Mekong River
Both the MODIS and TRMM have advantages in their ability to map surface water for the Lower Mekong River. The MODIS pixel size of 500 m x 500 m shows lots of detail that is not possible to see in the TRMM 10 km x 10 km pixels. However, the TRMM data was operating from late 1997 enabling it to capture the dry year of 1998 as well as the wetter years of 2000, 2001 and 2002.
The MODIS and TRMM data points were combined to compare results, and determine the best relationships to use with the modelled water volumes. Figure 12.12 shows the TRMM (1998-2002) and MODIS (2000-2002) monthly flood extent for the Tonle Sap Lake, plotted against the modelled Tonle Sap water volumes.
y = 0.1532x - 1193.5 R2 = 0.8331 y = 0.0896x + 277.78
R2 = 0.7892
0 2000 4000 6000 8000 10000 12000
0 20000 40000 60000 80000 100000 120000 Tonle Sap Storage (mcm)
TRMM flood extent (km2)
0 2000 4000 6000 8000 10000 12000 14000 16000 18000
MODIS flood extent (km2)
TRMM MODIS Linear (MODIS) Linear (TRMM)
Figure 12.12. Scatterplot of TRMM (1998-2002) and MODIS (2000-2002) mapped flood extent verses modelled monthly storage water volume for Tonle Sap Lake.
Figure 12.12 shows similar linear relationships between the TRMM extent and modelled storage as well as the MODIS extent and modelled storage. The TRMM has a similar correlation to the MODIS with an R2 of 0.79 compared to the MODIS R2 of 0.83. There is a large difference in flood extent with the TRMM having a maximum around 10000 km2 while the MODIS is around 15000 km2.
during wetter periods, and hence limiting the number of MODIS scenes available for mapping flood extent.
y = 0.0357x + 8375.7 R2 = 0.8089
10000 12000 14000 16000 18000 20000 22000 24000 26000
150000 200000 250000 300000 350000 400000 450000 Kratie annual water volume (mcm)
TRMM flood extent (km2)
15000 20000 25000 30000 35000 40000
MODIS flood extent (km2)
TRMM MODIS Linear (TRMM)
Figure 12.13. Scatterplot of TRMM (1998-2002) and MODIS (2000-2002) annual maximum flood extent for the Delta verses modelled Kratie annual water volume.
Independent research has mapped the Tonle Sap flood extent up to 15000 km2. Given this information, it is appears that the TRMM is most likely underestimating flood extent due to mixed water/non-water pixels, while the MODIS is giving a more realistic estimate. This is expected since the MODIS data has a higher spatial resolution, and hence can map more detail. Hence, to utilize the 1998 TRMM dry year and the good correlations between the TRMM and modelled flood data, as well as the more accurate MODIS flood areas, a combined approach is adopted for estimating flood area. For Tonle Sap, the monthly data for the TRMM and MODIS for 2000 to 2002 were regressed against each other (Figure 12.14).
y = 1.5034x - 940.36 R2 = 0.8911
0 2000 4000 6000 8000 10000 12000 14000 16000 18000
2000 4000 6000 8000 10000 12000
TRMM flood extent (km2)
MODIS flood extent (km2)
Figure 12.14. Scatterplot of Tonle Sap Lake monthly maximum flood extent for TRMM verses MODIS for 2000-2002.
This regression had an R2 of 0.89, with the linear relationship shown on the graph. This relationship was used to scale the TRMM data up to MODIS so both datasets could be used to generate a relationship between modelled monthly flow data and flood extent as measured by remote sensing imagery. This produced an R2 of 0.807, and a linear relationship shown in equation 12.3.
14 . 825
* 1426 .
0 −
= x
y ………..Equation 12.3
where x is the modelled storage (in mcm) and y is the mapped flood extent from the combined TRMM / MODIS dataset (in km2).
There were fewer points available to apply a similar approach for mapping the annual flood extent for the Mekong Delta using remote sensing data (Figure 12.13). Instead an offset was calculated such that the difference between the three common TRMM / MODIS points (2000, 2001 and 2002) equalled zero. This offset (13342 km2) was then added to the TRMM data to scale it up to MODIS flood extent (Figure 12.15). When the scaled TRMM and MODIS data were combined and regressed against the modelled annual flood volume measurements the correlation had an R2 of 0.662 and a linear relationship as shown in equation 12.4.
23428
* 0318 .
0 +
= x
y ………Equation 12.4
where x is the modelled annual flood volume for Kratie (in mcm) and y is the mapped flood extent from the combined TRMM / MODIS dataset (in km2). While this relationship is not as good as the one for Tonle Sap Lake it provides the best estimate available from the remote sensing data based on the limited years available, and it utilizes the advantages gained from adopting both the TRMM and MODIS data.
y = 0.1426x - 825.14 R2 = 0.8073
0 2000 4000 6000 8000 10000 12000 14000 16000 18000
20000 40000 60000 80000 100000 120000
Tonle Sap modelled storage (mcm)
TRMM scaled and MODIS flood extent (km2)
Figure 12.15. Scatterplot of the combined MODIS (2000-2002) and scaled TRMM (1998- 2002) monthly flood extent for the Tonle Sap Lake verses modelled monthly water volume.
12.4.1. Summary
Both the optical and passive microwave remote sensing data show a relationship with modelled water volume when they are used to map flood water extent. The TRMM passive microwave data has the advantage of being unaffected by cloud cover (while not raining), as well as having data for 1998 which was a relatively dry year in recent times. However its large pixel size (10 km x 10 km) does make it difficult to address the mixed water/non-water pixels. The higher resolution MODIS data was used here to develop a set of rules for interpreting these mixed pixels. The MODIS optical imagery was greatly affected by cloud cover, particularly in the beginning of the wet season, but when combined into 8-day composites it enabled a significant amount of data to be retrieved for the Lower Mekong River. Its pixel size (500 m x 500 m) is much smaller than the TRMM data, hence the water extent could be mapped in more detail. Unfortunately there were no MODIS data for 1998 so this relatively dry year could not be included in the analysis here. The final method adopted for mapping flood water extent in the Tonle Sap Lake and the Mekong Delta, was to combine the TRMM and MODIS data to overcome each instruments disadvantages and utilize their complementary qualities.