Model Reduction and Remote Sensing for Precision Irrigation Applications

• Author / Creator

  Sahoo, Soumya Ranjan

• Two-thirds of the world population faces severe water stress at least once per month in a year. The rapidly growing population exacerbates freshwater scarcity. Additionally, climate change, pollution, and bio-energy demands amplify the water demand problem. Since a large percentage of freshwater withdrawals are for agricultural irrigation, increasing water-use efficiency in irrigation becomes extremely important. Closed-loop irrigation is one of the precision irrigation techniques which has the potential to improve water-use efficiency than the traditional approaches. Due to the large-scale features of agricultural fields and significant uncertainty, there are significant challenges associated with closed-loop irrigation including sensor placement, data assimilation, and controller design.
  This thesis addresses some of the major challenges of applying closed-loop irrigation in the actual fields by proposing novel methods.

  Due to the availability of a limited number of sensors, in a typical agricultural field, it is essential to know the minimum number of sensors and optimal location of the sensors to estimate the field's soil moisture. The structure-preserving graph-based approach is used to reduce the order of a large-scale system model. A systematic approach has been developed to find the minimum number and best location of the sensors using observability and degree of observability analysis. In some irrigation implementing systems, the irrigation amount is non-uniform in spatial directions. In these scenarios, one reduced model for the whole time may not capture all the dynamics of a large-scale field or may increment the order of the resulting reduced model. Dynamic model reduction is proposed to handle these scenarios, where different reduced models are computed at different periods of time. Further, the sensor placement using dynamic reduced order is developed.

  Next, the framework for the state estimation of the large-scale field using the advanced optimization-based moving horizon estimation (MHE) is developed. The trajectory-based unsupervised machine learning method is proposed for adaptive model reduction of very large agricultural fields. Further, the algorithm of the existing MHE is modified to handle the adaptive reduced model. The proposed approach is applied to estimate the states of a real-agricultural field scenario located in Lethbridge, Canada.

  Afterward, an optimization-based closed-loop scheduler for large agricultural fields to provide optimal irrigation time and the amount is developed. The structure-preserving model reduction is used to decrease the dimension of the three-dimensional model. The scheduler has an objective similar to the economic zone MPC. In addition to that, time is also a decision variable to the optimization problem. The final objective is to maximize the yield while minimizing the water consumption and maximizing the time between the irrigation events.

  Further, the algorithm for surface soil moisture estimation using the thermal and optical remote sensing method is developed. The machine learning-based Long Short-Term Memory (LSTM) model is used to estimate the surface soil moisture. Due to the time-varying nature of the agro-hydrological model, the LSTM model is more preferred than the static neural network model. The LSTM based model is trained to obtain the surface soil moisture from the remote sensing images and the weather conditions. The developed method is applied to a real-agricultural farm located in Lethbridge, Canada using the experimental data collected in summer 2019.

  In this thesis, the details of the study area and experimental data collection procedure for experiments conducted in summer 2019 and 2020 at an agricultural field and a golf club are provided.

• Subjects / Keywords

  • Model reduction
  • Irrigation
  • Machine learning
  • Remote sensing
  • State estimation
  • Scheduler
  • Sensor placement
  • Moving horizon estimation

• Graduation date

  Spring 2022

• Type of Item

  Thesis

• Degree

  Doctor of Philosophy

• DOI

  https://doi.org/10.7939/r3-n2k7-zq42

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.