Interannual Change Detection of Mediterranean Seagrasses Using RapidEye Image Time Series

Abstract

Recent research studies have highlighted the decrease in the coverage of Mediterranean seagrasses due to mainly anthropogenic activities. The lack of data on the distribution of these significant aquatic plants complicates the quantification of their decreasing tendency. While Mediterranean seagrasses are declining, satellite remote sensing technology is growing at an unprecedented pace, resulting in a wealth of spaceborne image time series. Here, we exploit recent advances in high spatial resolution sensors and machine learning to study Mediterranean seagrasses. We process a multispectral RapidEye time series between 2011 and 2016 to detect interannual seagrass dynamics in 888 submerged hectares of the Thermaikos Gulf, NW Aegean Sea, Greece (eastern Mediterranean Sea). We assess the extent change of two Mediterranean seagrass species, the dominant Posidonia oceanica and Cymodocea nodosa, following atmospheric and analytical water column correction, as well as machine learning classification, using Random Forests, of the RapidEye time series. Prior corrections are necessary to untangle the initially weak signal of the submerged seagrass habitats from satellite imagery. The central results of this study show that P. oceanica seagrass area has declined by 4.1%, with a trend of -11.2 ha/yr, while C. nodosa seagrass area has increased by 17.7% with a trend of +18 ha/yr throughout the 5-year study period. Trends of change in spatial distribution of seagrasses in the Thermaikos Gulf site are in line with reported trends in the Mediterranean. Our presented methodology could be a time- and cost-effective method toward the quantitative ecological assessment of seagrass dynamics elsewhere in the future. From small meadows to whole coastlines, knowledge of aquatic plant dynamics could resolve decline or growth trends and accurately highlight key units for future restoration, management, and conservation.

Keywords: Cymodocea nodosa; Mediterranean seagrasses; Posidonia oceanica; Random Forests; RapidEye; Thermaikos Gulf; change detection; time series.

Figure 1

Location of survey site within (A) Thermaikos Gulf, (B) Aegean Sea, Greece. The displayed RapidEye imagery is a non-atmospherically corrected, true color (band 1 as blue, band 2 as green, band 3 as red) composite in UTM (zone 34) system/WGS84 projection. The imagery was acquired on 22/06/2016 (RE16 in text). The red polygon in (B) depicts the location of (A) within the Thermaikos Gulf. The deep water polygon represents a ~160 × 160 pixel window implemented in the water column correction of the image time series as it represents an area with very little water leaving radiance values in all three bands.

Figure 2

Schematic representation of the methodology. ¹L3A ortho products are the initial radiometric, sensor, and geometrically corrected RapidEye images in UTM/WGS1984 projection, ²R represent atmospherically-corrected (FLAASH module), at-water surface reflectances, ³R_rs are remote sensing reflectances, transformed from R using Equation (1), ⁴R_b are water-column-corrected, bottom reflectances using the analytical model of Maritorena et al. (1994).

Figure 3

Polynomial regression between the log-transformed ratio of blue and green remote sensing reflectances, R_rs, and in situ depth measurements from the Thermaikos survey site. The shown polynomial equation was implemented to estimate the bathymetry map displayed in (C) of Figure 4.

Figure 4

Methodological steps from atmospheric to water column correction in order of successive processing. All four panels are true color RapidEye image composites (22/06/2016; RE16 in text) projected in UTM (zone 34) system/WGS84. (A) Non-atmospherically corrected composite. (B) Atmospherically-corrected composite using the FLAASH module. (C) Satellite-derived Bathymetry map of the survey site draped over the atmospherically-corrected composite of (B) using the site-specific polynomial algorithm of Equation (2) as shown on Figure 3. We applied a 5 × 5 low-pass filter on the initial ratio-derived bathymetry (not shown here) to reduce potential noise which would be transferred to the water-column corrected product. (D) Water-column corrected composite following application of the water column correction algorithm of Maritorena et al. (1994) draped over the atmospherically-corrected composite of (B) and masked using the optically deep limit of 16.5 m to enhance bottom features and potentially increase classification accuracies.

Figure 5

Plot of Satellite-derived Bathymetry (SDB) vs. in situ measured depth for the validation of the bathymetry map of the Thermaikos Gulf (Figure 4C). SDB was derived from Equation (2). Regressed SDB have been previously smoothed with a 5 × 5 low pass filter to reduce unwanted noise.

Figure 6

Classified water-column-corrected RapidEye composites from the 4 years using Random Forest machine learning classifier (100 trees). The frames on the upper right of each panel indicate the date of each RapidEye image. (A) RE11—Overall accuracy: 73.5%. (B) RE12—Overall accuracy: 81%. (C) RE15—Overall accuracy: 78.5%. (D) RE16—Overall accuracy: 82%.

Figure 7

Interranual change detection of seagrasses in the Thermaikos Survey site between 2011 and 2016 using RapidEye satellite images. The trajectory plot displays change of area (in hectares; y-axis) over the years (x-axis) of Posidonia oceanica and Cymodocea nodosa species, and of total seagrass area. Linear regression black lines (m = slope) show approximate trend in area between 2011 and 2016. Posidonia oceanica seagrass is decreasing at 11.2 ha/yr, Cymodocea nodosa seagrass is increasing at 18 ha/yr, while total seagrass area is expanding at 6.8 ha/yr.

Figure 8

Change in seagrass distribution in the Thermaikos survey site between 2011 and 2016 for (A) Posidonia oceanica and (B) Cymodocea nodosa. Between 2011 and 2016, P. oceanica seagrass meadows have declined by 4.1%, while C. nodosa seagrasses have increased by 17.7%.