Evaluation of inexpensive wireless sensor networks for managing small vegetable farms

Brookfield Farm

As with most projects not everything went according to plan. The sensor equipment we spent most of grant supplies money on (the Libelium Standard board) became obsolete in November and was no longer available during the project. It was replaced with a much improved, more off-the-shelf version (Plug & Sense!) but at about 20% higher cost. However with a lot of experimenting with different scenarios we were able to make the standard set Libelium sensors function albeit it late in the season. We were able to get real-time readings for the five sensors, set and receive alerts and display reports. In addition the Digi, Cygnus and NHR sensors were deployed successfully and provided useful comparisons for temperature/humidity sensors. See the next section for how the readings were used by the farmers and researchers. Also, at the very end of the project Mr. Manire acquired and was able to test recently available cellular, Ethernet and USB gateways, repeaters and sensors from Monnit and Mojyle. These proved to be the best equipment evaluated in the project and are the basis for the primary recommendations on the outreach website.

From a technical point of view we learned a lot from this equipment about how to set up and deploy wireless sensor networks including the following:

-Overheating--even if a temperature sensor itself is in a solar radiation shield, the associated electronics which do not fit in the solar radiation shield can overheat in direct sun; we found that foam reflectors of the type sold in auto parts stores for windshields make excellent, inexpensive solar and rain/irrigation shields where airflow is not important

-Batteries--lithium batteries are a must especially when it is cold; when using 15 minute observation times, batteries will generally last all growing season if the sensors sleep at low power between observations

-Interference--forests, thick vegetables, concrete floors and metal walls are very hard to penetrate with wireless signals; use routers or antenna extensions to route the signals around them

-Signal Strength--for best signal strength place antennas vertically and try to get an unobstructed line of sight between end devices/routers and the gateway; use antenna extensions to get the antenna outside of a metal enclosed walk-in for example, but extensions reduce signal strength somewhat; if the antenna is near a router or gateway then the reduced signal is not a problem; an easy way to get a USB gateway/coordinator antenna outside of a building to improve line of sight is to use a long USB cable rather than an antenna extension cable; the signal loss is much less. Another approach is to use a directional antenna such as a Yagi that focuses the signals in one direction.

-Radio Power--the power of the radio and type of antenna (external/internal) make a big difference in how far away the components of the network can be from each other. The cost is directly related to the range.

-Insects--any holes needed for air flow must be covered with screening to keep insects out; we found an entire ant colony on top of a deep cycle battery and many spider webs and cocoons in most exposed crevices and corners of our equipment that was deployed in the fields.

-Gateway Computer--these computers must be on 24/7, and ideally be dedicated so that human use doesn't disrupt them. Turn off all automatic updates for windows, and applications, and set the power settings for the computer and hard disk to never stop both when powered and on battery; however the screen may be allowed to dim or turn off. A UPS (Uninterruptable Power Supply) is helpful especially if the computer is not a laptop. In fact increasingly available “Ethernet”or WiFi gateways eliminate this type of gateway altogether. These devices connect directly to your Internet router via a cable or WiFi. The best approach is to not use a computer for a gateway at all but select gateways that connect directly to the Internet router or connect via WiFi.

-Cellular Gateways--cellular based gateways are useful when fields are not close enough to a USB gateway but, of course, require cell tower access as well as a constant, solar charged power source. We initially installed one at Brookfield on the Small Ones potato field, but the signal level rarely exceeded one bar and was not stable enough. Moving it to the Brookfield Lower Field which was more open proved to be very stable because of better cell tower proximity.

-Routers--these must be powered all the time; when AC power was not available we were able to attach a router to a 12 volt deep cycle battery which was recharged with a 35 watt solar panel.

-Deploying Sensors--we found that farms present one of the most difficult environments for deploying sensors. Fields are scattered and often miles from Internet access and even cell tower access. Sources of interference abound. Insects, animals, tractors, dust, weather and even climbing plants are potential threats, and sensors often have to be moved after early crops are harvested. It is best to test a deployment before assuming that it will work at a particular location.

-Monitoring Conditions--the ability to monitor conditions on smart phones and send alerts to farmers we found makes a big difference in whether the sensor readings get used in a timely fashion. Vegetable farmers are extremely busy all the time so convenience is very important. Using an iPhone Brookfield was able to check readings at any time and received "push" email alerts when readings exceeded certain limits.

