Dialogue on Water and Climate in Yellow River Basin
INTERNATIONAL WORKSHOP
ON VULNERABILITY OF WATER RESOURCES
TO ENVIRONMENTAL CHANGE

Beijing International Convention Center, China September 16-20, 2002


PART IV HYDROLOGCAL PROCESS MODELING AND TOOLS TO CHANGE ENVIRONMENT

A Distribution Hydrological Modeling Applied to Heihe Basin in Western China

Jun Xia 1,2, Gangsheng Wang 1, Ge Tan1
1. Institute of Geographical Sciences and Natural Resources Research, CAS, Beijing 100101, China
2. Wuhan University, Wuhan 430072, China
Email: jxia_mail@263.net

The Western China Development is confronted with a quite serious and realistic problem on water resources because of the water cycle's district characteristics, the ecosystem's vulnerability and the increasing demand of water resources. Based on the traits and existent problem of water cycle in western China, the key scientific issues of water cycle in western China has been studied in this paper. Main contribution in this paper is to develop a distributed hydrological modeling coupled physical process grids with a Time Vafian Gain Model (TVGM) that was developed for system approach in University College of Galway, Ireland for river flow forecasting (1995). This model has three advantages:

(a) It can be available to the case of input information imperfection such as the lack of enough precipitation and evaporation observations for distribution modeling;

(b) It can be coupled the present physical process units or conceptual models with hydrological system approach such as TVGM, where time variant gain factor, G (t), could be linked with RS information to extend easily to application of hydrological prediction in basin scale;

(c) Parameter sets of the distributed hydrological model can be estimated in terms of system identification approach to reduce uncertainty

By preliminary application and examination in the Heihe River Basin of Guansu Province with area 1300km2 in arid & semi arid region, it was found that this model is successful with more advantages than conditional distributed model applied in China. Moreover, the new problems were addressed in the paper.


A Distributed Runoff Model for Inland River Mountainous Basin of
Northwest China

Rensheng Chen, Ersi Kang, Jishi Zhang
Cold and Arid Regions Environment and Engineering Research Institute, Chinese Academy of Sciences, Lanzhou Gansu 730000, China
E-mail: crs2008@ns.lzb.ac.cn

The Arid Area of Northwest China lies in the inland area of Asia far from oceans. It consists of several large inland river basins. Many enormous mountain ranges lie in the area and uplift the air current containing water vapor, making the mountainous area receive much more precipitation compared with the low land area in front of the mountains. Glaciers and snow storage develop very well in the mountains. Therefore, in an inland basin of Northwest China, mountainous watersheds are runoff generation areas, while the low land plains, basins and deserts in front of the mountains are the area of water resources consumption and runoff scattering. Thus, the runoff amount generated from the mountains, which run out off outlets of the mountainous watersheds, basically represents the amount of water resources of the inland arid area. Now the oases in Northwest China are diminishing, while the deserts are expanding. Therefore, it is very significant to forecast the runoff amount from mountainous area. For water management, environmental conservation and other applications, monthly runoff amount forecasting may be more important. As a case study, we tend to forecast the monthly runoff from mountainous catchments of Heihe mainstream. Heihe River, originating from the Qilian Mountains, running through the Hexi Corridor and scattering in the deserts, is one of the largest inland rivers in Northwest China, with a drainage area of 130,000 km2. Heihe river basin, like other inland river basin in Northwest China, posses distinctive vertical zonality of landscape: the mountainous area and the area in front of the mountains. The former can be divided into the high mountain ice, snow and permafrost zone and the mountain vegetation zone, while the latter can be divided into oases zone and desert zone.

Distributed hydrological model has been studied worldwide, while rarely used in large basin scale, because it requires too many parameters and too many hydrologic data, meteorological data, soil data, vegetation data etc. In this paper, the author want to use the routine hydrometeor logic data to create a distributed model, coupled with GIS and RS techniques and data. The model takes sub-basin as the minimal confluent unit, divides the main soil of the basin into 3 layers, and divides the vegetation type as forest and pasture. Because there is little data required by the standard distributed model, this model gives some conceptual parameters. The data used in the model are precipitation, air temperature, runoff data, soil weight water content, soil depth, soil bulk density, soil porosity, vegetation cover etc. The model defines and introduces some concepts: soil volume water content, soil water content capacity, soil water amount, and hold that if the water amount is more than the water content capacity, there will be surface runoff. The actual evaporation is proportional to the product of the potential evaporation and soil volume water content. The studied basin is Heihe mainstream mountainous basin, with a drainage area 10009 km2. The model time step is a month.

The data used in this simulation are 1980.1-1995.12, and are divided into 2 parts. The first 10 years' data are used to simulate, while the last 5 years' data are used to calibrate. The results are very well. For the simulation process, the NSE (Nash-Sutcliffe Equation), B (Balance Error) and EV (Explained Variance) are 0.8681, 5.4008 and 0.8718 respectively, while for the calibration process, the NSE, B and EV is 0.8799, -0.5974 and 0.8800 respectively. Thus, the model has been successfully used in Heihe mainstream mountainous watershed.

The results show that, the vegetation, especially the forest in Heihe mountainous watershed, could decrease the evapotranspiration of the watershed, adjust the runoffprocess, and could increase the soil water content.

Analysis of Relation Between Plant Ecosystem and Shallow Groundwater
Flow System in Ejina Catchment of Downstream Heihe Basin

Yanqing Wu1,2
1. CoM and Arid Regions Environmental and Engineering Research Institute of CAS, Landzhou 730000, China
2.Xi 'an University of Technology, Xi 'an China
Email: yqwu@ns.lzb.ac.cn

The study area is located the downstream Heihe basin, arid region inland river watershed, northwest of China. In recent 30 years, the desert oasis has degenerated because unreasonable use of water resources in the middle Heihe basin has resulted in the decrease of the downstream river discharge and the decline of groundwater level. Through site investigation of desert plant ecosystem, soil moisture and groundwater level in phreatic aquifer, we have found that the close connection between desert plant ecosystem and groundwater system.

