1st Edition
Optimisation of Dynamic Heterogeneous Rainfall Sensor Networks in the Context of Citizen Observatories
Precipitation drives the dynamics of flows and storages in water systems, making its monitoring essential for water management. Conventionally, precipitation is monitored using in-situ and remote sensors. In-situ sensors are arranged in networks, which are usually sparse, providing continuous observations for long periods at fixed points in space, and due to the high costs of such networks, they are often sub-optimal. To increase the efficiency of the monitoring networks, we explore the use of sensors that can relocate as rainfall events develop (dynamic sensors), as well as increasing the number of sensors involving volunteers (citizens). This research focusses on the development of an approach for merging heterogeneous observations in non-stationary precipitation fields, exploring the interactions between different definitions of optimality for the design of sensor networks, as well as development of algorithms for the optimal scheduling of dynamic sensors. This study was carried out in three different case studies, including Bacchiglione River (Italy), Don River (U.K.) and Brue Catchment (U.K.) The results of this study indicate that optimal use of dynamic sensors may be useful for monitoring precipitation to support water management and flow forecasting.
1. Introduction
1.1 Background
1.2 Motivation
1.3 Innovation
1.4 Objectives
1.5 Layout of this thesis
1.6 Highlights
2. Literature review and proposed framework
2.1 Introduction
2.2 Sensors and sensor networks
2.3 Models of precipitation for rainfall-runoff simulation
2.4 Simulation of Rainfall-runoff processes using lumped conceptual models
2.5 Classification of approaches for sensor network evaluation
2.6 Proposed framework for sensor network design
2.7 Conclusions
3. Case studies
3.1 Introduction
3.2 Bacchiglione River
3.3 Brue Catchment
3.4 Don River
4. Advancing Kriging methods for merging heterogeneous data sources in nonstationary precipitation fields
4.1 Introduction
4.2 Dealing with data of variable measurement uncertainty
4.3 Estimating uncertainty due to partial recording
4.4 Handling Non-stationarity in the kriging framework
4.5 Application in the Brue Catchment
4.6 Conclusions
5. Optimisation of static precipitation sensor networks and robustness analysis
5.1 Introduction
5.2 Formulation of decision variable encoding
5.3 Selection of decision variable encoding and of optimisation algorithm
5.4 Exploring relationships between various objective functions
5.5 Solving the optimal design problem for the selected objective functions
5.6 Analysis of robustness
5.7 Conclusions
6. Optimisation of dynamic precipitation sensor networks
6.1 Introduction
6.2 Posing the optimisation problem
6.3 Objective functions and corresponding strategies for deployment
6.4 Experimental setup and solution of the optimisation problem
6.5 Results and discussion
6.6 Conclusions
7. Conclusions and recommendations
7.1 Summary
7.2 Conclusions
7.3 Limitations
7.4 Outlook and recommendations
Bibliography
ANNEX 1. Overview of candidate algorithms for sensor network optimisation
ANNEX 2. Hydrological models used for the Brue catchment
ANNEX 3. Perturbation specification for simulating incomplete precipitation data
Biography
Juan Carlos Chacon-Hurtado is a Civil Engineer from the Pontificia Universidad Javeriana Cali, with an MSc Water Science and Engineering with specialisation in Hydroinformatics from UNESCO-IHE, and with various academic and professional interests related to the use of advanced ITC tools to address water-related problems (hydroinformatics). Juan Carlos has experience in areas of hydrological modelling, optimisation, uncertainty analysis, data assimilation, water loss control and scientific programming. Juan Carlos is currently a Postdoc researcher in multi-criteria decision analysis applied to sewer asset management at the Delft University of Technology, The Netherlands.
