Award Date


Degree Type


Degree Name

Master of Science in Engineering (MSE)


Civil and Environmental Engineering and Construction

First Committee Member

Sajjad Ahmad

Second Committee Member

Haroon Stephen

Third Committee Member

David James

Fourth Committee Member

Pushkin Kachroo

Number of Pages



This thesis estimates the relationships between the water surface levels and quantities of water in reservoirs and rivers using remotely sensed data without any field measurements. However, the accuracies of the estimates were validated using the field measurements from ground stations. The relationships between the water surface levels and quantities are fundamental for monitoring water quantities for the operation of hydraulic structures, as well as analyzing variability and changes in hydrology. Accessibility and transparency are big issues in establishing a monitoring system that can provide field validation of efforts to model global climate systems. Remote sensing has a capability of global spatial coverage and stable temporal frequency in data acquisition, and hence can be helpful. Two types of remotely sensed data are used: satellite images, and satellite altimeter elevations. The thesis has two parts.

First, the relationships among water surface levels, areas and volumes were estimated for reservoirs. A strategic procedure, which is missing in the current literature, was formulated to estimate the water surface area, and then the water volume. Water levels were derived from Hydroweb, a satellite altimetry database. Areas were estimated from Landsat surface reflectance images by classifying the Modified Normalized Difference Water Index (MDNWI) into binary images using an internally calibrated threshold. Internal calibration of the threshold was performed by computing the overall accuracies of classification from confusion matrices created for selected regions in the classified image. Finally, water surface heights from the lowest levels and areas were used to estimate volumes assuming an inverted pyramidal shape; then, second-order polynomials were fitted to compute relationships. The fits were tested to be statistically significant by performing t-tests for coefficients and F-test for overall significance at α = 0.05. Stage-area-storage relationships were developed for Lake Mead (LM) and Lake Powell (LP) that are reservoirs formed in the Colorado River. The study estimated the areas of LM with a Root Mean Square Difference (RMSD) of 17.8 km2 and LP with an RMSD of 53.7 km2 compared with in-situ measurements. The RMSD in volumes were 699 Million Cubic Meters (MCM) for LM and 1330 MCM for LP. The second-order polynomial fits between water surface heights and volumes were established with R2 = 0.999 for both LM and LP. The coefficients of the fit and the overall fit were tested to be statistically significant at α = 0.05. The RMSD is higher in LP than LM and were explained by comparatively more shadows and a higher number of mixed pixels in the LP Landsat images than LM.

Secondly, the relationships between the water surface levels and discharges for rivers were estimated. Two major rivers, the Mississippi River and the Colorado River, representing an alluvial and rocky terrain, were selected to highlight the differences in estimates between varied terrain and size of the river. A variant of Manning’s equation was used that required a channel cross-section, water surface slope, and roughness coefficient as input parameters. A parabolic cross-section was fitted for each river using the width of river estimated from the Landsat images at several water levels. Water surface slopes were estimated from water elevations at different locations on each river using two sources. For the Mississippi River, water surface elevations were obtained at virtual stations from the DAHITI database. For the Colorado River, elevations were extracted using the MAPS at river crossings. Roughness coefficients were estimated using empirical models that utilized meander length. Results showed that discharges were estimated to within 31.4% of the average discharge with root mean square error of 5700 cu.m/sec for the Mississippi River. Colorado River discharges were estimated within 30.5% of the average discharge with RMSE of 50 cu.m/sec. A linear relationship was fitted between the water surface elevation and discharges in the Mississippi River with R2 = 0.62. For the Colorado River, second-order polynomial was fitted for a relationship between water surface elevations and discharges with R2 = 0.99. The coefficients of the fits and the overall significance of the fit were statistically significant at α = 0.05 tested by performing t-tests and F-test respectively. It was difficult to estimate a cross-section for rivers with smaller channel widths or smaller changes in width with water level as in the case of the Colorado River. However, estimated accuracies were similar in both the cases in terms of percentage of error.


Discharge estimation; Landsat images; Remote sensing; Reservoirs volume estimation; Satellite altimetry; Satellite images


Civil Engineering | Environmental Engineering