According to the observation data of groundwater level and analysis of remote sensing images, the change relation between desert plant ecosystem and shallow groundwater level has studied in regional scale and proposed that river runoff entering downstream basin needs 7.5108m3/a to protect the safety of existing desert plant ecosystem.


Analysis of the Interaction Between Groundwater and Surface Water Using
Radon-222 in Middle Heihe Basin of Northwestern China

Yanqing Wu1,2:
1. Cold and Arid Regions Environmental and Engineering Research Institute of CAS, Landzhou 730000, China
2. Xi 'an University of Technology, Xi 'an China
Email: yqwu@ns.lzb.ac.cn

In arid regions of western China, water resources come from mountain watershed and disappear in desdrt plain. The exchange of surface water and groundwater takes place 2 or 3 times in a basin. It is essential to analyze the intdraction between groundwater and surface water in order to use water resources effectively and predict the change of water environment. The conventional method of analysis, however, measures only the flow of a stream and cannot determine groundwater seepage accurately. Since the concentration of Radon-222 (222Rn) in groundwater is much higher than in surface water, the use of 222:Rn was examined as an indicator for the analysis of the interaction between surface water and groundwater. Measurement of the 222Rn concentration in surface water was conducted to detect groundwater seepage into a stream in middle Heihe basin. Furthermore, the simultaneous movement of water both into and out of a stream from the underlying strata was quantified by solving the 222Rn and water balance equations. Meanwhile, river runoff was gauged to determine the exchange rates between surface water and groundwater.


Application of the AnnAGNPS Model with GIS Interface to Erosion
Accounting for Watershed

Ming-Shu Tsou1 and Xiaoyong Zhan1 ,Changyuan Tang2 ,Junhong Chen3
1. Kansas Geological Survey, Lawrence Kansas 66047, USA. Lyle Frees and Chad Volkman, US Departmdnt of Agriculture, NRCS, USA
2. Graduate School of Science and Technology, Chiba University, 1-33 Yayoi-cho, Inage-ku, Chiba 263-8522, Japan
3.Institute of Geography, Guangzhou, China
Email: xyz@ku.edu

Water availability and quality in many watersheds are becoming a threat to food security, human health, and the natural ecosystem. Watershed management depends on monitoring and modelling of watershed. This paper reports the progress of watershed modelling with emphasis on sediment transport in the Cheney watershed, Kansas by using the continuous-simulation Annualised Agricultural Nonpoint Source Pollution (AnnAGNPS) model with GIS interface developed recently.

Annualized Agricultural Nonpoint Source Pollution Modeling System (AnnAGNPS) is a joint colputer model of the USDA Agricultural Research Service and Natural Resources Conservation Service, and was developed to predict nonpoint-source pollutant loadings within agricultural watersheds. It is a popular model in surface hydrology in the USA and other countries. Water, sediment, and chemicals (nutrients and pesticides) are simulated leaving land areas (cells), flowing into the watershed-stream network at user-specified locations (reaches), and eventually arriving at the watershed outlet on a daily basis as final output of the model.

The first part of this paper provides an integrated overview of the multiple facets of watdrshed modeling-GIS-database issues of AnnAGNPS. A newly implemented enhanced GIS interface is discussed briefly. This ArcView extension facilitates the modeling exercise for AnnAGNPS. Also, an approach to fully use Soil Survey Geographic (SSURGO), one of national standard soil databases, was developed for AnnAGNPS.

The second part of the paper presents an application of AnnAGNPS to the Cheney reservoir watershed in south-central Kansas, USA. The watershed covers approximately 989 square miles. Ninety-eight percent of the watershed land is used as agricultural with varies farming practices.
The Cheney reservoir supplies over 60$ of the daily drinking water to the City of Wichita. Last decade sediment loading in the Reservoir was identified as one of the primary problems in this watershed. Result of AnnAGNPS modeling provides a planning tool to make recommendations for remedial action to improve and preserve the reservoir to its maximum potential, to identify which best-management practices (BMPs) minimize sediment load, and for assist in the development of a strategic plan to accomplish the ultimate goal of cleaner water.

Finally, potential applications to watersheds in China and Japan are briefly discussed because these two countries are strongly focusing on long-term best-management practices in watersheds for maximum eco-environmental benefit.

Assessment of the Indicators of Aridity for Uzbekistan Related to Climate
Change
Spektorman T. Yu

Under current conditions, water resource shortages in Uzbekistan, even a small but stable reduction of these resources presents a drastic problem. During the course of the last third of the 20th century, intensive irrigation from the flows of the Central Asian rivers caused the regrettable Aral Sea crisis: the drying-up of the Aral Sea, a reduction of the delta lakes and drastic aggravation of the ecological situation in the Aral Sea Region

Such factors as drought and human-induced changes (the existing irrigation practice and the global warming) are inter-related; this leads to increasing negative impact.

Climate study has shown that changes in various components of the climate system are already underway in the region. Wanning tendencies during both the cold and warm seasons have been observed. Observations in the mountain river basins demonstrate a stable decrease of snow cover. The degradation of glaciers is evident, as well as a reduction in over-all glacial areas. From other side the monitoring data of the past 70 years do not show any significant tendencies towards change in the precipitation regime in the territory of Uzbekistan.

Comparison of Regional Flood Estimation Models Over Some Catchments
in Kenya

Paul Ng'ala O1oo, Prof. Francis M. Mutua, Prof. Laban A. Ogallo

Many annual flood data are too short to allow for reliable estimation of extreme events. Big portions of most river basins are not gauged, yet design floods are needed at ungauged sites. Such short records further contain outlying flood values, which makes estimates of floods of a given return period to be highly uncertain. The short period data can also lead to different cumulative distribution functions to be fitted to stream flow data of neighboring gauges with the same hydrological characteristics.

Adopting regional flood frequency analysis can minimize some of these flood frequency problems. In such analyses, hydrologic information from several gauging stations is combined to provide regionally averaged hydrological information. This can improve or stabilize site-specific estimates based on limited site data. Regional information can also be used to estimate design flood at ungauged river basins, by using models, which can transfer regional information into site-specific information.

