Basin Summaries

All of the following basins had at least one discrepancy between calculated and reported area for a flow gauging station. The references and methods used can be found in the comment.
Data files will be available soon for those basins where the Graham et al. delineated areas are larger than actual areas. No files will be provided for those where actual areas are smaller. For these basins it is recommended that the basins be reconstructed from fine resolution DEMs.

The format for these gauges is as follows:
For each gauge, the area estimate and true area are given along with a possible (hopefully the most likely) explanation of why (qualitatively) this is the case. Then the method used to correct the error is given.

Severnaya Dvina

The problem:

This gauge is located on a channel connecting two outflows of this river system - flow divides in the downstream direction with part flowing in the main stem (Severnaya Dvina) and another portion flows north to the baltic sea (Kuloy). Since flows in these two channels are not readily accessible, the division of flow is assumed to follow the division of area. The Pinega-Kuloy system contains approximately 20% of the basin area (by eye). This subsystem is also father from the gauge than the Severnaya Dvina main branch. Also the eastern extent of the basin of the delineated basin is greater than the actual extent (~60deg delineated vs. ~56 actual). Presumably these are the causes for the area discrepancy.

The solution:

Remove those portions of the grid file that are in the Pinega-Kuloy system and that are on the eastern edge of the basin, recalculate a flowaccumulation file from the clipped flow direction file and determine the new area.

***clipping method***

To insure that the new clipped basin is in fact one single basin with one outlet (as opposed to several new smaller basins with multiple outlets), the clipped basin is run through the Arc/Grid basin delineate macro - which defines each basin by number. If there are multiple basins, the intended basin (usually the largest) is selected, used as a mask file for reclipping the fdr, and a new single basin fac is created. The single basin fac and fdr are then used with accarea to insure that the area is correct. This final fdr will be used as VIC input. Aside good clips are those that don't result in large sub-basins; this is used as a qualitative assessment of the clipping process.

Results:

After two iterations. Goodclip. Dvinafdr is the Arc/Info file that contains the corrected gridfile. Ultimately all will have to be recombined into a world 5min fdr.

Anaydr

The problem:

This watershed includes portions of three other watersheds: the Kanchalan and Velikaya on the east and the Peuzhina system on the west. The eastern watersheds were obvious, but the western one required some trial and error because of flat topography.

Results:

After 3 iterations. Goodclip. Anafdr is the Arc/Info grid file containing the corrected gridfile.

Yukon

The problem:

The most upstream gauge overestimates area, and the next downstream underestimates it. However the net effect is not apparent from the next two downstream gauges, which encompass most of the basin, where upstream area is less than 2% off.

The accarea derived file shows the main channel flowing almost exclusively east to west in the approximately 2 by 2 degree window created with gridwindow. The map from RivDis and the available maps in the Nat'l Geographic Atlas. This suggests that the original DEM data or the integration into Arc/Info was poor. There is no location along the river system upstream of this area that will produce the same drainage pattern with the correct area upstream. Therefore simply moving the gauge is not an option. A second potential source of the problem (though not the cause of the erroneous flow direction at the location of the gauge) is the upstream most area in the southern fork upstream from the gauge. It appears that slightly more area is included than should be.

The second gauge has a more intractable problem. The flow direction immediately to he east of the gauge has a northern draining pattern in the fdr, but is in fact draining to the west in a tributary upstream of the gauge. Another problem is the area to the immediate north east of this station. On the map it drains into the Peel river (part of the MacKenzie system), but in the fdr it is part of the Yukon.

Results:

After 1 iteration. Goodclip. Yukfdr is the Arc/Info grid file containing the corrected gridfile.

Correcting the area for the Carmacks gauge made the area at Dawson worse. However, the cause of the underestimation at Dawson is not readily fixed. The recommendation is to use the other two upstream gauges as the internal validation points and simply write off the Dawson gauge for this project. It does not seem much is lost in this basin from the exclusion of this gauge.

