Grantee Research Project Results
Final Report: Monitoring Regional-Scale Hydrologic Processes in the South Florida Ecosystem
EPA Grant Number: R825156Title: Monitoring Regional-Scale Hydrologic Processes in the South Florida Ecosystem
Investigators: Kasischke, Eric S. , Smith, Kevin B. , Richardson, Curtis J. , Bourgeau-Chavez, Laura L. , Romanowicz, Edwin
Institution: Environmental Research Institute of Michigan , Duke University
EPA Project Officer: Packard, Benjamin H
Project Period: October 1, 1996 through September 30, 1999
Project Amount: $896,086
RFA: Ecological Assessment (1996) RFA Text | Recipients Lists
Research Category: Ecological Indicators/Assessment/Restoration , Aquatic Ecosystems
Objective:
The goal of this project was to advance techniques for monitoring and predicting changes in the hydrologic condition of regional scale wetland ecosystems in the south Florida region. Our objectives were to integrate satellite remote sensing data with hydrologic models and field data to improve capabilities for monitoring and understanding processes controlling surface water flow in Florida wetlands. The specific objectives of this project were to: (1) conduct research to better understand the relationships between Synthetic Aperture Radar (SAR) backscatter and wetland ecosystem parameters (flood condition, vegetation stem density, vegetation height, biomass, etc.); (2) develop techniques for using single date and time-series SAR imagery to map spatial and temporal patterns of flooding in the herbaceous wetlands of southern Florida; (3) detect changes to the landscape using a time-series of multispectral imagery and relate these changes to natural resource issues; and (4) develop a theoretical hydrologic flow model that best describes the hydrology of the south Florida landscape and utilizes the information obtained from SAR imagery in refining predictions of hydrologic flow.To meet these objectives, we created a radar imagery database of more than 100 images and a TM database of several images. We set up sites to collect field measurements and installed data loggers to measure water level continuously over several years. Additional rain gauge and water level data were obtained from local government agencies to supplement our database.
Summary/Accomplishments (Outputs/Outcomes):
Much of our research was focused on developing techniques for using SAR data to monitor changes in hydrologic condition (Objectives 1&2). We developed methods to derive information on hydrologic condition from ERS SAR data collected over southern Florida wetlands. Our studies resulted in preliminary maps showing the regional extent of flooding and relative water depths. Additional maps of the duration of flooding (hydroperiod) were produced from a time series of ERS SAR imagery. While these products need further refinement and validation, they demonstrate the utility of SAR data for monitoring hydrologic condition in regional wetland ecsosytems. Our end goal was to have not only monitoring capabilities but also the ability to predict effects of changes in the landscape on hydrologic flow. Using Landsat TM imagery, we refined change vector analysis methods of monitoring landscape changes using a time-series of imagery (Objective 3). Finally, we developed a theoretical hydrologic model for predicting and hindcasting hydrologic flow (Objective 4).Objective 1
Theoretical Radar Modeling. Theoretical radar modeling work was performed using the MIchigan MIcrowave Canopy Scattering model (MIMICS) (Ulaby et al., 1990). The theoretical microwave scattering model was used to better understand how vegetation, soil moisture, and inundation influence the radar backscatter from herbaceous wetlands. In addition, MIMICS was used to investigate how variations in radar system parameters will influence the detection of hydrologic conditions. The model predicts a positive relationship between soil moisture and total backscatter. However, as water accumulates on the land surface, the amount of backscatter to the satellite decreases, due to specular reflection from the water surface away from the satellite. The theoretical microwave scattering model also predicts that at higher biomass levels, changes in soil moisture are not detected because the C-band wavelengths do not penetrate the canopy. These observations were validated using empirical data.
SAR Data Analysis. The results of the theoretical and empirical C-band backscatter analysis demonstrate the applicability of SAR imagery for detection and monitoring of hydropatterns in regional scale nonforested wetlands. The herbaceous wetlands of our study area exhibit a wide range of conditions, from dry to saturated soils to flooded, which are readily detected by the C-Band SAR sensor. In general, for herbaceous wetlands with dry soils, the total backscatter return is low, typically between -16 and -10 dB. As soil moisture increases in herbaceous wetlands, the backscatter to the satellite increases, roughly between -10 and -4 dB, depending on the percent soil saturation. As surface water pools, more incident energy is reflected away from the satellite, which results in a lower backscatter, ranging from -8 to -16 dB. These overlaps in dB range from the theoretical modeling and observed results indicate the possibility of having the same backscatter value from entirely different environmental conditions. For example, a backscatter value of -12 dB could be a dry bare soil or a highly vegetated wetland with variable soil moisture, or an inundated wetland. This is due to: (1) the influences of biomass, and (2) the large difference in signature from saturated soil to standing water. This study indicates that these discrepancies can be resolved by using ancillary field data on hydrologic condition and relative biomass levels to determine the environmental condition at the time of the satellite overpass.