Regional flood estimation procedures require the identification of homogeneous regions. These regions were delineated for every basin considered using the rotated solutions of Principal Component Analysis. A comparison of regional flood estimation models was then undertaken in this study and the basins targeted included the Lake Victoria, Tana River and Athi River in Kenya and the models included the Index flood, two component extreme value and empirical Bayes methods. The data used in this study consisted of the annual maximum flood series for sixty-one gauging stations for the period 1960 - 1994.

The General extreme value, two component extreme value and lognormal distributions were the regional distributions used together with the three regional flood estimation models. Their parameters were estimated using the moments, maximum likelihood and probability-weighted moment's methods. The performance of the best regional flood estimation model was determined using the Kolmogorov-Smirnov test, Chi-square test and Akaike Information Criterion. Simulation was further used to test if the best regional distribution could provide skilful estimates of floods over the homogeneous regions.

The results from the rotated solutions of Principal Component Analysis showed that the three drainage basins could be grouped into nine homogeneous flood regions with three regions in the Lake Victoria basin, four regions in the Tana River basin and two regions in the Athi River basin. The nine groups that were derived formed the fundamental basis for the development of the regional flood estimation models, which were used in this study.

The results from the Index flood method indicated that, the extreme value type 2 was identified as the regional distribution that could best fit annual flow records in all the drainage basins. The extreme value type 1 could also be used predominantly in the Tana River basin. The results also showed that the two-component extreme value distribution was inappropriate for modeling annual maximum flows, since the extraordinary flood events were very few in comparison to the ordinary flood events. The empirical Bayes method parameter estimates had very large standard errors, since the super population of flows based on regional regression relationship is weak. The General extreme value distribution was the best fitting regional distribution with the lognormal distribution giving the worst fit. The general extreme value distribution was therefore utilized in the simulation study. It was found that it could provide skilful estimates for the flows in all the nine regions.

It may be concluded from this study that reliable estimates of design flood can be obtained both for gauged and ungauged river basins by regional flood frequency analysis. The Index flood method with the General extreme value distribution was found to be the best regional flood estimation method in this study. The empirical Bayes method was the worst regional flood estimation method. The results from this study can be used for the general planning and management of most hydrological systems in Kenya.


Determination of Change Points in Nonlinear Hydrologic Time Series Using
Fuzzy Statistical Approach

Simon Wu1 ,Gordon Huang2
1.Saskatchewan Agriculture and Food, Regina, Saskatchewan S4S OB1, Canada
2. Faculty of Engineering, University of Regina, Regina, Saskatchewan, S4S 0A2, Canada
Email: swu@agr.gov.sk..ca

Hydrologic time series models are valuable analytical tools for assessing potential impacts of climate change on water resources systems. Most analyses have been based on the assumption that there is a constant regime throughout the temporal dimension. However, regional climate shifts have been observed around the world for many years. Efforts have been made to introduce change detection methodology to solving the problem in the application of time series in water resources. Change points are now normally searched and examined before time series modeling is conducted. In addition to classical sequential analysis, many innovative methods such as Bayesian estimation and Markov Chain Monte-Carlo simulation have been proposed to deal with change detections in the fields of engineering, financial mathematics and econometrics. A common disadvantage with most of the methods is their incapability of handling nonlinear models. Even more important is that those techniques base their detection on an instant time point or relatively very short period. Variations in phenomenon regime could thus be misidentified. A fuzzy statistical approach is applied in this study to nonlinear hydrologic time series analysis. Combining fuzzy clustering with a current change detection procedure forms the approach. Fuzzy change period concept in the form of a-level point is used to properly extend the detection time points. The application of the method to time series analysis for a river flow is presented in the paper. It shows the technique can be used effectively to identify the change points. More case studies are being undertaken to further examine its applicability to predicting environmental changes.

Genetic Analysis of Yellow River Tidewater Inflow Sharp-Reduction in 90s

Xuecheng Zhang, Ling Wang, Qimin Pan
Hydrology Bureau, YRCC, Zhengzhou 450003, China
Email: Xuecheng@371.net

Yellow River mean tidewater inflow in 90s is only 119.2 108 m3 which took 32% of mean tidewater inflow before 90s. With considering rainfall, human being activity such as water resources development and utilization, ecological and environmental construction etc., this paper searches about the causation of Yellow River tidewater inflow sharp-reduction in 90s. The result shows that action of rainfall variety taking 36% and action of human being activity taking 64%.

Human-induced Water Scarcity on Loess Plateau of China

Xiubin He, Fengli Zheng, Mingde Hao, Keli Tang
Institute of Soil and Water Conservation, Chinese Academy of Sciences, and Northwest Sci-Tech University of Agriculture and Forestry, Yangling 712100, China
Email: hexiubin@yahoo.com

Water scarcity is one of the most prominent issues of discussion worldwide concerned with sustainable development, especially in the arid and semi-arid areas. On the Loess Plateau of China, population growth and fast-growing cities and industries have caused ever-increasing competition for water. The present paper shows a down-scale analysis on how the regionwide mass action of food-security-oriented soil-water conservation ecologically influenced regional water scarcity on the Loess Plateau, the Middle Reaches of the Yellow River of China. Result shows a great progress has been achieved in erosion control and food production since the 1980s. About 24% of erosion area has been controlled. Grain yield has increased greatly and sediment in the Yellow River has decreased by about 25%. However, various evidences show that the soil-water conservation measures might have played a great role on regional hydrocycles, leading the depletion of deep soil water, no runoff in the Yellow River and consequently ecological problem. It implies that cropland and artificial forestry are increasingly suffering water stress ecologically. Advancing sustainable development further, or even maintaining the current situation, will be a great challenge given the burgeoning socio-economic development of the area combined with global climatic change.

Modeling Soil Water Dynamic at Watershed Scale in Hilly Area of Loess
Plateau, China

Hongmei Xu
The Institute of Resource Science of Beijing Normal University and The Open Research Lab. of Environment Change and
Natural Disasters of the State Education Commission, Beijing 100875, China
Email: xhm@irs.bnu.edu.cn

In arid and semiarid ecosystem of Loess Plateau of China, the average soil moisture is quite low and the patterns of seasonal precipitation and the quantity of soil water availability are highly variable. Because of this, water is a major determination of ecosystem processes.

