Research into the measurement and modeling of water movement and other hydrologic processes has been identified as a primary scientific need in support of Everglades restoration. To accurately simulate surface water hydrology in south Florida, scientists must understand the variation in vegetation cover and the role vegetation plays in surface water flow, surface water removal, and water quality.

Numerous efforts to map vegetation type and/or associations have been completed recently in the south Florida region (Welch and others, 1999). However, we need to account for the effect vegetation has on surface water flows, so additional vegetation characterization has been necessary to meet the needs of hydrodynamic models being developed for the Everglades. Field and flume experiments are measuring the importance of vegetation structure and amount on resistance to flow (Lee and others, this volume). Therefore, two objectives of this remote sensing effort are the creation of land cover maps to which resistance coefficients can be assigned (Carter and others, 1999) and the derivation of biomass or density maps for dominant vegetation types, such as sawgrass (Jones, 2000). By combining information extracted from the literature and field experience with a unique combination of satellite image data, we have generated a nine-class map of vegetation types to which preliminary vegetation flow resistance values have been assigned (fig. 1). The resultant flow resistance information has been used successfully in the hydrodynamic modeling of the Southern Inland Coastal System (Swain, 1999).

Although logistical constraints, variable water levels, and changing vegetation conditions present great challenges, a set of ground truth data points are being continually augmented for use in accuracy assessment. In this heterogeneous environment, however,

Figure 2. An isotropic measure of spatial autocorrelation (Geary's C) was calculated for vegetation index values generated from high resolution imagery for seven of nine evapotranspiration monitoring sites. The two sites shown exhibit the extremes of decay rates for autocorrelation. On average, significant correlation ends at lengths of about 30 m. Click for larger image.

comparison of point characterizations of land cover with variables derived from satellite imagery is problematic. Spatial analysis of high-resolution airborne imagery collected for this project indicates that the scale lengths over which sawgrass densities are strongly related are extremely short with statistically significant relationships ending at lengths of approximately 30 meters (fig. 2). Thematic mapper (TM) resolution is therefore at the limit required to capture inter-patch dynamics and represent the average distribution of sawgrass over large areas. Intrapatch dynamics are not resolved through conventional classification of TM pixels into nominal density classes. The importance of variation at these short scale lengths remains in question, given a planned hydrodynamic model resolution of 500 m. None the less, subpixel classification of satellite data presents one possible means of characterizing sub-30 m heterogeneity. Spectral un-mixing is being investigated for the estimation of fractions of cover type within TM pixels or the generation of continuous fields of vegetation density for use in model parameterization. These results will be compared with neural network based classification techniques also being developed through this project and detailed elsewhere in this volume (Lemeshewsky, this volume).

The characterization of other surface features is important for process modeling in the Everglades. The important role that periphyton composition plays in mercury cycling has been recently discovered (Cleckner and others, 1999). Its composition and placement in the water column vary in time and space, making it both an attractive target for and a complicating factor in remote sensing activities. For example, on an event timescale, periphyton can separate from the soil substrate and form mats that float on the water surface. The result can be a drastically different canopy substrate as imaged by the remote sensing instrument. Although limited in temporal and

Figure 3. A spectral library is being assembled using field and laboratory spectra collected from various land surfaces. These spectra are used in the analysis of airborne hyperspectral imagery. Click for larger image.

spatial extent, hyperspectral imagery collected from airplane platforms may provide data of sufficient spatial and spectral resolution to develop effective techniques for mapping periphyton characteristics. A spectral library is being assembled using reflectance spectra collected in the field over various vegetation canopies, water, bedrock, and other surfaces (fig. 3). In May of 2000, reflectance spectra were collected in the field using a handheld spectroradiometer within 24 hours of the collection of airborne hyperspectral remote sensing data. Periphyton field spectra collected at this time compare well with spectra drawn from calibrated imagery. These and other hyperspectral data are being used to develop periphyton mapping methods.

Measurement of evapotranspiration (ET), a dominant component of the Everglades water balance, has been under way at nine locations for several years (German, 1996). While the importance of the role of vegetation in wetland evaporation remains a source of controversy (Allen and others, 1997), results from these sites suggest that, as in many environments, available energy and water levels are the first-order variables (German, this volume). Therefore, remotely sensed data have been used to empirically derive spatially distributed fields of ET for selected image dates. Although this technique provides a glimpse of spatially distributed ET, its application is limited to the clear-sky dates for which both imagery and sufficient ET point estimates are available. Therefore, a simple, more widely applicable technique has been developed, in which the spectral characteristics of the regions immediately surrounding each ET monitoring site are used to generate ET regions under various water level conditions. ET values from ET site measurements or from the models developed from them are then used to provide spatially distributed values of ET.

The overall objective of this research is to develop and apply innovative remote sensing and geographic information system techniques to map and characterize vegetation and related hydrologic variables, such as evapotranspiration, through space and over time. The effective use of in situ measurements and the use of remotely sensed data from multiple systems allows us to derive regional fields of information about the land surface that are not possible through any other means. Continued hydrodynamic model development and water quality research will assess the value of biophysical fields generated through these techniques.

(This abstract was taken from the Greater Everglades Ecosystem Restoration (GEER) Open File Report (PDF, 8.7 MB))

Note: PDF files require the Adobe Acrobat Reader to be read. Download the FREE Adobe Acrobat Reader ®.

Back to Project Homepage