Figure 1 shows the relationship between water depth and backscatter for several different prairie wetlands. The results from these marl prairie areas give some insight into the explanation of how the transition from saturated soil to standing water influences SAR backscatter. The theoretical models show a distinct decrease in backscatter associated with flooding. The empirical results show that while, in general, this decrease occurs, it does not happen instantaneously. What appears to be happening is that there is a mixed scattering phenomenon resulting in a gradual decrease in backscatter. The results indicate that in order for the wetlands to exhibit backscatter values predicted by the
Figure 1. Relationship Between Water Depth and Backscatter for Several Different Prairie Wetlands
theoretical models, the level of inundation must be high enough (>15 cm) to eliminate scattering off unflooded soil areas and/or double bounce scattering off the bases or stems of the vegetation. These sources of backscatter gradually are eliminated as water levels rise, and more incident microwave radiation is forward scattered off the water surface, which results in a decrease in backscatter.
Objective 2
SAR Image Analysis. The temporal database of SAR imagery was used in conjunction with field data to determine relationships between changes in soil moisture, water level, biomass, and SAR backscatter. An algorithm was developed, based on our field measurements and coincident SAR data, to estimate relative water level (hydrologic condition) from SAR backscatter. This algorithm was applied to entire ERS scenes throughout the study period to produce hydropattern maps. Figure 2 presents an example of wet and dry season hydropattern maps from 1999. These are examples of a series of maps produced for the study area. A movie of 25 hydropattern maps is available for viewing at http://esg.erim-int.com/wetlands.
Figure 2. Example of Wet and Dry Season Hydropattern Maps From 1999
The time-series database of SAR imagery developed under this project allows us to monitor inter- and intra-annual variations in hydropattern for the Big Cypress region. The maps produced allow us to determine both the spatial and temporal variations in wetness and inundation throughout the region. The maps were created using our 12 study sites located in herbaceous and sparse forest canopy. For testing the accuracy, we focused on water monitoring stations installed by local government agencies?Big Cypress National preserve (BCNP), U.S. Geological Survey (USGS), and South Florida Water Management District (SFWMD). Unfortunately, we were only able to test the accuracy of the maps using 3 sites because the water monitoring sites of BCNP, SFWMD, USGS, etc., are all located in forest, where C-band SAR cannot penetrate the canopy to measure inundation. Although, the 3 testing sites showed our map as accurate, additional validation of these maps is desirable. Based on our knowledge of the region and microwave energy, a problem seems to exist in mapping areas with dry soils as inundated.
Our next step was to use a time-series of imagery to create a hydroperiod (time period of inundation) map for the region. Assuming this map is correct, we can improve our individual date hydropattern maps by using the hydroperiod map as a base. To develop a hydroperiod map, we evaluated several methods of conducting time-series analysis. While a time-series of imagery allows assessment of temporal variation in land cover condition, it is often cumbersome to deal with the large number of images. We worked on various methods of combining time-series imagery to better interpret the factors influencing the large variation observed in ERS backscatter from the south Florida wetlands in wet versus dry seasons. The most useful technique was Principal Component Analysis (PCA).
One negative characteristic of SAR imagery is image speckle that results from the coherent image formation process. The PCA technique reduces speckle noise in the first component image, and it becomes progressively greater in the subsequent component images, with speckle dominating PCs greater than 4. Thus, these first few component images are useful for analyzing 14 months worth of data. This statistical analysis technique allowed us to determine the sources of temporal variations in backscatter between different wetland types. For instance, the first principal component showed all the areas that were stable in all of the images, while each subsequent principal component image showed areas that were highly variable in backscatter between different images.
