152km2 basin on the west side of the Sierra Nevada. The temporal resolution of MODIS imagery allows us to observe the depletion of fSCA daily and makes it possible to compute the distribution of SWE within each 500m MODIS pixel. In addition to average SWE values, the degree of heterogeneity intrinsic to each pixel can be derived and we investigate how it is correlated with landscape characteristics.
Julie Kiang: Hydrologist, USGS
U.S. Streamgage Network Analysis - Gaps for Streamflow Prediction
The U.S. Geological Survey (USGS), in cooperation with the U.S. Fish and Wildlife Service, is conducting a streamgage network analysis to understand the utility of the current USGS streamgage network to support estimation of daily streamflow at ungaged or partially-gaged sites. The analysis will identify spatial and temporal gaps in the network using GIS analysis. In addition, the analysis will assess the ability of the existing streamgage network to support techniques for transferring streamflow information from gaged locations to ungaged locations. Basin similarity is needed to apply these techniques and is often assessed using correlation or by comparing watershed attributes between two locations. Cross-correlations of streamflows are being calculated and mapped for reference streamgages across the entire United States, including Alaska and Hawaii, to show regions where high correlations can be found. The distribution of watershed attributes in all watersheds throughout the U.S. is being compared to the distribution of these attributes in watersheds upstream of USGS streamgages to determine if the watershed attributes represented by the current streamgage network is representative of ungaged watersheds across the United States.
Dr Stefan Kienzle: Associate Professor of Hydrology and GIS, University of Lethbridge
ACRU simulation of the effects of climate change on streamflow in the upper North Saskatchewan River Basin
The ACRU agro-hydrological modeling system provided the framework, containing code to simulate all major hydrological processes, including actual evapotranspiration estimates, to simulate the impacts of climate change in the Cline River watershed, Alberta, Canada, under historical (1961-1990) and a range of future climate conditions (2010-2039, 2040-2069, and 2070-2099). Whilst uncertainties in the estimation of many hydrological variables were inevitable, verification analyses carried out for the historical baseline period resulted in good to very good simulations of a range of hydrological processes, including daily air temperature, snow water equivalent and streamflow. Five climate change scenarios were selected to cover the range of possible future climate conditions. In order to generate future climate time series, the 30-year baseline time series was perturbed according to predicted changes in air temperature and p 1 recipitation. Projected increases in air temperature and precipitation resulted in mean annual increases in potential and actual evapotranspiration, groundwater recharge, soil moisture, and streamflow in the Cline River watershed. Increases in both high and low flow magnitudes and frequencies, and large increases to winter and spring streamflow are predicted for all climate scenarios. Spring runoff and peak streamflow were simulated to occur up to four weeks earlier than in the 1961-1990 baseline period. Predicted changes were simulated to progressively increase into the future. A clear shift in the future hydrological regime is predicted, with significantly higher streamflow between October and June, and lower streamflow in July to September.
Nina Köplin: PhD student, Hydrology Group, Dept. of Geography, University of Bern
A method to quantify hydrological change for mesoscale catchments
A nationwide water resources management under scenarios of climate change should be based on quantitative information about an associated hydrological change. These comprehensive data, however, are not available for Switzerland at the moment. The study presented here, aims at establishing and implementing a method to quantitatively assess potential impacts of climate change on hydrological systems in Switzerland.We use the hydrological modelling system PREVAH (Viviroli et al. 2009a), a semi-distributed and conceptual yet process-oriented model. It runs on the basis of hourly meteorological input and at a spatial resolution of 500 x 500 m2. We calibrated 162 catchments and extended our hydrological database to ungauged Alpine regions subsequently. We parameterized those ungauged catchments using a regionalization scheme by Viviroli et al. (2009b) that combines three different regionalization methods by computing the median of their respective simulations. To a total of 205 catchments thus parameterized, we apply climate scenarios of expected changes (deltas) in the annual cycle of temperature and precipitation. The deltas between the control (1980–2009) and the scenario periods (2021–2050/near future, 2070–2099/far future) are provided on a daily basis for each meteorological station site in Switzerland (Bosshard et al. 2011). For every site, a total of 10 model chains from the ENSEMBLES-project (van der Linden & Mitchell 2009) were analyzed. All scenarios are based on the A1B emission scenario, and the differences between the 10 different model chains represent model uncertainty.
