An ēKo-nomic solution for Nursery Monitoring

If you took a look at the plants in my yard, or had caught a glimpse of the few potted plants I attempted to care for in college it would be quite obvious that my thumbs are not green. The soil would usually be too wet or too dry and the leaves wilted leading to my plant's eventual demise. Imagine having acres and acres of plants to monitor and care for...is there a way to do this ēKo-nomically?

The FLOW-AID project is working to contribute to the sustainability of irrigated agriculture by developing, testing in relevant conditions, and fine-tuning through feedback, an irrigation management system that can be used at farm level in situations where there is limited water supply and water quality. The FLOW-AID project in collaboration with the University of Pisa has installed an ēKo system at an experimental nursery in Tuscany, Italy to monitor soil moisture at eight different locations in the nursery.

The system is designed to serve as an assistant for communication with higher level water management systems at basin scale for long and short term water use planning and prediction. This project integrates innovative sensor technologies into a decision support system for irrigation management while taking into consideration several factors in a number of third country partners. The ēKo nodes have been deployed in eight locations over the nursery in Tuscany. The ēKo ES1101 soil moisture sensors are monitoring the ornamental shrubs and trees being grown to make sure that all the water is being used efficiently and effectively.

The project results yielded will showcase the development and testing of new and innovative, but simple and affordable, technical concepts for irrigation under deficit conditions used at the farm level in a large variety of set-ups and constraints. It will show the development of a water management support system (DSS) that contains an expert system (off-line/long-term) to assist in farm zoning and crop plan in view of expected water availability (amount and quality) with a link to Basin Management, as well as a crop response module that can be incorporated into the irrigation scheduler that allocates available water(s) among several plots and schedules irrigation for each one with a link to Basin Management.

The FLOW-AID project has set up four test sites in various market conditions with different irrigation structures, crop types, local water supplies and constraints. The hardware/software systems used must adapt the general concept of water management to the local situation by using appropriate parts of it at the global sites in Lebanon, Jordan, Turkey and Italy.

The information being collected at the site in Tuscany, Italy, by the researchers at the University of Pisa for container crops and nursery grown crops is available to users over the internet via ēKo's EG2100 gateway device and the ēKoView interface. This device provides, in a fully integrated package the connection between ēKo Sensor Nodes deployed and the ēKo Gateway. The work done by FLOW-AID will be carried out between 2006 and 2009 as a 6th Framework European project under the call for water in agriculture, new systems and technologies for irrigation and drainage. For more information on the ēKo system, click here.

Not just a pipe dream!

US water facilities and those around the world are faced with mounting operational and maintenance costs as a result of aging pipeline infrastructures. The ability to monitor and control the infrastructure is no longer a pipe dream but is on its way to becoming reality thanks to wireless sensor networks. PipeNet, is a system designed by researchers at Imperial College in London and CSAIL at MIT in conjunction with Intel Research, to collect hydraulic and acoustic/vibration data at high sampling rates as well as use algorithms for analyzing this data to detect and locate leaks. A study by the EPA (Environmental Protection Agency) estimates that water utilities will need $277 billion over the next 20 years (2003-2023) to install, upgrade, and replace infrastructure. Unfortunately, identifying the high priority areas is a non-trivial task because of the scale and age of the pipeline infrastructures. Failures of large diameter bulk-water transmission pipelines are of greatest concern as these are supply critical systems. When these failures do occur, there are dire consequences including loss of life, severe interruptions in service, degraded fire fighting ability, damage to adjacent infrastructure and buildings, and of course the multi-million dollar repair bills.

PipeNet is a system based on wireless sensor networks which aims to detect, localize and quantify bursts and leaks and other anomalies in water transmission pipelines such as blockages and malfunctioning control valves. The system was based on Intel Motes (the 1st generation of the Imote2 devices available today) to provide remote monitoring in near real-time with support for high data rate time synchronized data collection from multiple locations. This is a significant change from current data acquisition practices of using portable data loggers with a low duty cycle and limited remote monitoring stations that do not have the capability to do local processing or high-bandwidth transmission.

The Motes were responsible for the data collection, local signal processing and relaying of data to the second tier consisting of the Stargate platform that was to relay the data back to the backend server via a GPRS modem. A special sensor board was also designed to interface to the Mote to accommodate the various analog sensors used in PipeNet. The clusters contained pressure sensors and pH monitoring sensors combined with water level monitors and pressure monitors. The mote could be configured to trigger data acquisition on a channel remotely when a monitored channel exceeded some threshold. Acoustic and vibration data analysis was used to detect and locate small leaks which are difficult to identify with hydraulic data. Vibration signals collected showed differences in the mode of wave propagation if a leak was detected. Leaks also manifested themselves in the acoustic signal which propagates uniformly in both directions away from the leak by the escaping water flowing through the rupture in the pipeline.