Results at Brookfield Farm

Soil Moisture/Leaf Wetness--We found that we did not have enough time to make significant use of the soil moisture and leaf wetness sensor readings. We would have had to correlate farmer field observations with the sensor readings and establish some guidelines for how to interpret the measurements. Leaf wetness data are most useful when run through predictive models for diseases such as late blight and early blight. Dan concluded that, for Brookfield soil moisture sensors were not useful because, in the absence of rain he waters the required crops once a week anyway and that works for him. Ryan Voiland from Red Fire Farm concurred. In a situation where water is costly or difficult to deliver this sensor could be useful but it will take further research to understand how to interpret the readings. Likewise, by the time late blight struck the tomatoes we didn't have enough leaf wetness readings or experience in interpreting them to predict its occurrence.

Soil Temperature--Due to delays we were not able to get soil temperature readings in the spring when they would have been useful for scheduling planting of certain crops. However in the fall soil temperature readings were very helpful in scheduling the harvest of sweet potatoes. According to Dan's log, he noted that on October 10th the soil temperature dropped from 60 to 55.5 degrees in a week thus warning him to plan to harvest earlier than expected.

Air Temperature/Relative Humidity--these were by far the most useful sensors. Some of the more interesting entries from Dan's log are:

-received iPhone alert early in morning that Strawberry field was nearing frost conditions, took frost prevention measures April 14th

-received an email alert while I was out of town that walk-in cooler was too warm and was able to call staff and have it fixed on July 17th;

-noticed that the sensor temperature reading in the Harnois fieldhouse was at 112 degrees at noon because side roll-ups accidentally were still down; raised the roll-ups and opened the doors July 14th;

-was able to check the walk-in after a big load was put in without having to get out of bed on August 29;

-found that a sensor in the Butternut squash field showed the temperature to be four degrees colder than the local forecast and was able to take frost protection measures on Sept 16th;

-on numerous occasions found that the low temperatures in the Lower Field were four to nine degrees colder than the “local” weather forecast predicted; this was new information and will affect use of the field in the future;

-was able to check early morning temperatures from bed and avoid having to go check if the peppers needed frost protection; result, farmer got more sleep!;

-on numerous occasions was able to check temperature and humidity in greenhouse, walk-in cooler, and root-cellar to monitor vegetable storage conditions and received six iPhone alerts when temperatures exceeded limits;

-was able to assist Town of Amherst by providing a pdf of sensor readings that showed sub 28F temperatures for 2 hours at the farm which killed all mosquitoes and enabled town to lift an EEE based ban on outdoor activity October 13th.

Not only was Dan able to prevent field and storage crop problems in time but perhaps more importantly he gained peace of mind by being able to easily check conditions on his iPhone that previously would have required him or someone to check on in person.

UMass Low Tunnels Project

One of the low tunnel Cygnus temperature/humidity sensors failed in December 27, 2012. With the frozen ground and snow it was too difficult to repair. In retrospect we should have installed a “zipper” on the side of the low tunnel or used a sensor with probes that could be placed inside the tunnel so that the electronics could remain outside of the tunnel for easy repair. The failed sensor apparently had low batteries due to the fact that it was the farthest sensor from the gateway and on the edge of good radio reception and spent more power trying to connect. The batteries were replaced in the spring and the sensor started working again. The other two sensors were closer and maintained communication during the grant period. Even though the batteries were alkaline C batteries they did surprisingly well in the cold weather however we would still recommend lithium batteries.

The sensors allowed the researchers to immediately see the impact of differences in low tunnel construction. The onions tunnel used black plastic in addition to the two film covers whereas the other tunnel did not. By using the cloud software to display daily temperature range charts the impact of having the black plastic was striking. We will use the sensors to monitor temperatures in spring 2013 for timing removal of the outer plastic layer and inner row cover. See the Project Photos PDF to see these charts.

UMass Vegetable Storage Conditions Project

The two temperature/humidity sensors at Red Fire and Simple Gifts farms showed very stable and controlled temperature and humidity readings in the desirable ranges during the project. Both of these facilities were cooled and humidified with standard HVAC equipment.

The Brookfield root cellar had only a fan for air circulation and water was thrown on the floor for adding humidity. In the warm December and early January period the UMass researchers inspected the readings via the RT Cirrus cloud software and noted that the humidity was low due to the warmer temperatures in the root cellar and also found that the test carrots had lost some moisture. The farm made additional efforts to humidify the cellar and those efforts along with colder temperatures later helped stabilize the environment.