Bathymetric Change Study
Bathymetric changes of the Great Bay Estuary over the last 100 years
In this project we analyze and compare data sets, gathered in 1913 and 1953-54 from Great Bay Estuary, located near Portsmouth, New Hampshire (Figure 1). This implies that we have to consider the problem of combining multiple hydrographic data sets collected over many years with different techniques and standards and, therefore, various associated errors. The purpose of the comparison is to see if it is possible to determine significant changes over time in the bathymetry of the Great Bay area, and to develop a robust protocol for hydrographic data comparison while accounting for the possible random errors in the source data.
The two data sets were digitized from hydrographic survey sheets, so called smooth sheets. The comparison of the two data sets (1913 and 1953-54) was made by constructing Triangular Irregular Network (TIN) models using Z/I Imaging's tool Terrain Analyst (MTA) for each (Figure 2) and then computing point-wise differences between the TIN models and the original data points (Figure 3).
To simulate the effects of random data errors, we have repeated the comparisons using a number of hypothetical data sets derived from the real data by perturbing the digitized data points with random vectors proportional to the horizontal and vertical errors assessed for the surveys using the associated smooth sheet reports. The use of Monte Carlo modeling for simulation of the random error component in bathymetric surveys is explored by Jakobsson et al (JGR, in press). Subsequently we compute statistics of the depth differences, with the objective of showing where the mean difference is significantly far from zero (i.e., where the difference observed is likely to be due to real changes, rather than measurement errors).
Figure 4 shows a simple straight comparison between the TIN model created from the 1953-54 data and the data points from 1913.
In order to analyze if this comparison is significant rather than measured errors the Monte Carlo results were used to compute the standard deviation of the depth difference between the 1953-54 perturbed TIN models and the 1913 data points (Figure 5). These results were presented during the Shallow Survey 2001 conference organized by CCOM/JHC (Jakobsson et al., 2001).