First results indicate that the hydrological change in the near future period is unclear due to climate model uncertainty. The far future period exhibits a clearer delta signal that leads to a clearer hydrological signal, too. Temperature seems to be the determining factor in high Alpine and climate-sensitive catchments affecting the ratio of solid to fluid precipitation and the amount and timing of snow and glacier melt, for example. This substantiates the need to regionalize the model parameters, as some of the most sensitive catchments could not have been studied, otherwise. Apart from specific hydrological responses depending e.g. on the catchments’ mean elevation or geographic region, an overall tendency towards decreasing summer runoff and slightly increasing runoff during the rest of the year could be observed for the majority of the catchments considered.
Prof K Kumaraswamy: Professor and Head, Department of Geography, Bharathidasan University, Tiruchirappalli, India
Precision Agriculture : A Solution to Hungry States
Extensive, Intensive and several other forms of agriculture have been tried around the world to solve the world's hunger problem, but in vain. Precision agriculture, a multi-directional agricultural solution is found to be the bsck-bone of Second Green revolution, especially for tropical rain-fed agrarian societies. The case study demonstrates the utility of such a dimension to water resource scientists to look at the problem in a different perspective.
Dr Ian Littlewood: Editor Hydrology Research / IGLEnvironment
Don’t forget – uncertainty comprises imprecision AND inaccuracy
Conceptual rainfall–streamflow model parameters for responsive gauged catchments, calibrated from daily data, can be massively inaccurate even when they have good precision. This source of uncertainty is poorly recognised, eg in model parameter regionalisation schemes for estimating flows at ungauged sites. The poster illustrates the problem and how it might be overcome, thereby contributing to the objective of PUB to reduce predictive uncertainty.
Professor Suxia Liu: Institute of Geographic Sciences and Natural Resources Research, CAS
Advances of PUB research by IGSNRR, CAS, China
The advances of the PUB research done by Institute of Geographic Sciences and Natural Resources Research, IGSNRR), CAS, China have been reviewed, including estimation of annual runoff and peak flow in ungauged basins, prediction, estimation analysis of soil moisture in ungauged basins, estimation of environmental flow in ungauged basins and distribution of hydrlogical elements in ungauged basins.
Dr James McPhee: Assistant Professor, Dept of Civil Engineering, Universidad de Chile
Hydrologic Modeling in Ungauged Countries - Not Just an Interesting Problem
In many regions of the world, economic activity develops at a faster pace than the regulatory framework and public infrastructure required to adequately evaluate its costs and benefits. In Chile, a perceived energy crisis, plus extremely favorable market conditions, have driven the hydropower and mining sectors in the past few years. New projects as well as expansion of current activities are proposed in areas where little is known both in terms of natural hydrologic conditions and with respect to the interaction between hydrology and ecosystems. This work summarizes some of the major challenges facing a country where pressure on hydrologic systems far exceeds the capacity of regulatory agencies, and highlights some of the key hydrology and water resources questions in need of robust and reliable assessment.
Dr Michele Minihane: Post-Doc, University of Washington
Estimating historic streamflow in the Rovuma River (Mozambique & Tanzania)
In the absence of in-situ measurements, historic mean monthly flows in the Rovuma River are estimated by combining simple index gauge methods with monthly historical composite runoff estimates from the University of New Hampshire / Global Runoff Data Center. The effectiveness of the methodology is evaluated by comparing results with those from a short measurement record in a tributary of the Rovuma. The streamflow estimates compare favorably with estimates from a macro-hydrological model with a 0.25 decimal degree spatial resolution. While there is significant uncertainty associated with a lack of in-situ observations, these results provide a starting point for discussing water resources planning and future assessment of changes in water resource availability under climate uncertainty.
