Normalized difference phytoplankton index (NDPI) and spatio-temporal cloud filtering for multitemporal cyanobacteria pollution analysis on Erie Lake in 2014

Cyanobacteria pollution has a great damage on freshwater resources and detecting the pollution is attracting more attention from researchers. In this paper, aiming at mapping accurately and continuously the distribution of cyanobacteria pollution over a pollution period, we introduce the Normalized Difference Phytoplankton Index (short for NDPI) first for pollution area detection. The new index combines NDVI and NDWI to enhance the effective in cyanobacteria identification. A spatio-temporal filtering method is also proposed to remove cloud-cover effect in multi-temporal data. The filtering method consists two steps. The first step is to mask cloud on each single date image by using ACCA algorithm and Landsat-8 OLI cirrus band. The second step is to remove the cloud effect by applying the low rank tensor recovery method. Using the proposed methods, we analyze the water pollution incident of Erie Lake in 2014 with Landsat-8 multi-temporal imagery. The results demonstrated the NDPI and the 3-D tensor filtering methods are apparently applicable. The pollution area ratio under different pollution levels at individual observation times is derived and the trend of spatial and temporal changes of NDVI, NDWI and NDPI are also analyzed.