Urban Farms as Green Infrastructure for Urban Heat Mitigation

Green infrastructure provides many ecosystem services within urban areas, including the potential to help mitigate the urban heat island (UHI) effect. Green infrastructure to help reduce the UHI often focuses on increasing tree canopy cover, and for very good reasons: trees provide shade that reduces air and surface temperatures below the canopy; transpiration from trees increases the latent heat flux thereby reducing the amount of net radiation available to heat the air (sensible heat) or the ground or other structures (ground or stored heat); and evaporation of water from the pervious soil beneath trees also contributes to an increase in the latent heat flux and a reduction in the sensible and/or ground/stored heat fluxes.

An urban farm is a type of green infrastructure that may be especially important in urban areas that historically had less investment in tree cover and parks and that now experience higher urban temperatures compared to other parts of a city. Urban farms often lack significant tree cover, however, because tree shade reduces the amount of solar radiation needed for photosynthesis and crop growth. Despite their relative lack of tree cover, urban farms contain actively transpiring vegetation and evaporation from pervious soils that can enhance latent heat flux relative to the sensible and ground/stored heat fluxes, especially if the farms are actively irrigated. In this work I examine whether and to what extent a neighborhood urban farm can help moderate local temperatures.

This study focuses on the neighborhood around Frogtown Park and Farm ("Farm"), an urban farm in St. Paul, Minnesota, USA. Temperature data were collected for the six-year period 1 Dec 2015 - 30 Nov 2021 at the Farm and at three additional sites within about 600 m of the Farm: a smaller urban garden site ("Stones"), a residential front yard ("Blair"), and a residential backyard ("VanBuren"). Temperatures were logged at 15-minute intervals using HOBO U23-002 or U23-004 sensors that were mounted approximately 2 m above the ground on a wooden or metal pole and housed in naturally ventilated radiation shields. Data quality control included checks against local all-time temperature records (none exceeded local records), checking for data spikes (a temperature change of ±2^oC or more compared to previous and subsequent 15-min observations), and checking for temperature shifts (a temperature change of ±4^oC or more compared to previous or subsequent 15-min observations). All spikes or shifts at a given site that were not corroborated by at least two other sites were removed from the record. The remaining 15-min data were downsampled to hourly values to create time series with the same temporal resolution as the nearest National Weather Service stations at the Minneapolis-St. Paul International Airport (MSP) and the St. Paul Downtown Airport (STP), both of which are located within the Minneapolis-St. Paul urbanized area. The retained hourly temperatures were rounded to the nearest whole degree Fahrenheit to be consistent with the available precision of the MSP and STP data and all temperatures were subsequently converted to degrees Celsius before analysis.

Because the hourly time series contain outliers, I used the nonparametric Kruskal-Wallis (KW) analysis of variance (ANOVA) to test for statistically significant differences in mean monthly temperature between and among the Frogtown neighborhood sites (Farm, Stones, Blair, VanBuren) and the urban airport sites (MSP, STP). Due to missing data at Stones in 2018, the KW tests for August through November are performed using five years of data rather than the full six year period. When the null hypothesis of no temperature difference between sites was rejected (p≤0.05), I used the Tukey-Kramer test to determine which sites were significantly different from each other.

ANOVA results show that, for December through May, there are no significant differences in mean monthly temperatures across the six sites, with one exception. This result is consistent with previous work in Minneapolis-St. Paul that suggests that land cover differences have little impact on intra-urban air temperature when there is persistent snow cover, little to no actively growing vegetation, and low net solar forcing. The exception occurs in May, when the mean temperature at STP is significantly cooler (by up to 0.4^oC) than the other five sites. This may be due to the particular geographic setting of STP, which is located alongside the Mississippi River, with cool spring water temperatures likely contributing to the cooler May temperatures at this location. ANOVA results for October and November show no statistically significant differences between MSP, STP, and the urban farm/urban garden sites (Farm and Stones) but both MSP and STP are significantly warmer (by up to 0.8^oC) than the front- and backyard sites (Blair and VanBuren). This may be due, at least in part, to shading from trees or nearby buildings that reduce mean temperatures at Blair and VanBuren as compared to the two airport sites, which have no tree cover and few nearby structures.

Between June and September, mean temperatures at all of the Frogtown neighborhood sites are significantly cooler (by up to 0.7^oC) than at MSP, consistent with the expectation that cooler temperatures occur in areas with actively growing vegetation and pervious soils as compared to areas with little vegetation and more impervious surface (which is the situation at MSP). The neighborhood sites show no significant differences with temperatures at STP, which has less impervious surface as compared to MSP, in addition to being alongside the Mississippi River. There are no significant differences between mean temperatures at the Farm and the other neighborhood sites, which suggests that, on average, the relative lack of trees at the Farm does not negate its heat-mitigation capacity during the warmest months of the year.

Acknowledgement: My thanks to Frogtown Park and Farm and the Frogtown neighbors for their generosity with hosting the temperature sensors.