Issue link: http://winesandvines.uberflip.com/i/998810
64 WINES&VINES July 2018 GRAPEGROWING WINE EAST vineyard block. Data were stored onboard and retrieved after the flight mission. Geometric correc- tion was performed to correct the image geometry. Georeferencing was achieved by identifying con- trol points (targets) on the ground. Geometric distortions caused by changes in UAV attitude and altitude were corrected using the information provided by the inertial station. The series of images acquired during each flight was assembled into mosaics by selecting the over- lapping areas to limit the viewing angle and the problems of direc- tional effects. Once assembled and corrected, NDVI was calculated on mosaics. The NDVI pixel values corresponding to the field points were extracted and compiled into a geodatabase that included all field-based variables (e.g., leaf ψ ). An example of UAV correla- tions with other variables is shown in Figure 1 (page 60). It should be noted that mathematical relation- ships between NDVI and other variables of importance never will be exactly the same for every vine- yard. There is substantial variabil- ity between vineyards, varieties and years. That is the nature of making biological measurements and attempting to correlate those to a physical measurement such as leaf reflectance. The UAV NDVI and thermal in- dices were correlated with vine size in all Riesling vineyards in 2015, including that of the Buis vineyard in Niagara-on-the-Lake depicted in Figure 1a. The applica- tion of PCA provides a pictorial method of portraying correlations between multiple variables in a large data set. The variables repre- sented by lines (eigenvectors) are considered correlated if the eigen- vectors are parallel to each other, and inversely correlated if they are roughly 180 ° from each other. In Figure 1a, NDVI acquired by UAV was correlated with GreenS- eeker NDVI, soil moisture, leaf ψ , gs, and berry weight, and in- versely correlated to thermal data, FVT and PVT in Riesling. This sug- gests that zones of low NDVI might be linked to fruit with higher terpenes, and by inference higher-quality wine. Noteworthy associations in other Riesling vineyards included UAV indices and berry weight (five s i t e s ) , TA ( t w o s i t e s — B u i s , George), FVT/PVT (three sites), and GreenSeeker NDVI (four sites). Inverse correlations of note with UAV data included soil mois- ture and leaf ψ (five sites). Direct correlations also were noted for NDVI and at least one LT 50 mea- surement for four sites: LT 50 mea- sured in March (Buis), January LT 50 (Pondview), February LT 50 (Hughes, Cave Spring). This sug- gests that high NDVI was associ- ated with low winter hardiness, since LT 50 is a negative value. Yield was not consistently related to NDVI; yield and NDVI were in- versely correlated in three sites (Buis, Hughes, Pondview) but un- related in two others. For the most part we found similar patterns in all Riesling vineyards in 2016. In Cabernet Franc, UAV NDVI was associated with soil moisture, leaf ψ , gs, yield, and berry weight, and inversely correlated to color, anthocyanins and phenols in the George vineyard in Vineland in 2015 (Figure 1b). UAV NDVI was correlated with vine size in all other Cabernet Franc vineyards. Other noteworthy associations included UAV indices and soil moisture (three sites), leaf ψ (two sites), berry weight (five sites), TA (three sites), yield/cluster number (four sites), and GreenSeeker NDVI (four sites). NDVI was in- EASTERN WINE LABS Serving the Analytical needs of East Coast Wineries WWW.EASTERNWINELABS.COM Ph 609-859-4302 Cell 609-668-2854 chemist@easternwinelabs.com AOAC Member EasternWineLab_Mar09.qxp 1/22/09 9:47 AM Page 1 Figure 5: These maps were derived from UAV data and are examples of NDVI- based zones in three Ontario Cabernet Franc vineyards. 5a is Pondview in Virgil, Ont.; 5b is George in Vineland, Ont.; and 5c is Cave Spring in Beamsville, Ont. Scales are 5a: 1 = 1039; 5b: 1 = 1240; and 5c: 1 = 1581. a b c