The hilly area of loess plateau is well known for its high erosion rate. The hydrologic cycle in hilly area of loess plateau include: 1) low annual precipitation but high-intensity storms with significant variability, 2) high potential evaporation according to the high radiation and poor soil retention capacity 3) The groundwater don't include in hydrologic cycle because of the deep water table and the thick loess soil profile, 4) Surface runoff is mainly produced by infiltration exceed overland flow. All of these showed that rainfall is the main replenishment of soil water. So the ecosystem is particularly vulnerable to climate change, especially the variation of precipitation. A slight change in annual rainfall or the distribution of seasonal precipitation, or the frequency of extreme rain event could produce major ecological impacts. Therefore, modeling soil water dynamic with precipitation, radiation and other micro-environment elements based on the basic principle of the balance of water and energy, simulating the soil moisture spatial distribution at watershed scale according to runoff processes are very important for the arid and semi-arid ecosystem of loess plateau.

In this article we developed a daily time step, multi-horizon, distributed spatial model (SWDM - Soil Water Dynamic Model) to explore the spatial and temporal variability of soil water content determined by the hydrological characteristic of soil, precipitation and vegetation coven In SWDM, soil water is represent in 4 layers (the top layer-20cm, the second-40cm, the third layer-40cm, the bottom layer-100cm). Infiltration adds water to soil layers in a cascading fashion according to water holding capacities of layers and the redistribution among the layers defined as Darcy-Richards equation. No infiltration beyond 2m and no horizontal water movement is taken into account. Water is removed from the top layer by evaporation and from all layers by transpiration. The potential evapotranspiration was calculated by Penman-Montieth equation. Removal of water via evaporation was determined by energy to the surface layer and the total leaf area of vegetation. Removal of water via transpiration form soil is partitioned among the layers according the distribution of total roots in each layer and related to the total leaf area of vegetation and the transpiration capacity of the plant canopy. Calculations of runoff and run-on flows are important for a watershed scale simulation model, as the flows represent important spatial processes in a watershed. Runoff was generated when total rate of rainfall is larger than the steady penetrating rate and the surface soil gets saturated. Runoff from a grid cell is distributed to one of its neighborhood cells with lower elevation. Run-on to a grid cell is simply summation of surface runoff from higher neighboring grid cells.

The model was tested in Zhifanggou watershed, which belongs to Ansai city and covers an area 8.27Km2.The regional topographic types are loess hilly and gully landforms. The soil has developed over loess with fine silt to slit in texture, and slope vary between 10-30~. The present vegetation cover of the study area is composed mostly of natural grassland, artificial shrub, woodland and farmland. Climate variables and vegetation cover pattern were used as input driving variables for the model. In the modeling experiment, SWDM simulations have been shown to provide good approximation with field observation data.

On the Accuracy of Spherical Gages for Rainfall Measurements

Mingteh Chang1, Lee Harrison2
1. Arthur Temple College of Forestry Stephen E Austin State University. Nacogdoches, Texas 75962-6109, USA
2. Meteorologist-in-Charge, US National Weather Service Shreveport, Louisiana 71109- 7750, USA
Email: F_ changmi@sfasu.edu

Because of wind effects, point rainfall measurement is always deficient and a 10% error of rainfall input may result in a 15-30% error in runoff output in hydrologic simulations. Two spherical orifices were designed to modify the standard gage and other gages in use today (Chang and Flannery, 2001). Because of the spherical shape, the two orifices will catch rain with an effective diameter always equal to the actual diameter regardless of wind speed and direction. This reduces wind effect on rainfall catch to a minimum level. This report tested the spherical gages at two different locations, one at the City Landfill, Nacogdoches, Texas and the other at the NWS Forecast Office, Shreveport, Louisiana. Based on 131 storms at Nacogdoches and 94 storms at Shreveport observed between May 1998 and February 2001, the results showed: 1) spherical gages recorded an average 6-9% greater than standard gage and 3-4% less than pit gage, only 1-2% less than reported in the original study, 2) the catch of spherical gages was not significantly affected by three gage heights at 0.91, 1.83, and 2.74 m above the ground, but catch by the standard gage decreased with increasing gage height, 3) improvements of the spherical gages were most significant for larger storms and for winds at higher speeds, 4) the spherical gage with cylinders recorded 1-2% more rainfall than the spherical gage with vanes, and 5) correlation coefficients between catch deficiencies and wind speed were low and weak because of the distance and height of the existing wind sensor.

Due to greater surface wetting and evaporation loss, the spherical gages may underestimate rainfall catch by standard gage for small storms (generally less than 0.50 cm), especially in hot summer afternoon and for smaller storms. However, the underestimates do not over-shadow the merits of spherical gages because the differences are too small to be of hydrologic significance. Using polyethylene or other synthesized materials to construct spherical orifices may improve the catch for small storms. The results of the study agreed with the previous claims that spherical gages are effective in reducing wind effects on rainfall measurements and are suitable for large-scale applications.

Research on the Flood Forecast Model of Nonlinear Runoff and
Concentration for Luhun Reservoir

Xingyuan Song1, Shenglian Guo1, Zhicheng Su1, Yong Du1, Bei Wang1
Huale Jia2, Mibo Sun2, Yansheng Wang2, Jingwei Hao2
1. School of Water Resources and Hydropower, Wuhan University, Wuhan 430072, China
2. The department of Luhuen Reservoir Luoyang, Hunen 471000, China
Email: xysong@wuhee.edu.cn

The effects of basin nonlinear runoff and confluence are considered from the basin rainfall space-time changes and the basin topography as well as the river channel characteristics. This paper presents the basin models of nonlinear runoff and confluence and the method of real-time flood forecast. Using them to forecast the floods of Luhun reservoir, it shows that the models and method are effective in practice.

