Date of Award

Winter 2014

Project Type


Program or Major

Earth Sciences

Degree Name

Doctor of Philosophy

First Advisor

Larry G. Ward

Second Advisor

Susan Adamowicz

Third Advisor

David M Burdick


High-resolution and high-accuracy elevation data sets of coastal salt marsh environments are necessary to support restoration and other management initiatives, such as adaptation to sea level rise. Lidar (light detection and ranging) data may serve this need by enabling efficient acquisition of detailed elevation data from an airborne platform. However, previous research has revealed that lidar data tend to have lower vertical accuracy (i.e., greater uncertainty) in salt marshes than in other environments. The increase in vertical uncertainty in lidar data of salt marshes can be attributed primarily to low, dense-growing salt marsh vegetation. Unfortunately, this increased vertical uncertainty often renders lidar-derived digital elevation models (DEM) ineffective for analysis of topographic features controlling tidal inundation frequency and ecology. This study aims to address these challenges by providing a detailed assessment of the factors influencing lidar-derived elevation uncertainty in marshes. The information gained from this assessment is then used to: 1) test the ability to predict marsh vegetation biophysical parameters from lidar-derived metrics, and 2) develop a method for improving salt marsh DEM accuracy.

Discrete-return and full-waveform lidar, along with RTK GNSS (Real-time Kinematic Global Navigation Satellite System) reference data, were acquired for four salt marsh systems characterized by four major taxa (Spartina alterniflora, Spartina patens, Distichlis spicata, and Salicornia spp.) on Cape Cod, Massachusetts. These data were used to: 1) develop an innovative combination of full-waveform lidar and field methods to assess the vertical distribution of aboveground biomass as well as its light blocking properties; 2) investigate lidar elevation bias and standard deviation using varying interpolation and filtering methods; 3) evaluate the effects of seasonality (temporal differences between peak growth and senescent conditions) using lidar data flown in summer and spring; 4) create new products, called Relative Uncertainty Surfaces (RUS), from lidar waveform-derived metrics and determine their utility; and 5) develop and test five nonparametric regression model algorithms (MARS - Multivariate Adaptive Regression, CART - Classification and Regression Trees, TreeNet, Random Forests, and GPSM - Generalized Path Seeker) with 13 predictor variables derived from both discrete and full waveform lidar sources in order to develop a method of improving lidar DEM quality.

Results of this study indicate strong correlations for Spartina alterniflora (r > 0.9) between vertical biomass (VB), the distribution of vegetation biomass by height, and vertical obscuration (VO), the measure of the vertical distribution of the ratio of vegetation to airspace. It was determined that simple, feature-based lidar waveform metrics, such as waveform width, can provide new information to estimate salt marsh vegetation biophysical parameters such as vegetation height. The results also clearly illustrate the importance of seasonality, species, and lidar interpolation and filtering methods on elevation uncertainty in salt marshes. Relative uncertainty surfaces generated from lidar waveform features were determined useful in qualitative/visual assessment of lidar elevation uncertainty and correlate well with vegetation height and presence of Spartina alterniflora. Finally, DEMs generated using full-waveform predictor models produced corrections (compared to ground based RTK GNSS elevations) with R2 values of up to 0.98 and slopes within 4% of a perfect 1:1 correlation. The findings from this research have strong potential to advance tidal marsh mapping, research and management initiatives.