An unsupervised classification was conducted on four of the PCA images (Figure 3). This resulted in a map that discriminates not only changes in wetland hydrology, but also between wetland vegetation type. Independent validation showed that this technique is highly accurate.
Figure 3. Classification of Four PCA Images
Objective 3
Multispectral Data Analysis. Land use and vegetation changes have implications for hydrologic regimes and, thus, hydrologic modeling and flood control as well as land management (parks, reserves, and wildlife habitat). We conducted analysis of Landsat data to map changes in southwestern Florida between 1973 and 1995. Five winter scenes of Landsat MSS and TM data were acquired from 1973 to 1995 (MSS 25-Oct.-73, MSS 04-Nov.-78, TM 25-Nov.-82, TM 18-Jan.-88, TM 22-Feb.-95) and three summer scenes from 1985 to 1995 (TM 18-Jun.-85, TM 16-Jun.-90, TM 14-Jun.-95). These data were used to detect changes in land use and land condition.
A hybrid change detection procedure was applied to the first and last date of the winter scenes (1973-1995). The hybrid procedure is based on several data preprocessing steps and two types of change detection processing (delta-classification, and a radiometric change detection procedure called Change Vector Analysis, CVA). Figure 4 presents an example of the hybrid change detection procedure used to detect human induced land cover changes in southwest Florida. The red areas are areas of change. Using this method, changes can be detected between any two dates. Although Figure 4 shows only areas of change and not type of change, the latter also was determined. We worked with local agencies in planning our multi-spectral analyses to ensure our products would be useful to local land managers. The local hydrology and hydrologic change will have an impact on habitats and will need to be used with the multi-spectral data for assessment of habitat quality and condition.
Figure 4. Example of the Hybrid Change Detection Procedure Used to Detect Human Induced Land Cover Changes in Southwest Florida
Hydrologic Modeling. Current hydrologic flow models for the Everglades region, such as the Managed Systems Model, do not permit seasonal monitoring, since the input parameters are based on annual averages. In order to better model and ultimately manage the vast Everglades wetland ecosystem, a hydrologic model is needed that incorporates spatial and temporal variations in environmental conditions (hydroperiod, biomass changes, precipitation, etc.) into the model. Our goal is to integrate hydroperiod information derived from ERS SAR along with ground monitoring data so that they can be used as parameter inputs into a hybrid hydrologic model. In this manner, hydrologic flow will be modeled throughout the Big Cypress National Preserve/Everglades National Park region. The ultimate goal is to run the updated hydrologic model for different management and environmental condition scenarios, to more accurately assess how management regimes or global climate/sea level changes would impact the functionality of the Florida Everglades wetland ecosystem. This model remains in production. Results will be forwarded to EPA upon completion.
Journal Articles on this Report : 2 Displayed | Download in RIS Format
Other project views: | All 10 publications | 2 publications in selected types | All 2 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Michalek JL, Colwell JE. Monitoring development in Southwest Florida (1973-1995) using Landsat data and a hybrid change detection technique. Natural Areas Journal April 2000. |
R825156 (Final) |
not available |
|
Smith KB, Kasischke ES, Bourgeau-Chavez LL. Empirical and theoretical synthetic aperture radar analysis of seasonal wetland dynamics in Southern Florida. Remote Sensing of Environment April 2000. |
R825156 (Final) |
not available |
Supplemental Keywords:
ecosystem, restoration, habitat, ecology, hydropattern, hydrologic regime, hydrologic modeling, wetlands, overland flow, SAR, Landsat, remote sensing, Florida, FL, watersheds, water, south., RFA, Geographic Area, Ecosystem Protection/Environmental Exposure & Risk, Ecosystem/Assessment/Indicators, Ecosystem Protection, exploratory research environmental biology, State, Ecological Effects - Environmental Exposure & Risk, Ecological Indicators, aquatic, ecological condition, monitoring, remote sensing, wetlands, Florida Everglades, aquatic biota , biodiversity, ecosystem assessment, estuaries, satellite images, Southeastern Estuaries, regional hydrologic vulnerability, conservation, ecosystem condition, ecological assessment, estuarine ecosystems, regional scale, aquatic ecosystems, water conservation, restoration, South Florida ecosystem, FloridaRelevant Websites:
http://esg.erim-int.com/wetlands
http://www.nps.gov/bicy
http://www.nps.gov/ever
Progress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.