Sediment Yield Estimation in Cheney Watershed Using Annagnps-gis
Modeling System

Ming-Sbu Tsou1, Xiaoyong Zhan1, Li Zheng1, Lyle Frees2
Chad Volkman2, Changyuan Tang3, Junhong Chen4
1.Kansas Geological Survey, Lawrence Kansas 66047, USA
2. US Department of Agriculture, NRCS, USA
3. Graduate School of Science and Technology, Chiba Universityl-33 Yayoi-cho, Inage-ku, Chiba 263-8522, Japan
4. Guangzhou Institute of Geography, Guangzhou, 510070 P.R. China
Email: Zhengli_usa@yahoo.com

Water availability and quality in many watersheds are becoming a threat to food security, human health, and the natural ecosystem. Watershed management depends on monitoring and modeling of the watershed. This paper reports the progress of watershed modeling with emphasis on sediment transport in the Cheney watershed, Kansas, by using the continuous-simulation Annualized Agricultural Nonpoint Source Pollution (AnnAGNPS) model with GIS interface developed recently.

The first part of this paper provides an integrated overview of the multiple facets of watershed modeling-GIS-database issues ofAnnAGNPS. A newly implemented enhanced GIS interface is discussed briefly. This ArcView extension facilitates the modeling exercise for AnnAGNPS. Also, an approach to fully use Soil Survey Geographic (SSURGO), one of national standard soil databases, was developed for AnnAGNPS.

The second part of the paper presents an application of AnnAGNPS to the Cheney reservoir watershed in south-central Kansas, USA. Last decade sediment loading in the reservoir was identified as one of the primary problems in this watershed. Result of AnnAGNPS modeling provides a planning tool to make recommendations for remedial action to improve and preserve the reservoir to its maximum potential, to identify which best-management practices (BMPs) minimize sediment load, and to assist in the development of a strategic plan to accomplish the ultimate goal of cleaner water.

Finally, potential applications to watersheds in China are briefly discussed because this country is strongly focusing on long-term best-management practices in watersheds for maximum eco-environmental benefit.

Simple Methods of Hydrological Data Provision for Gauged and Ungauged
Catchments and Their Application in Developing Countries

Vladimir. U. Smakhtin
International Water Management Institute, PO Box 2075, Colombo, Sri Lanka
Email: v.smakhtin@cgiar.org

The need for continuous stream flow time series is constantly increasing with increasing pressure on water resources and emerging concepts of environmental flow management in many parts of the world. The conventional way of generating stream flow time series at ungauged sites in river catchments and under different scenarios of catchments development is through the use of deterministic rainfall runoff models, which represent a flexible but information consuming and labor intensive approach. In many developing countries, the input information required to run a detailed rainfall-runoff model may simply not be available and consequently, the use of such techniques may not necessarily result in a reliable output. In such circumstances, the use of simpler and quicker estimation methods may be more justified and, at the same time, equally successful.

This paper intends to illustrate the pragmatic alternative to deterministic modeling, which may be particularly relevant in data poor regions. The approach is based on a Flow Duration Curve (FDC) - a cumulative distribution of stream flow values and a graphical summary of stream flow variability at a site. The ultimate goal in pragmatic hydrological modeling concept is the same as in any catchments modeling - a continuous stream flow time series, representing a catchments hydrological response under natural condition or under certain scenario of development. The calculation of representative FDCs is the prerequisite for it and the starting point. The shape of the curve is determined by rainfall pattern, catchments size and physiographic characteristics, land-use type and the state of the water resource development. The effects of these factors on stream flow may therefore be built in the curve itself, prior to the simulation of actual continuous stream flow time series. In general, the methodology of pragmatic hydrological time series modeling includes:

Technique(s) by which to establish representative FDCs for different types of river catchments in natural conditions. "Natural" FDCs represent reference conditions of stream flow variability in a catchments, which existed prior to any catchments or water resources development. These techniques normally involve methods of hydrological regionalisation

Technique(s) by which to adjust the "natural" FDCs to match with the current state or possible scenarios of catchments and/or water resources development. These techniques allow different catchments development scenarios to be simulated (which also represent one of the goals of conventional hydrological modeling). These techniques involve correction of discharges of different exceedence values to account for effects of catchments development, such as irrigation, commercial forestry etc.

Technique(s) by which the established and adjusted FDCs may be transformed into actual continuous flow time series for any further analysis.

The latter group of methods is built around a non-linear spatial interpolation technique, which is based on the assumption that flows occurring simultaneously at sites in a reasonably close proximity to each other correspond to similar percentage points on their respective FDCs. The technique is designed to generate continuous flow time series at a (destination) site using either available flow data from the nearby (source) stream flow gauges or rainfall data from rain gauges in catchments. For each selected source site and for the destination site, tables of discharge values are generated for each month of the year (or for the whole year) for the 17 fixed percentage points on the corresponding FDC. Following the main assumption of the algorithm, the core of the computational procedure includes the estimation of the percentage point for each day's flow at the source site and the identification of flow for the equivalent percentage point from the destination site's FDC. The discharge tables are used to "locate" the flows on corresponding curves and log-interpolation is used between fixed percentage points. If several source sites are used, the procedure is repeated for each of them and the final destination discharge value on that day is calculated as a weighted average of values obtained using individual source sites.

If no suitable source sites with observed flow data are identified, the use may be made of rainfall records. If the source data from rain gauge(s) are used, the source flow time series is replaced by the time series of precipitation index, which reflects the current status of catchments wetness. This "current precipitation index" (CPI) is defined as a continuous function of precipitation, which cumulates in rainy days and exponentially decays during the periods of no rainfall. The process of rainfall-to-runoff conversion is then based on the assumption that current precipitation index values at rainfall site(s) in catchments and stream flow discharges at a destination site correspond to similar probabilities on their respective duration curves. AFDC at the ungaged destination site is established by regionalisation methods.

The pragmatic hydrological modeling has a limited number of parameters. It is transparent, simple to use, time efficient, not regionally restricted and is applicable for generation of both daily and monthly flow data. It could be used for hydrological time series patching and extension, for the restoration of natural daily stream flow time series in already modified catchments (for example in projects related to estimating ecological in stream flow requirements), for generation of inflows to coastal lagoons and estuaries, for data generation in ungagued catchments, etc. The paper presents examples of application of pragmatic time series modelling concept in several developing countries including South Africa, Sri Lanka and Nepal.

