Διερεύνηση και αξιολόγηση Τηλεπισκοπικών Δεδομένων Υψηλής Χωρικής Ανάλυσης στην Αμπελουργία ακριβείας

Abstract

Η παρούσα εργασία αφορά στη διερεύνηση και αξιολόγηση τηλεπισκοπικών δεδομένων υψηλής χωρικής ανάλυσης στην απελουργία ακριβείας.

The recent advances in optical remote sensing technology are facilitating several precision agriculture applications regarding crop monitoring and analysis. Particularly in viticulture, the main cost effective scenario is to produce vegetation related maps, during the veraison, as an additional input for the management of the upcoming harvest. This has been regularly employed both from a research and a commercial point of view.

To this end, the main objective of our study was to evaluate high spatial and spectral resolution satellite data for selective harvesting. From concurrent satellite and field campaigns multi-modal data was acquired, i.e. i) ground radiance data using a field spectrometer (GER 1500, Spectra Vista Corporation, 350-1050nm, 512 spectral bands), ii) WorldView-2 satellite data (DigitalGlobe, Inc. USA), GPS data (Trimble Spectra Epoch 25 L1/L2 RTK), orthoimages and DEM’s from KTIMATOLOGIO A.E. and other quantitative and qualitative data acquired from the collaborating wineries in the two study areas (Aigio & Atalanti, Greece). The satellite imagery had a spatial resolution of about 0.5m in the panchromatic band, and about 2m in the 8 multispectral bands which covered the range between 400nm to 1040nm. Data pre-processing included radiometric correction, atmospheric data correction (Modtran, Atcor), image fusion/pan-sharpening and orthorectification. In addition to that, ground reflectance data was calculated from the atmospherically corrected ground radiance. Also, simulated ground radiance and reflectance data, corresponding to the eight WV-2 bands, was computed and employed during the evaluation. Thus, it was made possible to evaluate the relationship between ground data ans satellite data, through correlation and linear regression models.

Afterwards, spectral signatures of the several (more than 20) vine varieties were calculated for all datasets (i.e. radiometrically corrected, atmospherically corrected, fused, simulated ground, etc.). Using linear regression models, the spectral signatures computed from satellite data (WV-2) were correlated with the ones computed from the simulated ground data (GER1500). In both study areas, the correlations of radiometrically corrected and fused data sets with the simulated radiance, gave similar results( R2 values of 81-85% and p-val of 0,1-0,2% in study areaA and R2 values of 61-81% and p-val of 0,2-2% in study area B). The results of theatmospherically corrected data to reflectance data correlation were significantly better (R2 valuesof 97-99% and p-val of 1,5*10-6-0,0025% in study area A, R2 values of 90-99% and p-val of 4*10-5-0,3% in study area Β). Furthermore, a number of vegetation indices (twenty-nine as proposed in the corresponding literature) were computed based on the fused satellite data. In cases where the indices had been proposed for use with hyperspectral data, their formulations were approximated to correspond to WV2 data. The calculated indices belonged to the following five general categories: Vegetation (NDVI, OSAVI, MCARI2, MTVI2, etc.), Chlorophyll (Gitelson Chl1-2, etc. ), Carotenoids (Blackburn Car1-2, Gitelson Car1-2), Carotenoid to Chlorophyll Ratio (NPCI, SIPI, etc.), Anthocyanins (Gamon Anth, etc.). Additionally, the green LAI (Leaf Area Index) was computed through a linear relation with the NDVI. Also, the maturity index IMAD and the color intensity index CIRG were calculated through a linear relation with the Gitelson Car2 index.

The above were compared with the actual decision making of the wineries both during harvesting and vinification. The qualitative and quantitative evaluation demonstrated that the final decisions made by winegrowers and winemakers were very close to the produced satellite map estimations. In particular, for the Syrah variety, in study area A, the proposed zoning and selective harvesting matched the final decisions regarding the three different qualities/products. Moreover, the estimated maturity condition (IMAD) was highly correlated with the organoleptic characteristics and the overall harvest management in both test sites/wineries.