Dr Scott Munro: Professor of Geography, University of Toronto Mississauga
Creating a Runoff Record for an Ungauged Basin: Peyto Glacier, 2002-07
Peyto Glacier basin runoff was gauged by the Water Survey of Canada during the 1965-74 International Hydrological Decade (IHD), but the gauge was removed soon after. In 1989 the first year-round automatic weather station was installed in the basin, the scope and quality of sensors improving over the years, so that hourly records of solar radiation, air temperature, humidity, wind speed and precipitation have been available since 2002. These data are the forcing function of a distributed basin model that generates potential runoff due to snow and ice melt for each element of a 25 m resolution grid. Basin runoff is the aggregate of yield from storage in each grid element of the grid, using a delay constant of 100 h for snow, 11.25 h for ice, where ice cover increases over the melt season. An interesting feature of model development is the use of runoff measurement data from the IHD archive to find a suitable delay constant for ice, a value that is consistent with the ~0.5 d used for ice by other modellers. Also of interest is recent supraglacial runoff work on Peyto, for which a 10.5 h delay constant is obtained, thus raising the possibility of significant storage in the weathering crust of the ice surface. Past association of runoff measurements with weather data pointed to the growing importance of solar radiation as the melt season progressed, air temperature being important throughout. Similar associations are evident for runoff estimates from this model.
Zoe Robson: Environmental Advisor, Nexen Inc.
Predicted Water Availability for Shale Gas Exploration in Northeastern BC
Nexen Inc. is currently exploring for natural gas in the shale formations of the Horn River Basin in Northeastern British Columbia. In support of our commitment to environmentally and socially responsible development, we are currently pursuing a long-term water license under the BC Water Act. This will replace prescriptive temporary use permits and allow for the management of surface water withdrawals with respect to environmental flow needs of the basin.
As part of this license application, a flow model was developed for the Tsea River basin in 2008. No published data is available for this area and flow estimates are based on Water Survey of Canada hydrometric records and climate stations maintained throughout the region. In 2009 and 2010, a monitoring program was implemented to record lake levels, local precipitation and stream discharge from several streams and rivers in the headwaters of the Tsea River. Presented here is the comparison between predicted and recorded flow volumes in a data-poor context.
Dr Karsten Schulz: Professor, Department of Geography, LMU Munich, Germany
Regionalization of eco-hydrological processes
To capture the spatial and temporal variability of (here) the gross primary production as a key component of the global carbon cycle, the light use efficiency modeling approach in combination with remote sensing data has shown to be well suited. Typically, the model parameters such as the maximum light use efficiency are either set to a universal constant or to land class dependent values stored in look-up tables. We exploit the machine learning technique support vector regression to explicitly relate the model parameters of a light use efficiency model calibrated at several FLUXNET sites to specific characteristics obtained by meteorological, ecological and remote sensing data. An automatic procedure which selects the relevant characteristics leads to an individual set of features for each parameter. The extrapolation scheme is evaluated with a cross-validation approach which shows the methodology to be well suited to recapture the variability of gross primary production across the study sites. The promising outcomes suggest a further validation with a larger data set of FLUXNET sites.
Prof Ric Soulis: University of Waterloo
Ontario Ministry of Transportation (MTO) phase 2 progress
The Ontario Ministry of Transportation (MTO) study to interpolate the MSC IDF curve station data for use throughout Ontario is in progress. This presentation will present the results from the 2004-2005 data as well as data from surrounding political jurisdictions. Some discussion of the investigation of the impact of climate change will be included.
Dr Cong Zhentao: Associate Professor, Dept of Hydraulic Engineering, Tsinghua University, Beijing, China
Attribution of actual evaporation trends with the Budyko curve
The actual evaporation trends in 5 major basins in China from 1956 to 2000 were investigated using the Budyko hypothesis. Using the Fu’s equation, an analytical expression of the Budyko curve, the actual evaporation trends were attributed to changes in precipitation, potential evaporation and the shape of the Budyko curve (hereinafter the Budyko curve trends). It is found that the contributions of the Budyko curve trends, which were considered unimportant in the previous studies, turn out to be significant in the basins with substantial change in land surface characteristics due to human activities. With the Budyko curve trends taken into account, the actual evaporation trends estimated using the Budyko curve are in good agreement with those obtained from the water budget method. It is also observed that the actual evaporation trends are more sensitive to precipitation trends in the northern basins, and more sensitive to potential evaporation trends and the Budyko curve trends in the southern basins.