Spatial Analysis of Rainfall Data Based on Self-Organizing Competition
Neural Networks

Xiang Zhang, Jun Xia, Xingyuan Song
Department of Hydrology and Water Resource, College of Water Resources and Hydropower, Wuhan University, Wuhan 430072, P. R China
Email: Scottzhx@yeah.net

The analysis of the spatial variability of rainfall has been of concern to hydrologist for many decades. The methods of analysis include the traditional methods such as Thiessen polygons, linear interpolation based on distance between sites and geostatistical methods such as Kriging. These traditional methods are mainly used to solve estimation or interpolation problems; much less work is available to deal with spatial classification problems. In this paper, we introduce a new methodology, based on artificial neural networks, to characterize spatial rainfall maps.

Artificial Neural Networks (ANNs) can be classified as two kinds such as supervised and unsupervised networks by their training algorithm. Self-organizing competition neural networks are belonging to the unsupervised networks. Unlike feed-forward and recurrent neural networks that are primarily used for approximation and optimization, self-organizing competition neural networks are typically used for patterns classification and projecting patterns from high dimensional to low dimensional space. In this paper, a new kind of self-organizing competition neural networks (SOCNNs) that hybrid ART II and Kohonen network is invested for classifying the inherent space variability of rainfall data.

SOCNNS have two layers that are input and output layer respectively. The input layer with n neural nodes passes the observed rainfall data sets that are structured as the input vectors with subset consisting of the daily elements. The neural nodes in output layer represent the classes of rainfall. Their number will be determined as 1 at the beginning of training. During the process of training, the number of neural nodes in output layer will add 1 when one new class of rainfall is found. After all the data sets are learned, the number of neural nodes in output layer is the classes' number of rainfall. The space distribution of different classes at the observing area shows the space variability of rainfall.

The training method of SOCNNs is unsupervised competition learning law. The first step is to compute a matching value between the input patterns and the patterns that the neural nodes in the competitive layer (i.e. the output layer) represent. Then the neural node that has closest match to the input is identified as a winning unit. Interconnection weights of the winning unit are updated to more closely match the input pattern. At last after all the rainfall patterns are learned, each set of the interconnection weights is the clustering center of each class. All of the rainfall patterns are classified into the different classes that the output units represent.

The space variability of rainfall in Hubei province is selected as the case study. The observed daily precipitation data at 77 gauging stations covering from June 20 to July 20 in 1996 during the flood season are analyzed. According to the training algorithm of SOCNNs, the neural nodes in input layer are 31 that are the time span of observed data (31 days). The total learning samples are 77 that is the number of gauging stations, i, e, each station's 31 days observed data are selected as one rainfall pattern. All 77-rainfall patterns are input to the network model. Through the unsupervised competition-learning algorithm, the winning units in output layer are five. So the 77 rainfall patterns are distributed into different five classes. The five classes show the space distribution and variability of rainfall in Huber province.

The space analysis of rainfall is important for many problems related to hydrological, water resources and environmental system. In this paper, a space analysis model, SOCNNs, based on ANNs, is invested. Its applicability in space variability analysis of rainfall is illustrated by a case study in Hubei province.

Study on the Modelling Method of Land Water Quantity and Quality and
Ecology Coupling System

Qiting Zuo 1 Ge Tan 2 Jun Xia 2
1. College of Water Resources and Environment, Zhengzhou University, Zhengzhou 450002, China
2. Institute of Geographical Sciences and Natural Resources' Research, CAS, Beijing 100101, China
Email: zuoqiting@netease.com

As the basic content of quantitative study on Managerfient of Sustainable Water Resources, it is necessary, and difficult as well, to build water quantity and quality and ecology coupling system model. On the basis of application study, this paper puts forward a new method called "Multi-Box Modeling". This method's clue is definitude, calculation is simple, and model is reliable. It is confirmed by the authors that the built model can stand for "land water resources system construction connection" and can be used in quantitative study on Management of Sustainable Water Resources.


Surface Overland Flow and its Alternation by Urbanisation.

Lev Kitaev, Elena Barabanova

Many authors, pointing out significant changes in hydrological processes under the human activities, mean first of all those, connected with agriculture. The Institute of Geography of Russian Academy of Sciences is engaged in research on intcrannual changes of surface overland flow depending on the influence of climatic and anthropogenic factors. To the latter is referred the rising urbanisation. We can estimate the contribution of urban territory to the surface flow of river basins. As an example we give the results of 13-year investigations in the upper part of the Seym basin the left tributary of the Desna river (the steppe zone of the East European Plain). The catchments area is 18100 km2, mcan annual discharge is 105 m3/c.

A distinguishing feature of the flow from urban territories is higher values of the surface flow, soil and dissolved matters washout. Compared to adjacent areas they may be 3-4, 8 and 20 times larger correspondingly. First of all this is connected with a significant change in soil absorbing capacity due to asphalt surfaces.

Winter period is characterised by negative trends of the surface flow and soil washout on different types of ground surfaces, this correlates to snow storage decrease by the beginning of spring flood due to snow evacuation and intensive melting during thaws. As this takes place, w4nter-spring precipitation and annual river runoff show positive trends emphasising the importance of anthropogenic impact on natural processes. In summer, at positive precipitation trend, surface flow decreases slightly within municipal building areas and does not change within villa building areas. The soil washout changes in summer in the same way as the flow docs.

In the upper part of the Seym basin the urban territories, covering near 5% of the total, make up a significant proportion of the net annual volume of the surface slope flow - near 15%:

Hence significant changes in hydrological processes are observed in a town compared to those taking place within agricultural and reserved territories. At a relatively small city area its contribution to the water balance of the river basins is essential.

