Farmers mentioned the need for more crop research, information-sharing, and stronger norms around acceptable environmental practices. While this theme did not translate easily into quantifiable spatial proxies, we focused on farmers’ expressed desire to grow in remote areas because of the opportunity to work the land in proximity to wild flora and fauna. We quantified this ruralness using the Human Footprint layer, which combines data on the built environment, population density, night-time lights, crop and pasture lands, roads and railways, and navigable waterways to create an index of direct and indirect human pressures at a 1 km2 resolution. We extracted the mean human impact value for each parcel using the exactextractr package in R . In the quote above, the farmer explained how some aspects of regulation were more impactful to his daily farm management decisions than others as he navigated the licensed industry. Most farmers did not perceive that enforcement influenced their land use decisions, although the farmers navigating the licensed recreational market said that regulations were often their first consideration. One unlicensed farmer compared law enforcement to wildfire risk, explaining both as factors that were constant background risks but ultimately outside of his control.Multiple farmers said that they started growing hemp, or had considered growing hemp, to avoid the legal hurdles of recreational cannabis. Others raised questions about what the new recreational market would mean for medical producers. Some interviewees mentioned that a rural location made things easier from an enforcement perspective, particularly in avoiding the Grants Pass area . Even those who were attempting to navigate the legal industry expressed that it was useful to be less closely monitored because of the difficulty in complying with all regulations, the time needed to demonstrate compliance, or fear that they may be breaking rules without knowing it. To translate the preference for distance from law enforcement into a spatial driver,vertical grow shelf we estimated this both with ruralness as well as the straight line distance from the Grants Pass Sheriff’s office to each parcel using the sf package in R . However, because these measurements were significantly correlated, we ultimately dropped distance to law enforcement as a variable in our models.
There were also a number of regulatory designations that cannabis farmers discussed as important when considering where to grow. Water rights and zoning were some of the most frequently mentioned. Water rights were considered critical for legal production but specifics of parcel-level rights were often hard to acquire or interpret. Water rights were not generally discussed by unlicensed farmers, but water access, storage, and application were all considered critical. Because of the mixed response to regulated water use, we assessed water access as part of Parcel Qualities below, rather than in Regulation. The shifting policies in Josephine County around zoning restrictions, particularly for Rural Residential zones, led farmers to identify exclusive farm zoned parcels as the safest and highest quality lands for cannabis production. One farmer also mentioned Farm Resource zoned properties. To translate this into a land use driver, we created a binary variable that assigned a ‘1’ to each parcel that was zoned for either EF or FR zones and a 0 for those that did not. Zoning information was provided by Josephine County . Farmers identified multiple biophysical properties of parcels that factored into decisions about where to produce cannabis. In the quote above, the farmer was expressing confusion as to why some cannabis producers selected parcels that required a large labor input to clear or terrace land to begin farming, when other, more open parcels seemed to him to be a more ideal choice. In addition to open/cleared areas with access to sunlight, some of the other factors mentioned included relatively flat slopes, and medium elevation zones as helpful qualities for production. Several interviewees mentioned that the climate in Josephine County was ideal for cannabis, while others expressed the belief that it was primarily grown in the region because of history and culture. One farmer mentioned that owning versus renting land for cannabis farming might change the relative importance of the physical factors of a parcel that a farmer prioritizes, as might living on the property where they are growing, but they weren’t sure how often producers rented versus owned their farms.We translated the above biophysical parcel qualities into multiple spatial drivers. First, we grouped land cover classes into a binary variable based on ease of clearing for crops. We included the following classifications in the easy to clear category, based on land cover descriptions: Developed Low Intensity, Grassland/Herbaceous, Developed Open Space, Pasture/Hay, Barren Land, and Cultivated Crops.
