Detecting Algal Blooms using ENVI

Using ENVI to Detect Algae

Utilizing Image Processing Methods In ENVI To Detect Algal Blooms

When working with multispectral satellite imagery it is important to utilize a set of robust and discrete processing methods for accurate analysis and interpretation. This case study showcases the utilization of multiple image processing algorithms including gap-fill correction, sensor calibration, atmospheric correction, and classification methods to identify an algal bloom using Landsat ETM+ imagery over Lake Auburn, ME.

Data Input And Pre-Processing

First, Landsat ETM+ data from are ingested to ENVI and single file gap fill triangulation is applied to remove data gaps due to instrument failure (Figure 1).

Figure 1: Top image: Landsat ETM+ before gap-fill. Bottom image: Landsat ETM+ after gap-fill. The image is centered over Lake Auburn, ME.

Once gap-filled, sensor calibration and atmospheric correction are applied to remove unwanted affects from the instrument and atmosphere prior to data analysis. These steps can include multiple methodologies including calibration to at-sensor radiance values based on gains and offsets, or to top of atmosphere reflectance which also considers solar irradiance, sun elevation, and acquisition time.

After the data are calibrated, atmospheric correction methods such as dark object subtraction, flat field calibration, Fast Line-of-sight Atmospheric Analysis of Hypercubes (FLAASH), QUick Atmospheric Correction (QUAC), or another method is applied. ENVI supports these and many other methods for data calibration and atmospheric correction. For this case study, Quick Atmospheric Correction QUAC was applied (Figure 2).

Figure 2: Spectral profiles over a vegetation pixel before (left) and after (right) QUAC.

Data Analysis Methods For Algal Bloom Detection

Once the data are preprocessed, ENVI analytics are used to delineate any existing algal blooms within the lake. The analytics applied in this case study are based on the assumption that vegetation strongly reflects radiation in the near infrared (NIR) portion of the electromagnetic spectrum and strongly absorbs in the red (R) portion; therefore a normalized difference vegetation index (NDVI) value greater than 0 may indicate the presence of vegetation. To implement this assumption, first an NDVI image is produced for each of the 2011 and 2012 images. NDVI is calculated from pixel values in the NIR and R image bands where:

NDVI = (NIR - R) / (NIR + R)

Resulting values from an NDVI calculation should range from -1 to 1. Bright pixels indicate higher NDVI values and the presence of healthy vegetation, dark pixels indicate non-vegetation, and shades of gray can indicate varying levels of vegetation health. The NDVI results below (Figure 3) indicate that there are vegetation pixels within the lake.

Figure 3: NDVI image from 2011 (left) and 2012 (right) over Lake Auburn, ME

In order to highlight the vegetation pixels within the lake, the ENVI Raster Color Slice tool is used. Various thresholds are applied to the NDVI histogram distribution to exclude non-vegetation pixels and highlight varying degrees of vegetation health (Figure 4).

Figure 4: Raster Color Slice of NDVI image from 2011 (left) and 2012 (right) over Lake Auburn, ME. Colored pixels indicate vegetation.

Based on these resulting images, there is algal bloom that begins in the NW corner of the lake which propagates along the boundary before becoming more widespread.

Summary

Locating and quantifying potential and historic algal blooms in residential water supplies are important steps in planning and implementing mitigation when necessary. ENVI is instrumental in providing an all inclusive set of tools and analytics for the ingestion, preparation, and interpretation of remotely sensed data to answer this and other real-world problems.