Issue link: http://winesandvines.uberflip.com/i/66152
WINEMAKING Wample's data collection was astonishingly thorough and elaborate. In all, 1,360 10- cluster samples were taken (650 on Aug. 24 and another 698 on Sept. 8, demonstrat- ing, happily, that early season sampling was found to match midseason sampling, vali- dating the reliability of the more practical pre-harvest determinations. Once again, geographic information system (GIS) maps controlled the har- vester conveyer belt based on GPS coor- dinates. Wample discovered that he could alter the volume ratio between regular and reserve lots, finding that tiny differences in the chosen cut-off point for anthocya- nins changed the amount available for the high-quality lot tremendously. Wample would later learn that his sampling procedure was major overkill. Through statistical analysis of the robust- ness of the dataset, Dr. William Browning of Applied Mathematics Inc. showed that the actual sampling necessary could be cut dramatically from 10 per acre to as few as two, resulting in a cost of around $50 per acre. Although, in one client's estimate, the resulting reserve lot sells for an additional $10 per bottle. STI generates 60 soil characteristics plus elevation, slope, aspect and other topo- logical data. At first, no obvious patterns emerged between the soil maps STI gen- erated and the grape composition data. Wample realized that the soil itself did not determine vine behavior, but rather its in- teraction with weather and cultural prac- tices. What was missing was a synthesis of measureable metrics into a dashboard that growers could understand and, hopefully, take action to influence. Cabernet's five principal components Principal component analysis (PCA) has become a powerful mathematical tool for digesting complex data sets. By extract- ing patterns, it often identifies the simpler forces that underlie large numbers of mea- surements. When STI's soil information system data was tossed into a PCA along with geo-lo- cated anthocyanin and Brix data, five sig- nificant dimensions emerged. "We derived five drivers for influencing distribution of root system based on holding capacity, when water is applied and how it moves through the profile," Wample says. At this point, Wample's ability to think like a grape began to merge the data with an ar- tistic sense of vine response. We could finally ask, "Where did the vines have a chance to A GIS-enabled Korvan harvester sorted fruit direction as it crossed over mapped boundaries develop a root system?" and, "How big a water bank account have we got?" Deriving names for the principal compo- nents was Wample's real accomplishment. "The statisticians looked at the data and couldn't make any sense of it," he said. "But with 20 years of grape irrigation studies un- der my belt, I could see meso-climatic vari- ability sources within the data." STI data Sell to More Wineries The Wines & Vines Online Marketing System (OMS) is a web application powered by the leading wine and grape industry database. See for yourself how easy it is to target and extract the data you need to grow your customer base. Improve marketing effectiveness | Identify new winery clients | Increase sales (866) 453-9701 winesandvines.com/oms 58 Wines & Vines sePTeMBeR 201 1