Japanese Barberry Invasion in SE Minnesota

Japanese Barberry is native to Japan and was introduced in the US in the 1800s as an ornamental and landscape plant used as a living fence and for erosion control. It was promoted as a substitute for common barberry that was introduced by early European settlers for hedgerows, dyes, and jams but was discovered to be a host for black stem rust (Puccinia graminis) of wheat (which causes severe crop loss), spurring a national eradication project of common barberry from 1918-1990 (Zouhar 2008, MN DNR). Grown for its interesting foliage colors, deer resistance and adaptability to urban landscaping, Japanese barberry has been found naturalizing in understory wooded areas in Minnesota (MN DNR).

Japanese barberry is a deciduous shrub ranging from about 1 to 10 feet tall, with multiple stems growing from the root crown that have numerous sharp spines. Cultivars have been developed in a range of colors from chartreuse, gold, maroon, and green, though most naturalized plants found have green leaves in summer that become vibrant orange and red in the fall. Small yellow flowers bloom in May, followed by red oblong berries that persist into winter. (MDA). Propagation is possible by seed, sprouts from rhizomes, and rooting of stems that touch the ground; this combination of propagation can result in multitudes of stems covering a large area, making it difficult to determine individual plants. It is adapted to many soil types and can grow in full sun to almost any shade level (Zouhar 2008).

Japanese barberry has become invasive where it has naturalized in wooded areas in Minnesota, mainly via seed dispersal by birds, prompting specific cultivars to be listed as Restricted Noxious Weeds by the state since 2015. Once established, Japanese barberry forms impenetrable, thorny thickets that shade out native plants and alter the soil properties (increased nitrate concentrations) and soil microbial communities (Zouhar 2008). Over time, the change in  soil pH and the higher nutrient levels can contribute to whole ecosystem changes of the area (MDA). Dense stands of Japanese barberry in the northeastern US have been found to harbor twice the number of ticks as non-invaded areas, potentially increasing in the occurrence of tick-borne diseases such as Lyme disease (MDA, Zouhar 2008).

Approximately 50,000 foreign species cost the US economy about $120 billion a year, and about 42% of Threatened or Endangered species are at risk primarily because of alien-invasive species, according to one study (Pimentel et al 2005). Because of the large impact to the economy, environment, and risk to health, and as new data becomes available, remote sensing is increasingly being used to monitor current distributions of invasives and predict where they are likely to spread. Surveys used in combination with remote sensing data can be used to map distribution and help with early detection and focus control efforts. 

The goal of this analysis is to use remotely sensed data and random forest modeling in R to identify the importance of landscape variables to the spread of Japanese Barberry and to predict areas at risk of future spread in the southeast corner of Minnesota.

Invasive species data is available from the Minnesota DNR, which collects GPS locations of invasive species as a multi-partner effort and publishes them via the  MnGeo  website. Winona, Houston and eastern Fillmore Counties have over 400 point location records of Japanese barberry. (According to the metadata, the accuracy is +/- 30 meters.) There are no points recorded in western Fillmore County, so I chose to only use the eastern portion of Fillmore County in the study area. I used the NAD 1983 UTM Zone 15N coordinate system for this analysis.

The random forest R model will compute the statistical significance of dependent variables on an independent variable. Each of the variables I chose are available as raster data. Each raster was resampled to the same spatial resolution and extent.

This analysis uses the random forest model in R that calculates the classification and regression based on a forest of trees using random inputs, evaluating the importance of independent variables on a dependent dataset and creates a predictive raster.

The random forest model calculates 57% variance explained by the included factors, with the importance of each of the landscape factors as: canopy cover 59%, distance to water 48%, landcover 39%, elevation 26 %, distance to development 23%, northness 22%, eastness 21% and soil 21%.

It is notable that the highest factor has an importance of only 59%. Research on naturalized Japanese barberry in the northeast U.S. found it is highly adaptable to almost all soils, landcover, and levels of shade. The low importance of soil and aspect at 21%-22% is not surprising and agrees with the research. The importance of elevation is 26% in the study area, which looking at the locations, I believe suggests that Japanese barberry has a slight preference to naturalize on the hills around river/stream valleys but not exclusively.

In the prediction raster, the hill tops close to the Mississippi, Root, and Whitewater Rivers (that are forested and have dense tree canopy) are at the highest risk for future infestation.

Checking the table data, 87% of the Japanese barberry locations are in Deciduous Forest and 96% are located in 70%-94% canopy cover. Before running the model, I hypothesized that landcover would be a high factor and canopy cover would not be as significant. Fifty-nine percent importance of canopy cover, which is higher than the importance of landcover (39%) suggests two possible reasons: the data was collected by employees of organizations specifically looking for invasive plants in forested areas so there are fewer data points in more open areas, or Japanese barberry cannot compete as well as with other species in areas with less canopy cover as in areas with 70% or more where it may grow faster and outcompete smaller and/ or slower growing shade-tolerant species. The 39% importance of landcover suggests to me that, as with elevation, Japanese barberry has a slight preference for forested areas but can grow in more open areas as well, which also matches the research. 

I was surprised by the results of the distance to development, 23%, and distance to water, 48%. I predicted that distance to development would have a high importance because the seeds originated from urban landscapes, and since Japanese barberry is highly adaptable I did not predict that distance to water would be very important. I hypothesize that these results are due to the habits of the birds that are spreading the seeds. The lower importance of distance from development could be because either the birds that eat the seeds in the landscaped areas are flying a fair distance away, or Japanese barberry has been incrementally spreading outward from these areas for long enough that it is not a large factor. I theorize that the higher importance of distance from water, which is also the second highest factor, is because the birds that are eating and spreading the seeds prefer to perch or nest in these forested areas along the river valleys close to a water source. 

This risk data could be used to monitor and control the spread of Japanese barberry in the wild in SE Minnesota. If I were to continue with this topic, I would talk to a biology or ecology expert of this area to try to identify any missing landscape factors that could help to further refine the areas a most risk. This data may be the most valuable if combined with native species locations/ habitats, especially those at risk of being outcompeted by such an aggressively invasive plant in dense forest areas.