Nelson

The problem:

Estimated area is slightly above the recorded area (already suspect since it is given at potentially one significant figure). The excess are is most likely due to the very flat topography to the east and glacial lake region to the north.

Results:

After 1 iteration. Goodclip. Nelfdr is the Arc/Info grid file containing the corrected gridfile.

Amur

The problem:

As delineated, this grid contains portions of the Taal and Orhon river systems at the extreme western edge and portions of the Kerulen river system to the southwest. Also much smaller regions to the north and northeast drain into other systems or directly into the ocean. These regions here, as elsewhere when they are discovered, are not clipped unless their extent is large enough to have an impact on the total gauge area ( >1-2% of total).

Results:

After 1 iteration. Goodclip. Amufdr is the Arc/Info grid file containing the corrected gridfile.

Saint Lawrence

The problem:

Headwater portions of the Rock and Wisconsin rivers are included in this delineation. This area is immediately west of Lake Michigan.

Results:

After 1 iteration. Goodclip. Stlfdr is the Arc/Info grid file containing the corrected gridfile

Frasier

The problem:

Too much area upstream, to little down. The Thompson is missing from the basin, and the Peace is included. An investigation of the map show that a large area to the north has been incorrectly delineated in the Fraser drainage instead of the Winston Lake/Peace and Skeena systems. This large area has been removed. And the Thompson was added.

Results:

After 1 iteration. Goodclip. The area correction did indeed correct the areas.

Volga

The problem:

As delineated this basin is clearly missing a great deal of area - the calculated area above is from the outlet at 46.12, 47.95, not at the above location. Also, the given location is above the outlet of the system not at the outlet as it is in this file.

Also, he location of the outlet in the Graham et al. file is not at the location of the outlet taken from a map (48, 46). When this basin is placed in the context of the surrounding basins (see figure in appendix), it's clear the missing area is from the northwest (not in any basin) and the northeast (the area excluded in correcting the Severnaya Dvina system). This combination may sum to the missing area. The actual adding of this area has to be done by altering the flow direction file.

Results:

After 1 iteration. Goodclip. Volfdr4 is the Arc/Info grid file containing the corrected gridfile. One basin delineation iteration was needed; however, many gridedit iterations were needed to insure that the new areas did indeed contribute to the existing flow system. The resulting fdr and fac were less than realistic looking (see appendix).

Recreating the basin using 30 arc second DEM data (GTOPO30) was able to create the actual drainage system at the 30 arc second resolution. However, any attempt to aggregate up to 5 minute resolution left large portions outside the basin - the same areas excluded in the Graham data set. The smallest scale that captures the true drainage is 1 minute; although, even at this scale the outlet is in an incorrect upstream location.

Don

The problem:

The areas in the northwest and along the coast of the Sea of Azov are clearly not part of the drainage. The coastline of the sea is most significant at the westernmost edge in the drainage (cut from the fdr) and the northwest is actually in the Oka(Volga) basin. These two areas easily constitute ~12% of the area upstream of the gauge.

Results:

After 1 iteration. Goodclip. Donfdr is the Arc/Info grid file containing the corrected gridfile.

Huange He (Yellow)

The problem:

A huge portion of the northwest part of this basin is in the Gobi desert (Ruo and Shule terminal lake systems), the southeast is part of the Ying-Huai system. The northeast includes parts of the Sanggan/Hutuo/Ziya/Fuyang systems. All of these areas taken together are enough to result in the errors above.

Results:

After 2 iterations. Goodclip. Huafdr is the Arc/Info grid file containing the corrected gridfile.

Colorado (Mexico)

The Problem:

The larger area at the downstream gauge is due to the inclusion of the ephemeral and developed Gila river system. The smaller upstream area results from missing a portion of the Green River (including Flaming Gorge Reservoir).

Results:

After 1 iteration. Clipping of Gila river system resulted in "jagged" looking southern basin boundary. Here, as everywhere else where area was ADDED to the system, extreme caution must be taken when using the coverage. Only the area estimate and a general atlas were used to improve the delineation. No attempt was made to reconstruct actual drainage patterns. That effort will follow after this project is completed.

