GISdevelopment.net ---> AARS ---> ACRS 1990 ---> Water Resource

Studies on satellite snow cover monitoring and snowmelt runoff forecast in the upper reaches of The Yellow River

Zeng Qunzhu, Fen Xuezhi, chen Xianzhang, Lan Yongchao Wang Jain, Jin Dehong
Lanzhou Institute of Glaciology and Geocryology
Chinese Academy of Sciences

Liu Yujie
Satellite Meteorology Center, State Meteorological Administration
Beijing, China

Song Qiang
Geography Deparment of Huazhong Teachers University,
Wuhan, China


Abstract
The formation and spatial and temporal variation feature of winter-spring snow cover in the upper reaches of the Yellow River and the effects on hydrologic situation made by snowmelt runoff were studied in this paper systematically by using NOAA, Land sat MSS, Tm images and ground truth data. The spring runoff Forecast Model was developed based on the Ridge regression and gray system theory employing satellite snow cover data and hydrometer logical data. The inflow forecast for each ten-day to longyang Gorge Reservoir from the first ten days of April to the first ten days of June has been carried out successfully for 3 years (1987, 1989, and 1990) The result shows that the forecast precision is within the permitted error range and can meet the need of user perfectly.

General situation of the basin and research content
  1. General situation of the basin.

    The upper reaches of the Yellow River is situated in the northeast of the Qinghai-Xiezang Plateau. To the southwest of the basin there is the Bayanhar MT. as a watershed between the Yangtse River and the Yellow River. And the Anyemaqen MT. lies passing through the basin from the northwest to the southeast. The highest peaks are over 3200m and 6262m a.m.l. respectively (fig 1) The study area was selected from the source of Yellow River to Tangnaihai hydrological station with 1172 km2 in length and an area of 121972 km2 Because of the abundant precipitation (300-750 mm/yr) the low annual mean temperature (-4,0-1 1 0C ) and the weak evaporation of the ground surface the runoff modules reaches 6.51dm3/s.km2 in this area which is about three times as high the mean value of the whole Yellow River basin , so the study area is one of the main runoff producing areas. In addition, the snow and glaciers are disTributed widely (fig 2) and 57 glaciers developed around Maqengangri peak with an area of 120.75 km2 The comparison of topographic man and Land sat MSS shows that the terminal of Halongglacier No.1 the largest glacier in this area has advanced by 720mm in recent ten years by 200m and 430m for the Halong glacier N0 2 and Walema glacier respectively. The melt water from glaciers and snow is one of the main supply sources of ring runoff in the upper reaches of the Yellow River and accounts for about 72.6% of the surface runoff in the same period.



  2. The Research Content

    The dynamic monitoring of winter-spring snow cover in the area and the operational forecast of spring runoff (fig 3 ) have been carried out under the support of Ice and snow Water Resources Information System (ISWRIS ) using snow cover information extracted from NOAA-9 10 , 11 AVHRR data Landsat TM MSS , image or CCT data and the ground data such as field in vestigation data experimental data and hydro metrological data the research results can provide a scientific basis for the management of large reservoirs in the upper reaches of the Yellow River and are very beneficial to their operational run.




Satellite snow cover monitoring and snow distribution in the basin
  1. Data and Processing Method.

    1. NOAA AVHRR data. The real time NOAA AVHRR positive transparencies from channel 1 to channel 4 with longitude latitude grid provided by the Environment Forecasting Center of National Oceanic Administration were proceeded by means of NAC-4200F Multicolor Data System in the following steps.

      Firstly the basin mask of the 1:1,000 000 topographic map and the water system chart the channel 4 (10.3---11.3 mm ) and channel 1 (0.58---0.68 mm ) images are set up successively in refresh memory of the system when the images of channel 1 and 2 have been matched and compo sited

      Secondly the mask channel 1 and histogram of the pixel values are displayed synchronously .Meanwhile the image contrast value c can be changed by histogram adjust if the C is rather small. The adjustment of image brightness should combined with the visual interpretation of snow information so that the proper snow brightness range can be determined on the image. The snow covered area can be extracted by density slicing and the graded coloring of snow covered area and the calculation of snow cover percentage is as follows.

      R=S/f/B/f %

      Where , P- the snow -cover percentage of the basin (%);
      S/f--- the snow- cover percentage within a rectangle range and
      B/f --- the area percentage of the basin relative to the rectangle range.


    2. AVHRR/HRPT digital data the preprocessing of the AVHRR/HRPR data is carried out as follows : The CCT " IB " data set is created after correction of solar zenith angles and limb darkening the stereographic mosaic is employed to AVHRR data of three orbits so that the image distortion can be eliminated according to the difference of that cloud and snow varies with time respectively the cloud effect can be eliminated by the method of 7 day minimum brightness the image data file can be created through the formula below:

      G(I,J) = Min[g1(I,J),g2 (I,J),…………,g7(I,J)] (2)

      Where, G(I,J) - the minimum brightness in 7 days at the point (I,J);
      gi (I,J)-the minimum brightness on certain day (i) at the point (I,J).

