To evaluate spectral indices as tools for separation of active aquaponds
filled with water and engaged in shrimp/fish production from empty aquaponds
using Landsat -8 data in coastal region of Guntur district, Andhra Pradesh.
Methodology: The active and empty aquaponds were demarcated with
Landsat satellite (Landsat-8) Operational Land Imager’s (OLI) multispectral
images using maximum likelihood classifier (MLC) algorithm and spectral
indices like Normalized Difference Vegetation Index (NDVI), Normalized Difference
Water Index (NDWI), Modified NDWI (MNDWI), Water Ratio Index (WRI) and
Automated Water Extraction Index (AWEInsh) by means of thresholds.
The supervised classification using maximum likelyhood classifier recorded
the highest active aquapond area whereas; NDWI, combination of indices and
WRI resulted in lower but almost similar extents. Evaluation of confusion
matrix using validation points revealed that NDWI, WRI and combination of
indices resulted in all most perfect agreement with a kappa value of more
than 0.9. Maximum likelihood classifier, NDVI and MNDWI could separate active
ponds and empty ponds from other land uses with strong agreement, while
AWEInsh could separate different land uses only with moderate agreement.
Interpretation: The study indicates that spectral
indices like NDWI, WRI and combination of indices are able to delineate
aquaponds that were cultured for shrimp/fish and kept empty at a given time
with noticeably high accuracy using satellite data for better managing of
resources in coastal ecosystem.
Aquaculture, Aquaponds, Decision trees, Spectral indices, Threshold values