A method based on satellite imagery to identify spatial units for eutrophication management

Abstract

Abstract A procedure based on analysis of time series of MODIS-Aqua satellite chlorophyll a images was developed in order to spatially delimit areas of the Alboran Sea with distinguishable annual cycles of productivity. For this objective, daily images of satellite chlorophyll were obtained from 2002 to 2013. Monthly means were calculated for each pixel and then the pixels were grouped by means of the statistical technique of non-hierarchical clustering analysis according to the features of their annual cycle. The outcome of the clustering analysis was validated by comparing the pixel grouping patterns with the spatial distribution of in situ chlorophyll a and nutrients as determined from the available data (about 2900 registers obtained in the 20 m upper seawater layer in 70 sampling stations). Six grouping areas were identified which reproduced the expected zonation according to the distribution of the main mesoscale hydrological structures that modulate the productivity in the study area (oligotrophic gyres and coastal upwelling). The grouping areas presented contrasting chlorophyll a concentrations as well as annual minimum and maximum occurring at different seasonal cycle timing. Seasonal means of in situ chlorophyll a and nitrate and phosphate concentrations calculated for the six grouping areas were significantly correlated, supporting that dissolved inorganic nutrients were the main factor controlling the productivity in the Alboran Sea. Consequently, the grouping of pixels based on satellite data reflected reasonably the underlying mechanisms that control the phytoplankton biomass in the study area. The utility of this technique to define spatial units for eutrophication management is discussed. These units can be used for spatial aggregation of eutrophication indicator data collected from in situ samplings as well as for calculating robust reference values and time trends. Furthermore, the pixel grouping is useful for optimizing the existing monitoring programs as it facilitates the selection and location of sampling stations in order to avoid collection of redundant and/or pointless information.