Switching Regime Analysis for Non-linear Hydrologic Time Series

Simon Wu1, Gordon Huang2
1.Department of Agriculture, Government of Saskatchewan, Regina, SK S4S OBI, Canada
2.Faculty of Engineering, University of Regina, Regina, Saskatchewan, S4S 0A2, Canada
Email: wusimon@env.uregina.ca

Time series models are valuable analytical tools for assessing potential impacts of climate change on water resources systems. Hydrologic time series has seen numerous applications in the past several decades. However, most analyses have been based on the assumption that there is a constant regime throughout the temporal dimension. In reality, regional climate shifts have been observed around the world for many years, which triggers changes in hydrologic behaviors of water resources. Efforts have been made to introduce change detection methodology to solving the problem in the application of time series in water resources management. Change points are now normally searched and examined before time series modeling is conducted. In addition to classical sequential analysis, many novel methods such as Bayesian estimation and Markov Chain Monte-Carlo simulation have been proposed to deal with change detections in time series modeling. There have been significant amount of literature in change-point detection methodology and their applications in various science, engineering and business fields. However, publications in change-point analysis for hydrologic time series have been very few and only limited to Bayesian approaches. A common disadvantage with the Bayesian methods is their incapability of handling nonlinear models. Linear time series models and their corresponding change-point detection methods basically do not support insight into switching regimes in time series dynamics. Linear models for hydrologic forecasting and simulation such as Auto-regressive Moving Average (ARMA) have been dominating in research literature, but actually gained limited acceptance in real-world practices. For example, when daily hydrologic data are of interest, linear modes are often hard to justify. Some innovative approaches such as neural networks, time-delay embedding, global dynamic models, have emerged to deal with problems associated with nonlinear time series. Non-linear time series analysis has been applied in many hydrologic studies. However, it is relatively hard to analyze the system dynamics due to the nonlinear nature of models, and few studies have been found applying switching regime analysis to nonlinear hydrologic time series forecasting or simulation.

In this paper we present an approach to detecting switching dynamics for hydrologic time series based on Hidden Markov Model. The Hidden Markov Model has been successfully applied to change-point analysis in data processing in other industries like financial market and industrial processes. A method proposed by Kohlmorgen and Lemm is adopted in the study and applied to a typical hydrologic case with Mackey-Glass time series. The basic theory of Hidden Markov Model is first introduced, and followed by description of change-point analysis algorithm. The case study with hydrologic time series is presented with discussion focusing on the applicability of the approach to hydrologic studies.

The Hydrology Process in Mountain Vegetation Zone in Inland River Basin in Northwest China

Kechao Song1, Ersi Kang2, Bowen Jin 1,2
1.COM and Arid Regions Environment and Engineering Research Institute, Chinese Academy of Sciences, 260 Donggang West Road, Lanzhou 730000, China.
2. Institute of Water Resource Conservation Forest in Qilian Mountains of Zhangye, Gansu 734000, China
Email: songkechao@msn.com

The water resource in northwest china mainly come from mountain vegetation zone and mountain glacier and frozen zone. In mountain vegetating zone the main vegetation type is mountain meadow, conifer forest, shrub wood. The main three vegetation types present in the approximate altitude range from 1800m to 3800m, but only over a horizontal distance a few kilometers. The average yearly precipitation in mountain is about 400mm-500mm and about 80% of precipitation fall from may to September.

In study, we characterize the differences of the water and heat regimes of different vegetation stands by comparing both water and heat that input into and output to different altitudinal vegetation stands, then conclude that it is the special vegetation types, special ground cover of litter moss layer, frozen layer, topography in vertical vegetation landscape in mountain vegetation area determine the special hydrology process in mountain vegetation zone of inland river basin. The difference in the amount of water and heat that input in conifer forest and shrub wood and meadow from the upper canopy is caused by the strong altitudinal gradients of precipitation and air temperature in mountain vegetation zone. The differences of distribution of water and heat in canopy and soil layer are caused by the difference in the geometry structure of canopy and the difference in the character of the soil poifle, so the frozen depth and soil water content and the composition of the runoff present the result that bigger runoff coefficient is presented in higher altitude belt of mountain vegetation zone.

The interception rate of precipitation in conifer forest, shrub wood are about 21%--35%,26%--41% respectively, if the precipitation is larger than 3mm and less than 20mm, account for actual measurement. Litter-moss layer in conifer forest or shrub wood reintercepts the through fall reaching the ground surface. The litter-moss layer's thickness is averagely 4cm-10cm, in local site can reach 15cm-25cm. The humus layer under the litter-moss layer can accumulate to 3cm-10cm in thickness and absorb infiltration water render the shallow soil water content larger than the depth soil's, so there have more subsurface runoff flow into the branch of fiver because of the larger antecedent soil water content. About 23-34% intercepted through fall by litter-moss layer infiltrate in shallow soil layer, about 66-87% intercepted through fall evaporate in atmosphere account for measurement. Only in rainy season subsurface runoff generate above seasonally frozen layer, during winter and spring, some snowmelt and through fall storage in shallow soil layer at the form of ice. The seasonal frozen soil layer in forest area occur at the end of October and thaw throughout in may, during this time, the net infiltration water all storage in shallow soil layer and exhaust in the transpiration and subsurface runoff.

The distribution regime of precipitation that falls on meadow is controlled by sparseness or density of grass. The infuriation water mainly used for grass transpiration if the grass stands is density; else the rainfall reaching to sparse grass is mainly exhausted for soil evaporation. Generally, in grass hill slope, there have no surface runoff in process of precipitation except in local site. In one measurement, the runoff coefficient is only 0.04 in a continuous 34mm rainfall process. The frozen layer in meadow area is manful for snowmelt runoff in spring, at the starting of summer the frozen layer does not exist.

The Interval Extension Evaluation Model of Water Environmental Quality and its Application

Jun XiaI , Baoqing Hu2
1.Institute of Geographic Science and Natural Resources Research, Chinese Academy of Science, Beijing 100101; Wuhan University, Wuhan 430072, P. R. of China
2. School of Mathematics and statistics, Wuhan University, Wuhan 430072, P. R. of China
Email: jxia_mail@263.net; bqhu@public.wh.hb.cn

This paper firstly introduces concepts of/E-distance, IE-locating value and /E-correlative function and proposes interval extension evaluation model (IEEM) and its application to water environmental quality assessment. This method solves assessment problem while quantity of assessed matter-element is an interval number produced by a variety of uncertainty.