Mississippi

The problem:

The area listing for the Tarbert Landing is wrong. This is good because to have missed that much area between the last two gauges would have required a rather peculiar looking addition (and one not justified by the topography). Areas were looked up in two reputable sources: USGS and The Water Encyclopedia.

Results:

1 - USGS station near Tarbert Landing(Lon: -91.66 Lat:30.66)

2 - Water Encyclopedia whole basin area

After zero iterations. The USGS area is the best readily available surrogate for the gauge area.

Shatt-al-Arab (Tigris/Euphrates)

The problem:

A large portion of the Syrian Desert is included in this delineation to the south and southwest. Also included is Van Golu in the north, which contributes to the gauge at Mosul.

Results:

After 5 iterations. Goodclip. Removing the Syrian Desert area had little impact on the gauge areas. The removal of Von Golu proved to be difficult due to the drainage patterns in the north of this watershed.

Rio Grande

The Problem:

Large arid areas to the southwest of Big Bend National Park are included in this drainage, as are smaller areas to the west. The southwest area either is outside the drainage or (from terminal and ephemeral streams on the atlas) produces no runoff. This is confirmed by the World Atlas of Geomorphic Features, which could be used to justify a much smaller basin. Making the delineation smaller makes no sense since the recorded area estimate sets a practical lower bound on basin area.

Results:

After 3 iterations. Goodclip. Riofdr is the Arc/Info grid file containing the corrected gridfile.

Indus

The Problem:

A large portion of the eastern side of the basin belongs to the Bramaputra (possibly explaining its lower than recorded value). Also a large part of the Great Indian Desert is included where it shouldn't be.

Results:

After 3 iterations. The Bramaputra region had to be extended to fit the Attock gauge. It is within reason, however. Goodclip. Indfdr is the Arc/Info grid file containing the corrected gridfile.

Brahmaputra

The problem:

The headwaters of the Brahmaputra are in incredibly narrow basins and were too wide, which explains the larger than normal area for the Panda gauge. The reason for the lower gauge's underestimation of area is unknown. Most likely it is simply incorrect. This basin (the Brahmaputra) is bounded by other catchments that have correct (to within 10%) areas. It is possible that these other basins are slightly too large. But this "excess area" would have to be added below the Panda gauge.

Another problem with this basins was the nodata values within its borders.

Results:

After 1 iteration. Goodclip. The downstream area estimate is now worse than before, as would be expected by the upstream improvement in area.

Senegal

The Problem:

The basin includes a large portion of the western Saraha desert below the Galougo gauge. This area either does not drain into the Senegal, or it does so only in small ephemeral streams.

Results:

After 2 iterations. The second iteration was done to minimize the clipped area upstream of Galougo. Goodclip. Senfdr is the Arc/Info grid file containing the corrected gridfile.

Irrawaddy

The Problem:

The gauge was located below the confluence instead of above it. This was determined by finding the town of Sangaing on an Atlsa and assuming that the gauge was located in/near that city.

Results:

The correction of the area error strongly suggests the validity of the above assumption.

Salween

The Problem:

The headwaters erroneously include a mountainous lake region in the Tibetan Plateau. This region is also near the headwaters of the Chang Jiang (Yangtze) possibly explaining it's slight (and uncorrected here) underestimation of area.

Results:

After 1 iteration. Goodclip. This area was indeed responsible for the mismatch.

Niger

The Problem:

Like the other major rivers adjacent to arid lands, the Niger basin includes parts of the southwestern Sahel region (part of Saraha?? internally draining, anyhow ... lake Chad).

Results:

After 2 iterations. The second iteration was done to because the first left the areas for Malanvilla and Gaya too small. Goodclip. Nigfdr is the Arc/Info grid file containing the corrected gridfile.