      The graded reflectance of NOAA AVHRR in visible channel can be calculated as follows

      Rk= SiG + Ii

      Where RK - the reflectance of ground surface after the correction of solar zenith angles.
      G - The brightness of any point on the composite image.
      Si,Ii --- the calibration of visible channel.

      Each level is 10% in the graded reflectance from level 0 to level 8 ( the 9 level is defined as the pixels outside the basin ) the levels 0,1 and levels 2,3 represent the background and snow respectively based on the logic analysis of season and the characteristic of ground object the test shows data of the basin extracted with the interpretation thresholds above. The result of interpretation is put in to forecast model of snow melt runoff as a variable.

    3. RR /APT image for the AVHRR/APT images with large distortion provided by provincial meteorological observatories the snow covered areas can be transferred from the image on to the topographic map and then measured after doing geometric correction.

    4. The 1 :1500000 snow series charts were mapped based on the AVHRR APT images and snow data in the study area and nearby from 1971 to 1980.


  2. Snow Distribution Characteristics of the basin.

    Generally the reflectance of snow cover falls down with the decrease of snow cover percentage within a pixel and finally it becomes the reflectance of background within a test site 0f 1.0*1.0 m2 at 3800 m a.s.l. near the Anyemaquen Mt. it was studied that the reflectance if a fresh snow in melting process varied with the snow cover percentage. The result provide a reference for snow -cover interpretation of AVHRR data.

    According to the analysis of NOAA APT and AVHRR image for recent ten years it is known that the stable seasonal snow cover begins at the end of September or the beginning of October the snow covered area extends gradually from high to low elevation and from south to north and reaches the maximum in the first ten days of April of the next year then the area begins to contract day by day gradually. The seasonal snow cover will melt away in the first ten days of June expect the basin of Anyemaqen Mt. table 1 shows the mean value of snow cover percentage in the study area over the years (1976-1990) from the last ten days of March to the last ten days of May on the basis of statistics APT and AVHRR data.


Table 1. The mean snow-covered area in the study area for each ten-day from March 20 to May 30 over the years (1976--1990)


Month March April MAY
Snow Area III I II III I II III
Snow cover (%) 9.8 10.3 7.6 7.4 5.0 7.1 3.7
Snow area (km2) 11953 12562 9269 9025 6098 8659 4512


Spring Runoff Forecast
  1. Main Factors affecting on spring Runoff.

    1. Relationship between snows covers percentage and spring runoff. The precipitation condition can be made not approximately through the analysis of snow cover percentage data obtained from AVHRR image. When the parameter of annual accumulated value of spring snow cover percentage, SS(%) is determined the winter spring precipitations, PS (mm) can be calculated by following

      Ps=1/(0.0067+158176.6e-SS)---------------------------------(4)

      The analytical results of AVHRR images since 1982 show that there was abundant snow in the upper reaches of the Yellow River for the years of 1982, 1983,1989 and 1990 with corresponding large spring runoff amount and the reverse was true fig 6 shows the close relationship between the accumulated value of winter snow cover and spring runoff .


    2. spring precipitation and air temperature effects on runoff . The analytical results show that the precipitation of snow melt period had little on the runoff of current month because the coefficient of correlation between the area precipitation from March to June obtained by Thiessen Polygon Method and the runoff in the same period was only 0.02,0.10,0.31 and 0.51 respectively for each month . The main heat comes from solar radiation first the sensible heat and warm air advection second .Under certain circumstances the air temperature is one of the main factors affecting spring runoff.

  2. Simulation and Forecasting of Spring Runoff2

    1. Selection of reference stations and forecasting factors. To forecasts each ten days inflow (Y) to Longyang Gorge Reservoir from the first ten days of April to the first ten days of June the Principal components analysis and Cognate Degree analysis were used to select reference stations and forecasting factors the result are as follows.