The Motes communicated from the manhole to the Stargate based gateway deployed on a nearby lamppost. The Stargate, the GPRS modem and 802.11 radio were powered from the power lines at the lamp post. The Intel Mote was connected to the Stargate through its UART interface acting as a bridge between the Stargate and the motes on the pipes. This cluster head was responsible for forming the sensor network, converting the configuration data coming from the Stargate and passing it to the correct sensor node as well as delivering the data collected via the reliable transport protocol to the Stargate where it was converted back into data files. These files were periodically sent to the backend server running in the lab via the GPRS link. The Stargate was equipped with an 802.11 link to facilitate drive-by access for on-site configuration and debugging as well. Data transfer was handled via standard Linux tools, and the data files were then loaded in a Postgres database that stored the individual sensor readings. This gave users the ability to browse these sensor readings by connecting to an Apache Web server running on the server. The web site used Google Maps/Google Earth to allow users to select and browse the sensor locations of interest. Once users selected a sensor location, they could retrieve data corresponding to a user-specified date / time range and sensor type to visualize the data.

Using the leak localization algorithms they developed, the research team was able to localize leaks to within 30 cm. The long term monitoring of pressure and acoustic signals in particular pipes will also facilitate the development of more accurate pattern recognition and classification models in the future. The next revision of the PipeNet system is using the Imote2 platform which integrates many essential components to enable high performance and energy efficient data processing. The XScale processor on the Imote2 has dynamic voltage and frequency scaling capability to allow applications to balance performance and energy needs by selecting speeds between 13 and 624 MHz. In addition, the processor includes a DSP co-processor to accelerate common data analysis primitives (e.g FFT, compression) thereby greatly improving performance and energy efficiency. This performance advantage will allow users to carry out the analysis and data reduction in real-time, thus reducing storage and power. Finally, the Imote2 includes 32 MB of SDRAM and Flash enabling the decoupling of data collection and communication with a richer peripheral support that will provide higher data acquisition rates and improve sensor integration.

PipeNet is the future of pipeline monitoring providing the capability to automatically detect leaks and bursts of water in the transmission pipelines; real-time operation with few false alarms; inexpensive to produce, install, and maintain; high-frequency data collection; the ability to differentiate between sensor and system faults; and a flexible reusable data-flow based programming environment. This system will not only improve our ability to monitor large scale water supply systems, but to conserve our natural resources and use them efficiently.

Heated up about your Energy Bill? Motes to the Rescue!

by Ralph Kling, Chief Architect, Crossbow Technology, Inc.

When I opened up my Gas and Electric bill from our local utility (PG&E here in Northern California) I was shocked: it was over $100 more than I expected. Some further checking confirmed that it was also substantially higher than during the same time last year, by about that amount. So what happened? Sure, energy prices likely have increased since last year and maybe it was a bit colder here the last month. But it still didn’t add up… 

A few years ago I had installed a shiny new programmable digital thermostat to replace an old mechanical gizmo that must have been around since the house was built in the 1950s. And I actually did read the manual (well, some of it anyways) and spent quite a bit of time programming it with different cycles for days and nights and weekends and so on. And, though I was very proud of that accomplishment at the time, the promised energy savings that I had hoped for actually never really materialized. My utility bills stayed pretty much the same before and after the thermostat installation. 

So after receiving the last bill, I investigated the thermostat but it seemed to work just fine and none of the settings had been changed since I had originally made them. After contemplating this for a bit, I had an idea. Since I am working at Crossbow, the leader in Wireless Sensing technologies I have access to the essential tools needed to solve this mystery: I took home a Wireless Sensor Starter Kit which consists of two sensor nodes and a base station. I also added a Stargate Netbridge gateway to record and visualize the data from the sensor nodes without the need to install anything on my PC (or even have it running and consuming energy during the measurements). 

The next day, I installed the “Sensor Network” in my house. It was really simple, just plug the gateway into my home router, place the sensor nodes in the rooms I wanted to monitor and turn them on. All networking and data gathering happen automatically from then on. The sensor nodes are battery powered thus they can be placed pretty much anywhere eliminating the search for outlets. I placed the nodes in the kids’ rooms in the center and back of the house. The nodes have all sorts of sensors including light, humidity, atmospheric pressure etc. but I was particularly interested in the temperature readings. 

