You are on page 1of 9

Advanced Geographic

Information Systems
_______________________
Watershed Analysis
On the Tibetan Plateau

Author: Austin Gates


Ti
bet
anPl
ateau Hyd roSHEDBas i
n&Ri
ver Net
w ork
LambertAzimut halEq ual
Area 0 30 60 120 180 240
Cart
ography by Aus t
i
nGat es Ki
lomet
ers

Lakes
Ri
ver Net
w ork
Bas i
n

TheTibetanPl ateau is home


tothew orl
d’s l
arges tgroup
ofhigh-al
ti
tud eend orheic
lakes .

µ Dat
aSourc
es :
UCLACCLE,USGS
DEM-Based analysis can be used to automatically delineate watershed boundaries and
extract drainage networks in hydrology. This process is referred to as watershed analysis.
Watershed analysis is important in understanding the balance of water levels within an area and
is especially useful when studying lake levels in a topographically uneven areas.
This paper will derive and analyze drainage basins and networks on the Tibetan Plateau,
one of the largest groups of high altitude lakes in the world. Using a hydrological process in
ArcMap, the final products will then be compared to USGS watershed products to gage the
accuracy of the derived process.
The data used in the paper came from two sources. The DEM for the area as well as a
shapefile for the lakes was acquired on the UCLA Geog 170 CCLE class website. These will be
used for the derivation. The USGS basin and drainage network shapefiles were acquired on
USGS hydroshed webpage, and are in 15 arc second format. In order the begin the derivation
process, only the DEM is required. The first step is to derive basins from this DEM. First, run the
Flow Direction tool on the DEM and then run the Basin tool on that output raster. Doing so
produces a Basin raster than can be converted to polygons with the Raster to Polygon
conversion tool. As shown in figure 1a, this produces much too many drainage basins. To solve
this issue, the DEM must be filled or corrected for errors in the DEM or low water levels in the
DEM. In order to define an appropriate Z value for the fill tool, four separate fills each with a
unique Z values were performed, and 4 different drainage basin polygons produced. Figure 1 is
a comparison of the different fill values ran on the original DEM. After comparing these to the
HydroSHED USGS Data, it became apparent that a Z value of 40m was the most appropriate.
The next step involved deriving flow accumulation rasters from the flow direction. Using
the 40m flow direction raster, the Flow Accumulation tool is then ran. This produces a raster with
pixel values between 0 and 2.9 million. Flow accumulation raster values represent the total
number of cells in the surrounding raster are flowing into that specific cell. Due to the nature of
the raster, there are very few cells with a value of zero. Figure 2a illustrates problems with using
the original accumulation raster. Because of this huge amount of data, it is necessary to filter
out some of these smaller valued pixels to apply a symbology that is legible. Using the Con tool,
it is possible to change the accumulation raster into a binary raster based on the accumulation
values. Once again, running this step multiple times to compare binary accumulation rasters is
required to decide on a correct threshold for the conditional tool. The Con tool has the following
inputs: the flow accumulation raster previously generated, a SQL expression reading ‘VALUE >
threshold’, an input for the true raster values, and an input for the false raster values. Those
input values can be can be either rasters or constants. Because the following steps do not
require the true value of the original raster, plugging in 0 for the false values and 1 for the true
values results in a binary raster with values of 0 or 1. In the SQL expression, using threshold
values of 1000, 2000, 3000 and 4000 generates four legible accumulation rasters. The Stream
To Feature tool can then be ran on new accumulation rasters to generate polylines to represent
the river network. Figure 2 is the different flow accumulation thresholds ran on the flow direction
raster.
After comparing these to the USGS hydroSHED river network, the 4000 threshold raster
appeared to be the most accurate of the thresholds. Now both drainage basins and a river
network has been derived using only a DEM. Further information can be gather from the derived
river network. Running the Stream Order tool with the Strahler method assigns a numeric order
to segments of the raster representing branches of a linear network. All streams begin with a
value of one. Stream order only increases when streams of the same order intersect. Figure 3
illustrates the final derived drainage basins overlaid on the filled DEM along with the lake
shapefiles. Directly next to it, Figure 4 is a map of the DEM derived river network along with the
boundaries and the lakes. The stream order shapefile provides visual representation to the
natural flow of water and the stream network in these mountains.
Figure 6a and 6b contain the derived basins and river network overlaid with the USGS
Hydroshed products. The similarities are very apparent in the overlay map. There are also areas
with major differences. These fluctuations can be attributed to differences in input data as well
as differences in procedures between the two process. It is easy to see that the USGS product
is more precise in general. The size of the files that their processors can handle is much higher
than the capabilities used in this paper. USGS use a corrected DEM that averages 1 arc second
and 3 arc second data, to correct for errors in the acquired data. (usgs.gov) The DEM used in
this paper had 100m spatial resolution which could also be a limiting factor. USGS also utilizes
more advanced algorithms for filling voids in DEMs, such as the CIAT algorithm for filling the
DEM. (usgs.gov)
The processes described above do provide an accurate model for hydroshed analysis,
but the USGS process provides a more appropriate model for deeper analysis. In order to
improve this process, more Z fills should be tested and more thresholds should be ran to
compare with the USGS process. Using higher spatial resolution would also improve the
preciseness of the model. A more advanced algorithm should be used to fill voids in the DEM.
Finally, maintaining the accumulation values in the raster can illustrate areas where more water
will theoretically flow through.
Figure 1b. Figure 1d.

