OVERVIEW

The soil pylon is a 1D sensor array deployed vertically in the soil to measure moisture and energy fluxes, along with chemical fluxes of interest.  In this case, we are concentrating on nitrate, which is the contaminant of interest at the Palmdale wastewater reuse test bed.  The approach is to develop and test the pylon hardware and software locally, then install 2 pylons at Palmdale in a pilot deployment, and finally in a scaled up Palmdale deployment.  At each step, we are embedding 1D soil simulation models in the system to support in situ pylon calibration and irrigation management strategies.

APPROACH

The systems and experimental sequence for this project is as follows:

• Design, fabricate, and test a prototypical soil pylon in the CENS and UC Merced labs, beginning with moisture and temperature sensor arrays
• Incorporate commercial and CENS-fabricated nitrate sensors into the soil pylon design
• Develop and test 1D unsaturated flow, energy and solute transport models for use in conjunction with soil pylons
• Develop and test multiobjective irrigation management algorithms for use in conjunction with soil pylons

SYSTEMS / EXPERIMENTS

(i) Soil pylon prototype.  The prototype soil pylon is being developed in a 12-inch diameter 2 m long PVC column (Figure 1).  The pylon is currently configured with 3 ECH20 (Decagon Devices, Pullman, WA) soil moisture sensors (1, 3 and 5 ft depths), 3 Watermark (Irrometer Co. Riverside, CA) moisture sensors (same depths), and 6 thermistors as 3, 6, and 9 inches; and 1, 3, and 5 ft depths (BC Components 10K Ohm).  Two types of moisture sensors are being compared, with the Decagon sensors representing the more expensive type characterized by a faster response time and the Watermark type representing less expensive, slower response sensors.  A MICA2 (Crossbow Technologies)) mote with the sensorboard (MDA300) is connected to the programming board  to allow continuous powering in the laboratory setting (MIB510; see photo). :Prior to field deployment, the power supply will be switched to batteries recharged via Heliomotes.  Data are transmitted from the wireless motes to a SPB400 Stargate Gateway, and then delivered to a MySQL database via the network connection. The Stargate accommodates a GPRS card for remote DAQ.

Figure 1.  6-m tall soil pylon test column at UC Merced (left) and hardware setup for continuously powered laboratory DAQ (right).

(ii) Incorporation of nitrate sensors into soil pylon.  The potentiometric nitrate microsensors are being incorporated into the soil pylon design.  Before this can happen, more progress must be made in the microsensor packaging with respect to environmental conditions.

(iii) 1D soil flow and transport simulators  An algorithm for solving the 1D version of the Richards equation (unsaturated flow through soil) was developed and is being incorporated into the DAQ system to provide real-time parameter estimates based on sensor data.  The model and parameter estimation routines were tested using the soil moisture data acquired at the Palmdale test bed.  Soil moisture parameters (K_s, θ_s, θ_r, α, n) were optimized by minimizing the sum of the squared-error using the Levenberg-Marquardt algorithm.

where b is parameter vector with elements of (K_s, θ_s, θ_r, α, n), θ_j* is measured water contents at a specified depth and times t_i.  For the first estimation trial, n was chosen as fixed at 1.7, and K_s, θ_s, θ_r, α were estimated. The soil moisture content was measured with TDR in a controlled irrigation experiment on October, 2004. The simulation  (Figure 2) for optimization ran for four hours and the deviation from the measured data was determined using a one minute sampling interval.

Figure 2.  Soil moisture data obtained from the Palmdale TDR sensor deployment and the best-fitting simulation by the 1D soil moisture propagation model.