The next day, I pulled up the temperature profile gathered so far and I was shocked:

 

The Thermostat was set to 70°F, (about 21°C) during the morning and evening and to 65°F, (18°C) during the rest of the day and at night. The measurements show an average of 24°C during the high and 20°C during the low period, a full 2-3 degrees more than expected. Why were the readings so far off from the preset values at the thermostat? And then it dawned on me: the thermostat was in the living room at the other end of the house. And I remembered that last year in the summer I had closed the heating air vents in that room to prevent the kids’ Lego pieces from falling into the shafts (they are really hard to retrieve from there). Of course I had forgotten about that in the fall when I restarted the heater. 

So the warm air had to make its way through half the house before the living room (and the thermostat) warmed up thus raising the temperature in the other rooms way to high (as well as the heating bill). But this was still a theory that needed to be proven. So I opened up the living room vents all the way, let the sensors collect more data and waited impatiently for the data from the next day. The results were nothing short of astounding:

 

The temperatures in the kids’ rooms now averaged 21°C during the high and 19°C during the low periods, much closer to the desired thermostat settings. Beyond that, the data also shows that one room warms up more poorly than the other (it has more outside walls with poor insulation). Adjusting the vents in those rooms as well could further equalize the temperature readings. 

So, with a very simple setup (it took literally 5 minutes to set up) I was able to gather a lot of valuable data and hopefully significantly reduce my heating bill (we will see next month). And all that without lowering the temperature settings on the thermostat! Beyond that there are lots more insights that can be gathered from the data: The slope between the day and night settings shows how fast the house cools down and how well it is insulated. Other data like the humidity measurements can point to potential problems with mold if the readings are too high. Do your kids always “forget” to turn off the lights – well a plot of the data can very convincingly show how much energy is wasted. 

I am sure there are lots more good ideas – reader’s suggestions are welcome!

Dr. Ralph Kling is currently the Chief Architect of the Wireless Business Unit at Crossbow Technology. Ralph is leading the Wireless engineering team at Crossbow and is responsible for new product strategies, technical directions and Standards activities.

Previously, Ralph was Principal Architect and Director of Sensor Network Operation at Intel Corporate Research. In this capacity, he was responsible for ground-breaking research in the area of Wireless Sensor Network Platforms that resulted in such novel designs as the Intel Mote and Imote2. Before joining Intel Research, Ralph's previous assignments include managing the Itanium® Processor Family microarchitecture/ performance group and the Microprocessor Research Lab (MRL).

Ralph obtained his Master's and Ph.D. degrees from the University of Illinois at Urbana-Champaign. His thesis research focused on Simulated Evolution, a new global optimization method for integrated circuit designs. Prior to coming to the US on a Fulbright scholarship, Ralph studied Electrical Engineering in his hometown of Hanover in Germany. He enjoys skiing in the winter and beaches in the summer.

Wireless Soil Moisture Tension Measurements for Irrigation Management

Irrigation is defined as the artificial application of water to the soil usually for assisting in growing crops. In crop production it is mainly used in dry areas and in periods of rainfall shortfalls, but also to protect plants against frost. Irrigation management in agriculture and landscaping is of growing importance as the growing global population puts more demand on finite fresh water supplies. Managing irrigation optimally improves yields and quality while reducing water user and pumping energy costs. Optimal irrigation management requires reliable knowledge of plant water stress and soil moisture status. Many different devices and techniques have been used to gather this type of information, but perhaps none as successful as one of Crossbow's beta ēKo deployments in California's Napa Valley wine country.

Mark Holler is the owner of Camalie Vineyards in Napa, Califorina. He is a viticulturist and a technology enthusiast who has been working closely with Crossbow in the development and testing of the ēKo platform over the past two years to increase the quality and quantity of his grape harvest by using and controlling his water resources. With the data he collected from the ēko platform, Mark has been able to minimize his water use and maximize his yield despite the low water season we saw this past year in 2007. This achievement was not only due to the ēko system's ability to collect data, but Mark's ability to analyze the data and apply it to his growing techniques. Mark has written a white paper on High Density, Multiple Depth, Wireless Soil Moisture Tension Measurements for Irrigation Management. Below is an extract regarding his application and findings. To read the entire article click here:

When sampled sufficiently at appropriate depths soil moisture tensions were found to correlate well with pressure chamber measurements of midday leaf water potential in Cabernet Sauvignon grape wines. Sampling 2-3 sites per acre across a 4.4 acre hillside vineyard produced a substantial correlation of midday leaf water potentials to soil moisture tensions at 24" depth. The correlations were performed on soil moisture data and pressure chamber data from the 2007 irrigation season on the Mount Veeder hillside vineyard on the western slopes of Napa Valley. The data suggests that  soil moisture tension measurements may be able to replace many leaf water potential measurements which are significantly more labor intensive. A strategy for use of soil moisture tension measurements in manging regulated deficit irrigation of grape vines and the monitoring of other irrigation system parameters using the ēko Pro Series is described in this overview.