Zaire (Congo)

The problem:

The basin includes Lake Victoria and a region south of the lake.

Results:

After 1 iteration. Goodclip. These two areas were responsible for the mismatch in area.

Victoria

The problem:

This drainage includes many ephemeral streams to the south - in fact the headwaters of the Victoria itself are ephemeral. Also this region is in the Tanami Desert. Only the gauge area provides a clue as to the practical extent of this basin.

Results:

After 1 iteration. Goodclip. Vicfdr is the Arc/Info grid file containing the corrected gridfile.

Section 2.

Two methods will be used for the remaining six (the second if time allows). The first is to simply add very crude flow direction cells to the existing flow direction files. This process is iterative and wholly interpretive. Adjacent basins will be used to clip areal extent of these added area when they exist (so that no area will be counted towards two watersheds). These will consist of a single "main channel" and the surrounding area will drain directly to it (all non-channel topography will consist of perfectly flat planes). The main channel will be located in such a way as to capture the general trend of the physical system to the nearest 45 degree angle. The second method will be to find another DEM source for the basins in question and re-delineate the basins from this other source. This would have the advantage of being reproducible, and it would hopefully give better results .... esp. if the DEM is at a finer resolution. This will take longer and may not be done satisfactorily for that reason. Thanks to Keith and Laura for the idea for the second method.

Parana (Plata) and Uruguay

The problem:

The Guaria and Posadas gauges are upstream of the Corrientes on the same major tributary(Parana/Paranaiba). This suggests that the the other main trib area is too large while the gauged trib is too small resulting in the Corrientas gauge having a relative error of less than 10 percent.

Results:

1- area value taken from RivDis

After 1 iteration. Goodclip. parfdr is the Arc/Info grid file containing the corrected gridfile. The correction of the upstream gauges made the downstream area worse; however, the downstream area estimate from RivDis shows an improvement in the accuracy of the calculated area. If for no other reason than consistency, this value should be used.

Murray

The problem:

The basin area given contains areas which do not produce runoff during normal years. This fraction of the total area is 2/3 (see below), so a realistic value of runoff producing area for this basin is: 33033km2 resulting in a percent error of -7.47 which is within 10%.

The following is taken from: http://www.murray-river.net/facts.htm

/*********omit information below and just use the reference?? ***********/

This information was compiled from:

* Who owns the Murray

Peter S. Davies

* The SA Recreational fishing guide

* Groliers Encyclopedia

* The Murray River Charts

MURRAY RIVER

Murray River Facts and Figures

Size

The River Murray Basin of 1,057,000 square kilometers is about one-seventh of the total area

of Australia and comprises three-quarters of New South Wales, over one-half of Victoria, a

small portion of South Australia, and an area of Queensland greater than the total area of

Victoria.

Rainfall

Of the above area, only five percent has an average rainfall in excess of 760 mm and nearly

two-thirds normally supplies no water to the two trunks. The catchment of the upper Murray

above Albury contributes more than one-quarter of the total flow in the Murray system, from

an area which is less than two per cent of the catchment area.

The average rainfall over the Murray basin is 430 mm compared with an average of about 1270

mm in the Tennessee Valley, USA.

Flow

The grade or fall of the Murray varies from 14 cm per kilometer near Albury, to about 2.5 cm

on the last 160 kilometers to the sea.

The highest point of the Murray watershed is Mount Kosciusko, with an altitude of 2231

meters.

Since gauging began at Swan Hill in 1909, the Murray ceased to flow at that point for short

periods in 1914, 1915, and 1923. In April 1915 no flow was recorded for a whole month.

Since the Hume Reservoir was built in 1936 a flow has been maintained throughout the length

of the Murray at all times, despite several severe drought periods.

Length

The length of the Murray from its sources to its mouth is about 2560 kilometers, including

2185 kilometers below Albury. The Murray is continuously navigable for 1986 kilometers from

Goolwa to Yarrawonga Weir. In this length there are thirteen weirs incorporating locks.