|Because of the lack of spatially detailed data on lake moderation of climate and macroclimate, the model was constructed using primarily topographic and soil factors.
3) Evaluate the accuracy of the GIS-derived variables at selected sites through field survey techniques and laboratory soil analysis.
4) Use statistical techniques (correlation, principal components analysis) to identify patterns of physical characteristics in existing vineyard locations. Predict other potential vineyard sites the using identified characteristics. The statistical method used most extensively was spatial correlation. Because of the small number of sites, verified vineyard areas were compared to the rest of the county on a 30-m pixel spatial basis. Highly correlated variables were excluded from the sets of variables, and variables with a high power of differentiation were retained in the model.
5) Review identified characteristics and potential locations with extension experts and industry representatives. Refine model as needed. This process continues. Please contact Tracy Aichele (firstname.lastname@example.org) with concerns, suggestions, or questions.
Results and discussion
Based on the data analysis described above, four differentiating, relatively uncorrelated, and surprisingly simple characteristics (and the resulting rules for identification purposes) emerged:
Spatial aspect – Vineyards were preferentially located facing directions other than northward. (Aspect not between 315 and 45 degrees.)
Potential for cold air drainage or local lake moderation of cold temperatures – Most vineyards were located in areas of relatively high relief, and at higher elevations in those areas. Some vineyards were located very close to a large lake; the presence of a large body of water moderates cold winter temperatures. (Range of elevation in a circular area centered on this point with a radius of 1,500m is greater than or equal to 30 m and greater than the mean elevation in that area.)
Sandy soils – Vineyards were preferentially located in areas with high sand content. Field samples obtained from vineyards also tended to be sandier than the representative mapped soil. (Mapped sand proportion of A horizon of predominant soil series in the mapping unit is greater than or equal to 30%.)
Low organic matter content – Vineyards were preferentially located in areas with low organic matter content as mapped. The soils in vineyards examined, however, tended to have higher organic matter than the representative mapped soil. (Mapped organic matter content of A horizon of predominant soil series in the mapping unit is less than or equal to 2%.)
In Berrien County, sites with all of these characteristics were 4.3 times more likely to be vineyard areas than non-vineyard areas. Sites with two or fewer characteristics were 13.8 times less likely to be vineyards than to be non-vineyards.
In Leelanau County, sites with all of these characteristics were 2.5 times more likely to be vineyard areas than non-vineyard areas. Sites with two or fewer characteristics were 30.1 times less likely to be vineyards than to be non-vineyards.
In both counties, the most common characteristic to be missing from a site was the non-north facing aspect (i.e., some portions of vineyards faced north). This was more prevalent in Berrien County than in Leelanau County, perhaps because of the differences in terrain and surface formations or because of the difference in macroclimate. In addition, because the data were collected on a 30-m cell basis, vineyards may have had small areas facing north, while the predominant aspect was in another direction.
After visiting vineyard sites, we found that publicly available data of soils and topography data generally represented the sites accurately. However, the resolution of the data did not allow some finer details that were apparent in a site visit to be fully captured by the GIS. We found some differences in soil layering from the mapped data, but none that would suggest a different management regime. These differences were described as inclusions in the major soil mapping unit. The topography of the sites was more complex than could be represented with the United States Geological Survey’s 30-m National Elevation Database; therefore, the NED should not be exclusively used for planning planting patterns. The larger topographic patterns (principal aspect, range of slope, local relief) on the sites are accurately depicted, however.
Based on these results, we mapped the common soil and topographic variables across the counties and classified areas based on the number of conditions met.
(Click on the map image below to view a larger, pdf file showing roads.)
|Leelanau County map
||Berrien County map
The distinction between vineyard areas and non-vineyard areas is not clean. Potentially, many of the non-vineyard areas could be suitable vineyards. From this study, we would anticipate that areas meeting more conditions are more similar to existing vineyards, and potentially have a probability of success for vineyard plantings. Because of this complication and the small number of sites, field verification and insight is welcome in refining the model, and potential investors in real estate should examine each parcel carefully before purchase.
This work was funded by the Michigan Grape and Wine industry Council and Project GREEEN, with additional support from the Michigan Climatological Resources Program.
This work was accomplished from October 1, 2004 to September 30, 2005.