To effectively monitor macrophytes in inland water ecosystems it is crucial to retrieve information on their spatio-temporal dynamics and some key biophysical parameters, such as density and biomass. Most of the studies on this topic are still based on terrestrial vegetation and agricultural crop, and therefore there is the need to test the efficiency of remote sensing based estimation models over aquatic vegetation, which exhibit specific physical and spectral features compared to terrestrial vegetation. This works focus on the implementation of semi-empirical models based on broadband multispectral satellite data (Landsat 8 OLI and Sentinel-2 MSI) for estimating and mapping macrophyte fractional cover, LAI and biomass (dry matter weight) in freshwater ecosystems. The study areas are two: Mantua lakes system (Northern Italy), and Lake Hídvégi (Western Hungary). Both sites are shallow hypertrophic ecosystems, hosting different macrophytes groups (emergent, floating-leaved, free-floating species) and are test site for the EU FP7 INFORM project. Macrophyte canopy in situ spectra collected (ASD FieldSpec Pro) during 2014-2015 field campaigns in both study areas have been resampled to Sentinel-2 MSI bands and used for deriving 19 Spectral Indices (SIs) selected from scientific literature as being sensitive to specific plant morphological traits. Best performing Sentinel-2 MSI simulated SIs were assessed in comparison with in situ fractional cover, LAI and biomass data collected simultaneously, using linear regression coefficient of determination (R2) as indicator of goodness of fit. Finally, the best SI for each morphological trait was used to calibrate a linear regression model to be applied to Sentinel-2 data: gWAVI for fractional cover, MCARI705 for LAI and MTCI for biomass. Sentinel-2 MSI and Landsat 8 OLI multitemporal data available for 2015 and 2016 were pre-processed though spatial resampling, radiometric calibration to TOA radiance and atmospheric effect compensation using ATCOR (based on MODTRAN 4). Radiometric accuracy of surface reflectance products was assessed over macrophyte target by inter-comparing MSI and OLI spectral bands and also using in situ spectra. The radiometric inter-comparison highlighted a good matching between sensors and with the reference spectra, especially in the VIS and SWIR wavelength ranges (<5% error), while relatively higher errors (5-10%) were found for NIR range bands. Semi-empirical models implemented were applied to the Sentinel-2 scenes for producing multitemporal maps of macrophyte morphological traits of Mantua lakes system and Lake Hídvégi. Resulting maps were finally validated against in situ data collected during 2016 in both sites (n=28 for fractional cover and LAI, n=23 for biomass). Good results in terms of mean absolute errors (RMSE) were found, with scores of: 11% (0.13) for fractional cover, 0.20 m2 m-2 (0.24) for LAI, and 0.06 kg m-2 (0.08) for above-water biomass. We demonstrated that the radiometric quality of Sentinel-2 MSI data, combined with the availability of spectral bands in the red edge range, allow estimating with sufficient reliability some key morphological traits of aquatic plants. Resulting maps could be useful for managing macrophyte in freshwater and wetland systems.
Mapping Macrophyte Morpholoigcal Traits Using Sentinel-2
Paolo Villa;Monica Pinardi;Mariano Bresciani
2016
Abstract
To effectively monitor macrophytes in inland water ecosystems it is crucial to retrieve information on their spatio-temporal dynamics and some key biophysical parameters, such as density and biomass. Most of the studies on this topic are still based on terrestrial vegetation and agricultural crop, and therefore there is the need to test the efficiency of remote sensing based estimation models over aquatic vegetation, which exhibit specific physical and spectral features compared to terrestrial vegetation. This works focus on the implementation of semi-empirical models based on broadband multispectral satellite data (Landsat 8 OLI and Sentinel-2 MSI) for estimating and mapping macrophyte fractional cover, LAI and biomass (dry matter weight) in freshwater ecosystems. The study areas are two: Mantua lakes system (Northern Italy), and Lake Hídvégi (Western Hungary). Both sites are shallow hypertrophic ecosystems, hosting different macrophytes groups (emergent, floating-leaved, free-floating species) and are test site for the EU FP7 INFORM project. Macrophyte canopy in situ spectra collected (ASD FieldSpec Pro) during 2014-2015 field campaigns in both study areas have been resampled to Sentinel-2 MSI bands and used for deriving 19 Spectral Indices (SIs) selected from scientific literature as being sensitive to specific plant morphological traits. Best performing Sentinel-2 MSI simulated SIs were assessed in comparison with in situ fractional cover, LAI and biomass data collected simultaneously, using linear regression coefficient of determination (R2) as indicator of goodness of fit. Finally, the best SI for each morphological trait was used to calibrate a linear regression model to be applied to Sentinel-2 data: gWAVI for fractional cover, MCARI705 for LAI and MTCI for biomass. Sentinel-2 MSI and Landsat 8 OLI multitemporal data available for 2015 and 2016 were pre-processed though spatial resampling, radiometric calibration to TOA radiance and atmospheric effect compensation using ATCOR (based on MODTRAN 4). Radiometric accuracy of surface reflectance products was assessed over macrophyte target by inter-comparing MSI and OLI spectral bands and also using in situ spectra. The radiometric inter-comparison highlighted a good matching between sensors and with the reference spectra, especially in the VIS and SWIR wavelength ranges (<5% error), while relatively higher errors (5-10%) were found for NIR range bands. Semi-empirical models implemented were applied to the Sentinel-2 scenes for producing multitemporal maps of macrophyte morphological traits of Mantua lakes system and Lake Hídvégi. Resulting maps were finally validated against in situ data collected during 2016 in both sites (n=28 for fractional cover and LAI, n=23 for biomass). Good results in terms of mean absolute errors (RMSE) were found, with scores of: 11% (0.13) for fractional cover, 0.20 m2 m-2 (0.24) for LAI, and 0.06 kg m-2 (0.08) for above-water biomass. We demonstrated that the radiometric quality of Sentinel-2 MSI data, combined with the availability of spectral bands in the red edge range, allow estimating with sufficient reliability some key morphological traits of aquatic plants. Resulting maps could be useful for managing macrophyte in freshwater and wetland systems.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
