Task 3: Biophysical Models

Dr. Deepak Mishra in the Department of Geography at UGA is leading the biophysical and flux data analysis and satellite based modeling task. Mr. Shuvankar Ghosh conducted a portion of his PhD research with funding provided from this project.

Synopsis:

Robust multi-scale models will be developed by combining biophysical data, eddy covariance tower data, and satellite data to estimate GPP, net ecosystem exchange, biomass, and chlorophyll content to be used for forecasting and hindcasting. Many of these biophysical vegetation processes remain poorly understood at the landscape scale for salt marshes and are difficult to predict because of non-linear responses. We use NASA's Moderate Resolution Imaging Spectroradiometer (MODIS) 250 and 500-m datasets and 30-m Landsat data for developing a biophysical parameter centered Gross Primary Production (GPP) model for Gulf salt marshes. GPP has a direct relationship with the photosynthetic capacity and other biophysical parameters such as leaf area index (LAI), canopy chlorophyll content (Chl), vegetation fraction (VF), and above- and belowground biomass (BM). After successful validation, we will apply the models to satellite data covering the study site and develop weekly and bi-weekly composite biophysical products so that comprehensive site-specific phenological analysis can be performed. Fifteen years of temporally dense Landsat and MODIS data will be used to map the spatial distribution of biophysical characteristics.

Methods

Field Data Collection

The models developed for mapping tidal wetland biophysical characteristics were based on establishing statistical relationships between MODIS 250m and 500m surface reflectance products and in-situ estimates of the tidal wetland biophysical characteristics or Vegetation Indices (VIs), such as Above–ground Green Biomass (GBM), Vegetation Fraction (VF), Green Leaf Area Index (GLAI), and Canopy Chlorophyll (CHL_c). Field sites were selected in areas that demonstrated extensive homogenous patches of tidal wetlands potentially covering multiple MODIS (250m and 500m) pixels. Within each 250m or 500m MODIS pixel, multiple (~4–8) mono–specific sub–plots (1.83m × 1.83m) were selected for field data acquisition based on accessibility. The data from all the sub–plots within a specific site were aggregated to represent a MODIS pixel during model calibration and validation (Figure 2).

Model Calibration and Validation

The main goal of model calibration was to establish relationships between several well established VIs and wetland biophysical characteristics. In this study, an extensive and comprehensive in situ data collection over a period of two years enabled us to perform a comparative assessment of the existing VIs which is crucial for providing insight into both selection of the best VI for mapping wetland biophysical characteristics, and the possible explanation for their respective performances.

Following the initial pre–processing of MODIS data, in situ sampling locations were used to extract pixel values from MODIS images. Scenes were chosen based on the proximity of the dates between the image acquisition and field data collection. Following successful calibration and validation, 8–day time–series composites were generated using ERDAS Imagine for each biophysical characteristic (GLAI, VF, CHL, and GBM) using the best fit models. Map composites were generated for both 250m and 500m data. For an entire year, almost 46 composites were obtained for each biophysical characteristic. These composites can be used for qualitative assessments of both site-specific and landscape-level tidal wetland conditions before and after significant natural and anthropogenic events such as hurricanes and droughts. Phenology charts for site specific tidal wetland patches were derived from these time–series composites, using R-Studio for the growing seasons over the course of fifteen years.

Results

Time–Series Composites

The time–series map composites that we generated for the 16 year time period using the best fit models for both 250m and 500m data provide relevant qualitative assessment of the biophysical status of the tidal wetlands (Figure 3). In particular, the models were able to illustrate the effects of large scale natural disasters affecting the region. For example, a comparison between the time–series composites pre and post hurricane Katrina clearly illustrates the impact of the storm surge generated by hurricane Katrina, in the tidal wetland habitats in Grand Bay National Estuarine Research Reserve, Mississippi. The 8-day composite derived from the MODIS image of August 13, 2005 showed relatively high levels of CHL, GBM, GLAI and VF for those regions which is expected during the middle of the growing season. Post hurricane composites (September 14) showed significant reduction in the levels of all biophysical characteristics, indicating severe short-term physical impact of the high energy phenomenon on the tidal wetland habitats. Such high frequency time-series map composites can not only help to identify extent and magnitude of physical damage to wetland patches after similar natural or anthropogenic disasters, but can also facilitate restoration and conservation measures. The high temporal resolution of the MODIS products allows for frequent monitoring, leading to rapid initiation of restoration efforts after disturbances and accurate monitoring of the restored habitats.

Phenological Analysis

Phenological charts derived from the time–series composites illustrate the trends in the biophysical values (Figure 4). The MODIS based biophysical models enables us to develop long-term high frequency phenology for tidal wetland sites. The site specific phenology shown in this study for tidal wetlands of Pascagoula capture not only the natural seasonal variability, but also the effects of various natural and anthropogenic events that have occurred between 2000 and 2016 on the wetland biophysical status. The growing season of tidal wetlands along the Gulf Coast usually begins in March/April and reaches peak growth and photosynthetic activity in August/September, followed by period of senescence and dormancy from October until the beginning of the next growing season. Natural or anthropogenic disasters induce both short and long–term stress in these wetland habitats. Hurricanes and similar high energy phenomena such as tropical storms have been known to cause moderate to severe short-term physical damage to wetlands. For example, effects of the storm surge generated by Hurricane Katrina on the tidal wetlands of Pascagoula are clearly visible by sudden decline in the levels of biophysical characteristics in the month of September, much before the normal end of the growing season. In addition, the effect of the drought conditions on the growth cycle of the tidal wetlands is also evident by the absence of a prominent peak in the level of biophysical characteristics in the middle of the growing season in 2011.

Conclusion & Further Research

The MODIS based models developed in this study have been able to map these biophysical characteristics effectively, and can serve as baseline for developing satellite based models of NPP/CSP potential of the tidal wetlands. The MODIS based time-series biophysical products (2000 – 2016; still ongoing) for the tidal wetland areas thus developed during the initial phase of the research, will be used to develop MODIS GPP models by linking these biophysical characteristics to in situ flux measurements. We aim to investigate and establish statistical relationships between measured GPP from flux towers and MODIS 250m and 500m derived GLAI, VF, CHL and GBM, through proper calibration. Post calibration, we intend to choose the best-fit models based on uncertainty analysis, percent root mean square error (%RMSE), residual trends, and sensitivity analysis. Once validated, we will apply the models to MODIS images covering the entire GA coast tidal wetlands and develop 8-day composites of MODIS GPP for the entire Gulf Coast. Mapping GPP will enable us to determine whether these ecosystems are functioning/likely to function as sinks or sources of carbon in the environment, with changing climate.

Recommended Citation

Shuvankar Ghosh, Deepak R. Mishra, Anatoly A. Gitelson. 2016. "Long-term monitoring of biophysical characteristics of tidal wetlands in the northern Gulf of Mexico — A methodological approach using MODIS." Remote Sensing of Environment 173: 39-58.

https://doi.org/10.1016/j.rse.2015.11.015.