Z value = 10 Z value = 30

Deriving Drainage Basins


on the Tibetan Plateau
Comparing Z Limits used to "fill" a DEM when
deriving watershed drainage basins.
Let Z = the maximum vertical distance in meters
between DEM pixels to be filled

Figure 1a.
Original DEM

0 70 140 280 Kilometers 0 70 140 280 Kilometers

Figure 1c. Figure 1e.


Z value = 20 Z value = 40
0 62.5 125 250 Kilometers

Lambert Azimuthal Equal Area


Cartography by Austin Gates
Data Sources: UCLA CCLE, USGS

Legend
µ
Elevation Lakes
Value (m) Basin Boundary
High : 6398

Low : 4448
0 70 140 280 Kilometers 0 70 140 280 Kilometers
Comparing Different Flow
Accumulation Thresholds on a
40m Z Filled DEM

Legend

µ
Flow Accumulation

Elevation
Value (m)
High : 6398

Low : 4463

Figure 2a. Flow Accumlation without a threshold.

Figure 2b. Flow Accumlation with a 1000 pixel threshold. Figure 2d. Flow Accumlation with a 3000 pixel threshold.

Figure 2c. Flow Accumlation with a 2000 pixel threshold. Figure 2e. Flow Accumlation with a 4000 pixel threshold.
0 70 140 280 Kilometers
Data Sources: UCLA CCLE, USGS
Lambert Azimuthal Equal Area - Cartography by Austin Gates
DEM Derived Basins with Lakes
Legend
Elevation Lakes
Value (m) Basin Boundary
High : 6398

Low : 4463

DEM Derived River Network with Lakes

²
Figure 3
0 50 100 200 Kilometers

Legend
Stream Order Lakes
1 Basin Boundary
2
3
4
Lambert Azimuthal Equal Area
Data Sources: UCLA CCLE, USGS 5
STRAHLER Method of Stream Ordering
Cartography by Austin Gates
Figure 4 0 50 100 200 Kilometers
Tibetan Plateau River Network Comparison
HydroSHED vs Derived
Data Sources:
Cartography by Austin Gates - Lambert Azimuthal Equal Area UCLA CCLE, USGS

Figure 6a.

Inset 1
0 10 20 40 Kilometers

µ
0 35 70 140 210 280
Kilometers

Derived River Network Map Inset


HydroSHED River Network

Inset 2
0 15 30 60 Kilometers
Tibetan Plateau Basin Comparison
HydroSHED vs Derived
Data Sources:
Cartography by Austin Gates - Lambert Azimuthal Equal Area UCLA CCLE, USGS

Figure 6a.

Inset 1
0 40 80 160 Kilometers

µ
0 35 70 140 210 280
Kilometers

Lake Map Inset


Derived Basin
HydroSHED Basin Inset 2
0 25 50 100 Kilometers

You might also like