The Ratio of Transformation From Precipitation to Water Flux into the Sea of the Yellow River, China, as Influenced by Human Activities

Jiongxin Xu
Institute of Geographical Sciences and Natural Resources Research, Beijing 100101, China
Email: xujx@igsnrrac.cn

To quantify the transformation from precipitation to water flux into the sea, an index of coefficient of water flux into the sea (Qw, s) is proposed in this study, which is defined as the ratio of water flux into the sea divided by the precipitation by volume over the river drainage basin. It has been found that Qw, s has declined since the 1950s, as a response of water cycle system of the Yellow River to the changing human activities and precipitation. Due to sharp increase in water diversion by man for irrigation, industrial and domestic uses, a strong lateral-branch water cycle subsystem has been well built in the Yellow River water cycle system. In the meantime, basin-wide water and soil conservation measures have been practiced, significantly enhancing the local water cycle from precipitation, infiltration to evapo-transpiration.

Based on the data from 1952 to 1997, a multiple regression equation has been established that relates Qw, s to water diversion by man, area of land terracing and tree- and grass planting and annual precipitation. The equation indicates that, to the decline of Qw,s, the factor in the first place is water diversion by man, the factor in the second is water and soil conservation measures, and precipitation is in the third. These results may be used as reference in decision making for the drainage basin water resources and delta environment management of the Yellow River.

The Realization of Quasi Real-Time System of Economical Operation of Hydropower Plant Based on Neural Network Method and Dynamic Programming

Qiang Huang, Chenguang Xu ,Maihuan Zhao
Xi'an University of Technology, Xi'an, Shaanxi Province 710048, China
Email: mhzhao@163.com.cn

Economical operation of hydropower plant researches are mainly about the balance of generator output, runoff and water head; the dynamic characters and the hydrologic characteristics of turbine generator; the reasonable distribution of load among generators; the economic combination of generators; the start-stop schedule of generators. Compared with short-term and long-term Economical Operation of Hydropower Plant, the principal difficulty of the implement of Economical Operation of Hydropower Plant (EOHP) is how to realize real-time system. In the past the study on algorithm laid particular stress on its accuracy while ignored its computing speed, which is essential to EOHP. For the sake of the hydropower plant to take response in time for the change of water level and electrical load, precise and high-speed algorithm should be established.

Traditionally the efficiency of water turbine generator comes from the efficiency characteristic curve of water turbine generator, which depends on the practical test data under some typical head. Although this method is fast enough, its error is evident. In this paper, the technique of artificial neural network (ANN) is proposed. After the efficiency test data under some typical head having been trained, groups of water head and generator output of the turbine generator are suggested as inputting structure parameters into the input layer, the efficiencies are used as outputs. Then the proper weights and biases are obtained, which can be used to forecast the efficiency of water turbine generator. Practically, the water head and generator output of last period can be used to modify the weights and biases of ANN.

Spatial optimal operation of hydropower station, which is based on the changing conditions of the water level and electrical load, can be realized by dynamic programming. If the changes achieve a certain domain, spatial optimal operation of hydropower station will begin, else the operating way of generators in last period will be continued.

The improvements in these two respects can increase the accuracy EOHP. And the simulation results show that they ran fast enough, which are faster than the hardware do. Thus the quasi real-time system of economic operation is achieved.

The Theory and Structure of SWAT Model: A Case Study in He;he Basin

Zhonggen Wang1, Youbo Huang 2, Xianfeng Wu 3
1. Institute of Geographical Sciences and Natural Resources Research, CAS, Beo'ing 100101, China
2. Hydraulic and Electric Engineering Department of Wuhan University, Wuhan 430072, China
3 Institute of environmental science, Beijing Normal University, Beijing1008 75, China
Email: zhg-wang@163.net

SWAT (Soil and Water Assessment Tool) is a physically based, long-period distributed hydrological model. It was developed to predict the impact of land management practices on water, sediment and agricultural chemical yields in large complex watersheds with varying soils, land use and management conditions over long periods of time. This paper addresses the issues of the hydrological recycle theory, the structure and functions of SWAT model. Finally, the SWAT model is applied in the heihe (yingluo valley) basin.

To Ascertain Design Flood of Longmenzhen Hydrologic Station in Yi River by Using Observed Storm Data

Xianghui Yang
Hydrology Bureau; YRCC, Zhengzhou 450003, China
Email: Yangxh@371.net

Flood is usually formed by storm in the most region of China. In actual practice, sometimes discharge-rainfall correlation method can not be used for interpolating and extending discharge data because of the lack of observed discharge data or the less series representative, sometimes design flood can't be ascertained by using discharge data because of obvious changed runoff forming conditions and breakage of series coincidence caused by human beings activities. Design flood need to ascertain by storm data in these cases or in the middle and small basin without discharge data, while can add the multiformity for calculating design flood in order to proving the rationality of design results. Design flood results are usually ascertained by using discharge data in Yellow River basin. In this paper, the design flood of Longmenzhen hydrologic station in Yi River of branch Yellow River is ascertained by using storm data. It is an attempt of the method applying in Yellow River basin.

It is a main process for calculating design flood as follows. Firstly, collecting 1953~ 1997 years observed discharge series and precipitation series as calculation series, three respective big storm and flood process in 1954,1958 and 1982 years as the typical storm and flood process, restoring 1960~1997 years observed discharge data influenced by Luhun reservoir regulation in Yi River upstream, and analyzing the representative of rainfall stations. Secondly, applying Xin Anjiang Model as runoff yield and flow concentration calculation model and ascertaining model's parameters. The model has 4 parts including evaporation calculation, runoff yield calculation, runoff's composition calculation and flow concentration calculation, with 14 parameters. Thirdly, according to 1953~ 1997 years average area rainfall, the maximal l day, 3days, 5days, 12days design area rainfall Xp(P=0.01%, P=0.1%, P=l%) are calculated by using P-III frequency distribution curve, then the design rainfall process corresponding to 3 typical storm process could be ascertained by the homogeneous frequency enlargement. Then, we could calculate the design flood process by the design rainfall process using the runoff yield and flow concentration model with its parameters and initial conditions. Fourthly, to ascertain Longmenzh's design flood volume by discharge data and compare with the results of design flood volume by storm data.

The calculation results express that there are not obvious difference between the design flood volume calculated by storm data and by discharge data with Xin Anjiang Model, its parameters and initial conditions. Therefore, it is thought that the method of ascertaining design flood by storm data would be possible, could be applied to the middle or small basin lacking discharge data.