      X1 the mean snow cover percentage (MSCP) (%) of study area in Oct. of last year; X2 the MSCP ten day early before issuing forecast; X3 the mean discharge (m3/s) ten day early before issuing fore cast at Tangnaihai hydrological station X4 the mean max temperature (0C) in Oct of. Last year at the Guoluo Meteorologic Observatory (GMO); X5 the precipitation (mm) is Oct. of last year at GMO X6 the mean max. temperature (0C) ten day before issuing forecast at GMO; X7 the precipitation (mm) ten day early before issuing forecast at GMO. The calculated results of the cognate order are listed in Table -2

      Table -2. the results of cognate degree and cognate order.
      Factor Forecasted Forecasting factors
      Item Y X1 X2 X3 X4 X5 X6 X7
      Cognate degree 1 0.69 0.78 0.87 0.57 0.62 0.46 0.67
      Cognate order 3 2 1 6 5 7 4

    2. Mid range forecast of spring runoff and its verification According to the need of user and actual possibility , lots of spring runoff forecast models were developed with snow information from NOAA and hydrometeor logic data by the use of mathematical statistics grey theory variance analysis The estimated coefficient value of ridge Regression Model (RRM) is

      B(K0) = 1(X'X + K0I)-1 * X'Y

      Where: I-the unit matrix; Ko-Ridge parameter matrix; B(ko-Ridge mark matrix that can be calculated by the following formula:

      Where i = 1,2…………,p;
      p ----- the number of the forecasting factors; lj and Sij are the characteristic value and eigenvector of X'X augmented matrix respectively. The forecasted value is



      Where : X the forecasting factor ; B(ko) Rige regression coefficient;
      e --- The residual.

      The RPM was used for forecasting the inflow of each ten day to Longyang Gorge Reservoir from April to the first ten days of June the verification results by measured data of 1989 and 1990 show that the forecast precision is within the permitted error range and can meet the need of user perfectly.

    3. Simulation of spring runoff under the support of ISWRIS in the typical area of the upper reaches of the Yellow River the Qushain basin the first order stream taking its source in Anyemaqen Mt. was divided in to three elevation zones zone a( below 3900 a.m.l) zone B (3900 -4650m a.m.l ) and zone C (above 4650m a.m.l ) according to elevation and the characteristics if snow . The snow cover percentage in each zone was extracted from AVHRR data by means of NAC -4200F Then the parameters and coefficients of selected Snow melt Run off Model were adjusted based on the actual conditions in study area and finally the following models were established in Qushian stream which could calculate the mean discharge un April and from the last ten days of March to the first ten days of June respectively:

      Qt (l-k) (- 143.39 + 0.345PA + 1.21SA + 0.3PB+ 0.72SB + 0.5SC - 12.98TC)

      Where Qt-the mean discharge (m3/s) from the last ten days of march to the first ten days of June at Damitan hydrological station in Qushain stream ; k-the retreat coefficient ( assigned to be 0.5 ) of discharge ; P-the precipitation (mm) at Guoluo meteorologic station : SA the snow cover percentage (SCP) (%) in zone A; PB the precipitation (mm) at Renxiamu meteorologic station SB the SCP in Zone B; SC the SCP in zone and TC---the air temperature (oC) of 500 hpa at Germu meteorologic station.

      Qt (l-k) (- 48.5 - 0.47SA + 0.3PA - 0.31SB - 0.2TB -0.213SC)

      Where Q4 - the mean discharge ( m3/s) at Damitan hydrological station
      k-the retreat coefficient (taken to be 0.2) of discharge;
      TB the daily mean temperature (oC) at Renximu meteorologic station and the rest of the symbols are same as in formula (8)

      The test by measured data shows that calculation precision of zoning is increased notably.(Table3)

Table 3. The calculation precision of zoning and unzoning.
Year Period Measured calculated value calculated value error (%)
  Data
(m3/s)		 (zoning)
(m3/s)		 (unzoning)
(m3/s)		 Zoning unzoning
1988 20 March 20.6 20.3 16.5 1.4 19.7
1989 to 10th June 31.7 30.9 23.4 2.4 26.0
1988 April 13.9 11.2 9.7 19.2 30.6
1989 April 11.6 10.8 13.8 7.1 18.9

Acknowledgements

The other persons who were involved in some of the field expeditions and indoor data analysis are Li Wenzhong , Zhang Shunying , Tang Han , Liang Fengxian , wang Guangyu Liu Jinhung Pu Yibin and the others The satellite Meteorology center state Meteorological Administration the Environment Forecasting Centre of national oceanic Administration the Meteorological Bureaus of Gansu and Qinghai Provinces provided meteorological data and the other relative data. The Longyang Gorge water power Plant gave the assistance with the testing and verifying of spring runoff forecast .All the acknowledgements are given to those described above.

Reference
  1. Zeng Qunzhu and Zhang Shunying, 1987, Satellite Snow Cover Monitoring in the Qilian Mountains and an Analysis for Characteristics of Stream Snowmelt Runoff in the Hexi Region, Gansu, China. Annals of Glaciology, International Glaciological Society. (1987).

  2. Martinec J. and Rango A., 1979, Application of Snowmelt Runoff Model Using Landsat Data. Nordic Hydrology 10, pp.255-238.

  3. Hall K. and Martinec J. 1985, Remote Sensing on Ice and Snow Publised in the USA by Chapman and Hall Ltd. (1985)