Using 30 years of climate data to determine where to grow grapes

This time around I decided, after much fruitless searching for a climate related topic, to look at where are ideal locations in the United States to grow grapes (without excess irrigation and manipulation). It is well known that California has been in a long standing drought, and yet is one of the largest producers of both wine and various vegetable crops. I settled on grapes, however because I just find viticulture more interesting.

Figure 1 - Fuzzy Overlay map showing ideal area to grow grapes in blue based on climate data using average temperature, solar radiation, and precipitation. This fuzzy overlay used a fuzzy And processing type.

Essentially this project boils down to a nationwide site survey. To do this I pulled in climate data compiled from 30 years of data from WorldClim. I chose three variables: average temperature (°C), solar radiation (kJ m^-2 day^-1), and annual precipitation (mm). Each variable had a raster for each month of the year, so I built a small model to process all of these rasters using an iterator, then ran extract by mask to clip them to the shape of the US, then re-projected and resamples them into USA Albers Equal Area Conic and to 10,000m respectively. The growing season for grapes is between March and September, so I created rasters of the average of solar radiation and temp for the growing season, as well as a cumulative raster of precipitation.

Figure 2 - Top: map of solar radiation through the grow season (kJ m^-2 day^-1); Bottom left: average precipitation during the grow season (inches of rain); Bottom right: average temperature during the grow season (°C).

To decide where would be best given these three variables I used fuzzy logic. Fuzzy logic differs from binary logic in that instead of something being true or false (1 or 0), membership within the set (or something being true) is measured on a scale from 0 to 1 with a range of values of how likely something is to be a member (more on fuzzy logic here, here, and from ESRI.). I converted the variables raster into fuzzy members and then ran a fuzzy overlay (fuzzy And) using all three. What it produced were these maps showing ideal places where grapes should grow well.

Figure 3 - This fuzzy overlay used a fuzz Or method, which shows a larger portion of the country as ideal. 

Figure 4 - The predictive power of our model shows that California is not exactly a great place to grow grapes, at least not naturally, yet California is largest producer of wine in the US by a massive margin. It can be seen that the areas where grapes are grown in California have fairly low values.  

Figure 5 - Yearly precipitation in four wine regions: Napa, Ca; Sonoma Co., Ca, Lodi, Ca; and Barboursville, Va for comparison. Note that All three areas in California have a similar trend of very little rain during the grow season, whereas Virginia has a higher average, and good amount of rain during the grow season.