(iv) Multiobjective irrigation management algorithms.  After an extensive literature review, Receding Horizon Feedback Control (RHFC) was selected as a method for managing the multiobjective application associated with the Palmdale wastewater reuse/irrigation test bed.  The basic idea of RHFC is to execute the optimization over a time interval known as the prediction horizon and repeated by moving the same prediction horizon one step forward. The first step optimal control value is applied to update state information and the updated information becomes the initial condition for the next optimization, where the feedback comes into play.  To test the RHFC scheme, a numerical solution to the 1D Richards equation was embedded in Feedback Control scheme and executed to control an irrigation regime. The optimization function for this example is

where T is prediction horizon, θ (t) is moisture content at a depth of interest, and θ_threshold is an arbitrary number (in this case, 0.25), which is estimated as an indicator that too much water is accumulating at the lower end of the simulation domain. As the problem formulation expands to include solute transport, an analogous nitrate threshold concentration will be investigated. The objective function honors the two key objectives of this problem; by allowing the moisture content up to the threshold value (without violating it) the other goal of maximizing the wastewater input is satisfied.

Figure 3 below presents the results of RHFC with 0.25 of soil moisture threshold. (a) water content in soils with respect to depth at the end of 50 management steps (b) (c) (d) optimal value, optimal application rate, and soil moisture content at the node of the maximum moisture content with respect to depth at the end of each management step.

Figure 3.  Test of the RHFC algorithm in the context of a simple irrigation problem where the volumetric moisture content is contrained to be less than 0.25:  (a) final moisture distribution in the soil profile, (b), (c), and (d) objective function value, application rate, and maximum water content in the soil profile over the course of the management horizon.

Major ACCOMPLISHMENTS

The main accomplishments associated with this project have been:

• Design and development of the soil pylon sensor array for precipitation, soil moisture, temperature, and water quality sensors, such as nitrate
• Developing unsaturated soil simulation and management models to support sensor network calibration efforts
• Characterization of the geospatial soil hydraulic conductivity properties at the Palmdale field site
• Completion of 1D infiltration simulator for pivot geometry

Satisfactory results are being obtained from laboratory (2 meter column) soil pylon, and field deployment at the Palmdale, CA experimental agricultural plot (30 acres) in summer 2005.  The deployment plan calls for the staged installation of 12-15 pylons.

FUTURE GoalS

• To extend real-time simulation capabilities to support in situ sensor calibration.

These goals will be realized by achieving the following objectives/milestones:

• Develop in situ and self-calibration methods and algorithms for differentiating between sensor signals and variations imposed by one-dimensional flow, transport and reaction processes.
• Develop accurate and compact methods for representing noisy information from the sensors and sensor network, and for incorporating uncertainty about inter-nodal environmental media.
• Extend the modality of the soil pylon device to additional sensors (plug-and-play pylon reconfiguration features).

The general approaches and timelines associated with these objectives are as follows:

• One-dimensional (column) studies and accompanying real-time simulation and parameter identification routines will be used to develop in situ calibration schemes for the soil domain.    [July 2005 – January 2006] 
• In situ and self-calibration methods will be tested in the field based on one-dimensional findings.  [June 2005 – June 2006]
• Include additional sensors for redox active chemical species (major water quality ions, above-ground microclimate sensors (e.g., evaporation, relative humidity, etc.); this will be initiated in collaboration with the Weizmann Institute through a 1-yr BARD grant. [July 2005-July 2006]

PEOPLE

FACULTY

Prof. Thomas Harmon, School of Engineering, UC Merced
Prof. Deborah Estrin, Computer Science, UCLA
Prof. Jennifer Jay, Civil & Environmental Engineering, UCLA

STAFF

Dr. J. Eric Haux, School of Engineering, UC Merced
Mr. Mohammed Rahimi, CENS – UCLA

GRADUATE STUDENTS

Ms. Yeonjeong Park, Civil & Environmental Engineering, UCLA
Mr. John Ewart, Environmental Systems, UC Merced
Ms. Nithya Ramanathan, Biology, UCLA
Mr. Tom Schoellhammer, Computer Science, UCLA
Mr. Naim Busek, Computer Science, UCLA

UNDERGRADUATE STUDENTS

Mr. Hendra Tjahayadi, Computer Science, UCLA
Mr. Obimdinachi Iroezi, Computer Science, UCLA