In addition to clearing, we created a binary variable to describe if the majority aspect of a parcel was southern-facing, to reflect parcels with greater sunlight access, using the raster package in R. We also used maximum elevation per parcel to capture elevation as a potential limiting factor, using a 10 m DEM and the exactextractr package in R . We calculated maximum roughness to capture potential preference for overall flat parcels using the ‘terrain’ function in the raster package in R . In the raster package, roughness measures the difference between the maximum and minimum elevation value of a cell and its surrounding cells. Farmers discussed parcel size as a potential factor that could influence where to locate a cannabis farm. One farmer mentioned that parcels in Josephine County were smaller than in other regions where he had farmed cannabis, while other farmers implied that they had looked for larger parcels within the county. Multiple farmers discussed the importance of space on the property, whether directly for cannabis production , multiple kinds of cannabis production , or for other reasons, for example to provide a treed buffer or space for a fence between the farm and its neighbors, to have enough room for setback distances required by regulation, or to accommodate other land uses on the same parcel . To translate this into a spatial driver, we used the calculated area of each parcel polygon using the sf package in R. Not all farmers interviewed operated licensed production sites, and many were in a “gray zone” of legality, and so for some, proximity to water on a parcel was more important than specific water rights. Most farmers mentioned that in 2016, regulations on cannabis farming were not yet enforced, and so access to water at that time point might have had more to do with physical parcel qualities than legal access. Because of this, we used proximity of farmed parcels to water as a spatial driver instead of specific water rights on a given parcel for our model. We used the NHDplus flowlines database, filtering to include rivers and streams, as well as artificial paths . We then calculated distances using the sf package in R . While some farmers mentioned that soil quality mattered to them when selecting a site, most said that existing soil was not a primary concern for them, or for most farmers that they knew. Instead, most reported that the industry standard was to grow with imported soils in grow bags or boxes. Some farmers did report growing in native soil, but that they still had to add amendments to do so. Given the mixed comments on soil quality, we did not include this as a potential spatial driver of cannabis grow indoor land use.While all farmers interviewed discussed the difficulties of supporting themselves or their families economically in the cannabis industry, none of them specifically mentioned land prices as a factor in their decision making, and we did not ultimately include any drivers based on this theme. In the quote above, the farmer expressed that it was difficult to make a secure living with cannabis farming, which often made it risky to attempt new sustainable techniques. In this case, the farmer was also explaining that in their own attempts to grow with lowered environmental impacts in mind, it sometimes meant an income trade off. Thus, farmers reported that economics primarily influenced their decisions on specific land use practices, as well as whether or not to enter the licensed market. The farmers did see broader drivers of supply and demand being important for the industry as a whole, but for their individual decisions, economics was influential in deciding how much to grow, how much to spend on equipment or labor, how to balance different types of production , or when they might have to leave the industry altogether. Most expressed that the industry, both licensed and unlicensed, was full of uncertainty, and economic vulnerability. Many expressed concerns that when operating under economic uncertainty, farmers were unlikely to take a risk on more sustainable or less ecologically-impactful farming practices.
All interviewed farmers said that the cannabis farming industry had expanded with legalization, and expressed concerns or uncertainty for the future of the industry. In the quote above, the farmer was looking at their own long history in the cannabis industry and seeing an uncertain future, and comparing it to the other major land-based industry cycles in Josephine County. Most interviewed farmers compared the cannabis industry to the gold rush, and expressed concern that its rapid increase might not be sustained in the long term. Many farmers, both legacy producers that associated themselves with hippie culture or renegade counter-culturalists, as well as younger farmers that came from more indoor or urban production cultures, described a shift in the industry from one that was culturally or spiritually motivated to one that is primarily economically driven. They expressed concerns that the industrialization of cannabis with the legal market would lead to further ecological harm, while the money involved in the black market would encourage other criminal activities . Many farmers expressed a desire for more research and education, particularly around best growing practices. Most of those interviewed agreed that there was a general lack of knowledge or research-supported farming practices. While few were optimistic about the future, most expressed a belief in small-scale farms to produce in a way that was less harmful to the environment than conventional agriculture, and for persistence of a “craft cannabis” market.For the model of cannabis development onto new parcels post-legalization in 2016 , we found that the following hypothesized drivers had a significant relationship with parcels that developed new cannabis: larger parcels, lower human footprint, lower distance to nearest cannabis, higher density of local cannabis, easily cleared land cover, non-farm zoned, lower elevation, less rough, lower distance to rivers, and mapped in 2013 . All significant drivers performed in the direction we predicted , except for farm zoning, which was negatively associated with the development of new farms, and image year, which did not have an associated prediction. Distance to nearest cannabis, local cannabis density, parcel elevation, and distance to rivers or streams all had approximately linear relationships with the probability of new cannabis development . Parcel area and roughness on the other hand had non-linear relationships with possible threshold effects . The change in probability attributable to individual covariates was generally small , except for parcel area and human footprint . Rural cannabis land use in the western US has traditionally been a difficult topic for research. In this study, we demonstrated the effectiveness of an interdisciplinary approach to identify, assess, and contextualize drivers of cannabis land use and development. We combined generative cannabis farmer interviews with three models of cannabis land use in Southern Oregon during the early period of recreational legalization , to examine the relationship of spatial covariates with cannabis distribution, new development post-legalization, and plant density over time. The majority of our covariates were significant in at least one model, and combined with the context from the farmer interviews, suggest that they are likely reliable predictors of land use in this system. Previous studies examining cannabis land use and land use change have relied on biophysical covariates. Building on this foundational approach for understanding cannabis distributions, the addition of interview data to inform and contextualize models adds depth to the interpretation of modeling results, and generates new covariates that might otherwise be missed. For example, in Butsic et al. , the authors noted strong network effects on the distribution of cannabis production, and postulated that producer networks might be important in the development of the industry.