Correlations were done between leaf water potentials and soil moisture tensions acquired at 12" depth and 24" depth. Data from all locations and times were combined for these correlations. Sample size was 43 points per depth.The soil moisture data at 12” depth does not correlate with the leaf water potential measurements but, at 24” depth there is a “substantial” correlation.  This data suggests that deeper placements of the soil moisture sensors might produce better correlations with the leaf water potentials. 

From the data gathered one could also conclude that the vines were getting their water from deeper depths and that the vines have not concentrated their root growth around the sub surface dripper which is co-located with the soil moisture sensor at 12” depth. This information was useful in deciding not to move the subsurface drippers further from the vines or deeper to encourage root growth.This type of correlation could be used to optimize locations for soil moisture sensing. In an initial deployment many sensors could be placed at different depths at a few locations for the first season. At the end of the season correlations with leaf water potentials could be done and the root zone locations with best correlations determined. The following season more sites would be added with fewer soil moisture sensors per site only at the optimal location(s) in the root zone determined. 

The general success of the 2007 growing season at this vineyard in terms of yield, ripeness and reduced water use supports the use of the modified regulated deficit irrigation though indirectly because there are many confounding factors which affect yield. In 2006 data from the soil moisture sensors was used to optimize irrigation durations and intervals. Soil moisture sensors provide good insight into how water moves within the soil – hydraulic transport, something that leaf water potentials cannot provide. The delay between wetting at 12” depth and 24” depth is a measure of how long it takes water to move downward within the soil. From this the vertical hydraulic conductivity can be inferred. The slope of the drying transient indicates how fast water is moving away from the sensors either due to diffusion or plant uptake. 

Irrigation durations and intervals were optimized to achieve desired average soil moisture at 24” depth. This soil moisture target was based on leaf water potentials as described above. Total available water supply for the season was also considered. We adopted the premise that the vines benefit from reduced variability in soil moisture over time. The best uniformity over time would be achieved by very short durations at frequent intervals. Short durations and frequent intervals, however, do not allow the water to penetrate very far between irrigations. Short intervals also result in non-uniformities across each block because the line pressures are below spec for constant drip rate during start and stop transients. The total start up and shut down transient time for this irrigation system was determined to be about 30 minutes. We set the minimum irrigation duration to 2 hours to make the transient effects less than 25% of the irrigation duration. We then checked to see that the water was reaching the 24” deep sensors consistently with an interval equal to the time it took the 24” depth to dry out to the level before the last irrigation. The interval was then varied to bring the average soil moisture level at 24” depth to the target value. We then looked at the water consumption rate of our optimized duration/interval times and forecasted total use for the season.

If this use was in excess of our water resource we lengthened the interval to the consumption rate we could afford. We then monitored the new average soil moisture and spot checked leaf water potentials to determine if we could keep the vines from becoming over stressed. If the leaf water potentials continue to drop to –15 bar and beyond as was the case in the 2007 season we purchased additional water and trucked it to the vineyard. In 2007 in light of a very dry winter rainfall we delayed irrigation until a higher stress level was achieved to reduce canopy growth and subsequent water consumption by the vines. Our yield and fruit maturity results suggest that this was a good approach. We feel strongly that high water stress transients during the growing season can damage the vines not only in the short term but over several seasons as well.

Camalie used a prototype network during the 2005 and 2006 growing seasons to guide irrigation decisions in the 4.4 acres of Camalie Vineyards. Yield per vine in 2005 was double that of the 2004 yields for same age vines yet the water consumption was kept constant.  Typically water consumption goes up with canopy size which more than doubled for these 2.5 year old vines in 2005. The grape quality was excellent. Of course, some of this success was due to generally better than average weather in 2005 but, Mark and others at Camalie believe their visibility of the soil moisture played a significant role.  Extra drippers were added to some areas of the vineyard based on the soil moisture data.   Also irrigation intervals were shortened based on sensor data to reduce the amount of water that penetrated below the root zones where it would be wasted. In 2006, the third year for their vines, the yields again doubled from 4 tons to 8 tons. In the 4th year, 2007, the network was upgraded to the latest Crossbow technology, the yield again doubled to 16 tons of fruit that was sold and another 1.5 tons that the vineyard made into wine themselves. The yield was 3.97 tons per acre which is very rare on Mt. Veeder especially with water limited due to less than half the normal rainfall in the winter of 2006/07.  Water had to be purchased but thanks to their precision irrigation the vineyard minimized water purchasing and still had great yields.  Fruit quality was excellent as before.

For more insight into the methodology used at Camalie Vineyards, be sure to read the complete white paper here.