The first day of the summit was focused on reviewing current activities across Melbourne and San Diego. Matt gave an update and led the discussion on the RISER project activities, and Kevin led a discussion ab
out Phoenix RapidFire. Ilkay Altintas summarized activities in the WIFIRE project from the UCSD side, and gave a demonstration of the Kepler workflow software being developed and used for managing fire modelling workflows.

Webcams, traffic counters, closed-circuit TVs, smartphones, iBeacons… We are surrounded by sensors and information systems that are capturing and communicating live data about particular phenomena, events, and trends around them. Sometimes, data sensed for one purpose may be useful for another purpose, called “opportunistic” sensing.

In order to interpolate a large surface from massive scattered data more efficiently, tessellated spatial windows have been implemented in parallel on the data streams. A divide-and-conquer approach has been used. The scattered data in each tessellated region are processed to interpolate the subsurface within the same region. The subsurfaces are then recombined to generate the whole surface. The interpolation in the tessellated regions can be processed in parallel in InfoSphere Streams. Peter tested the Intra-PE loading split parallelization of the spatial interpolation operator on a dual CPUs machine (see the figure below where the color scheme denotes processing elements). Processing the streams in parallel improves the efficiency of spatial interpolation significantly. Peter is currently working on parallelisation over multiple hosts now. This is expected to further improve the performance since more CPUs are available on more hosts.

Peter upgraded the stream operator for spatial interpolation. The new version of this operator uses in-operator memory more efficiently, which allows multiple operators to run in parallel. Peter also added an option to the operator which controls how the grid positions are loaded and saved. If the number of the grid positions is not too large, they can be loaded and saved in the in-operator memory before the scattered data are received. Once all the scattered data are received (a “flush” event of a stream’s tumbling window), the operator can begin interpolating the surface immediately. Otherwise, data can be saved in persistent memory (a file or a database) and be queried after the scattered data are all received. This improved operator uses less memory, at the cost of increased latency, spending more time waiting for grid positions.

Peter and James modified 13 sensor motes for the new external battery scheme, and more motes will be modified soon. These motes will replace a number of motes in the field Olinda that suffered complete battery depletion. Effective battery management is one of the most challenging aspects of the practical network deployment. Peter also made and tested a prototype of the waterproof battery case and a third-party USB charger (see the photo below). Replacing batteries can be highly time-consuming. The new scheme should dramatically streamline vital battery maintenance, allow USB-charging of batteries at the base, with hot-swapping of charged in the field. Further battery kits will be implemented soon, following field testing of the new scheme.

Allison attended in Taiwan a workshop on UAVs at the National Cheng Kung University. The outcome of this is helping in the selection of a UAV for the RISER project to support our ongoing research in grassland curing monitoring. Allison also attended the UN FIG conference in Malaysia where various discussions were held on disaster management technologies and their transfer to developing countries. Finally, RISER is currently working on supporting an initiative by the CFA in Anglesea to deploy a sensor network in collaboration with local primary schools as part of a disaster management education outreach project. RISER will provide training and support in the deployment and use of these sensors.

Peter implemented polygonal, extent-based spatial windows in InfoSphere Streams (see below). The spatial window can be easily combined with a temporal window to create a spatiotemporal window. Peter is testing several spatial interpolation methods (IDW, natural neighborhood, thin plate spline, and kriging) to interpolate a temperature surface within the spatial window from simulated massive scattered data. The efficiency and accuracy of different interpolation methods under different configurations of spatiotemporal window is being compared.

To test our procedure of fitting fuel moisture content (FMC) data to the different models in the literature, we have been using some data supplied by Kevin. The data involves field observations of FMC and associated weather observations. The original data was collected every 3 hours from 7am to 7pm over a number of days in an open area and in a forested area. The data was interpolated to hourly observations. Peter then fitted two hourly models of FMC (based on Catchpole et. al. and Matthews et al.) to the hourly observations. As anticipated, the predictions show a good fit to the models (see below). A similar process will be used to fit further models to the data, and to fit the models to data collected in our coming chamber and field experiments.

Allison is one of the chief investigators on a new funded NDRGS (Natural Disaster Resilience Grants Scheme) project. The project, led by RMIT, concerns using GNSS to improve prediction of severe weather, and involves support from DEPI, BoM, the CRC-SI and the UK Met Office.

Planning for the WiFIRE/RISER research summit at UCSD has progressed well. The summit is planned for August 11th/12th at UCSD, with Allison, Matt, Peter, and Kevin planning to represent RISER. The program includes a day of discussions on collaboration, followed by a day of writing and concrete collaboration planning.

Last update we reported on a new collaboration emerging with other researchers working in the area of fuel moisture content (FMC) monitoring at the University of Melbourne, including Tom Duff, Petter Nyman, Craig Nitschke. Inspired by Petter’s works on estimating FMC by inserting soil moisture probes (VH400) into the fuel layer, Peter designed a plan to log the analogue output signal of the VH400 probes using a waspmote in RISERnet.

Matt gave a talk about the RISER project at the Institute for Geoinformatics (iFGI) at the University of Münster, Germany as part of a symposium to celebrate the 20th anniversary of the institute. The symposium was entitled “Geoinformatics: Solving Global Challenges,” and Matt was one of five speakers from around the world invited to share their views of how geoinformatics research can contribute to solutions for grand challenges like emergency response.

The past fortnight saw two meetings with UCSD RISER collaborator Jessica Block as part of the process of planning further close collaboration between the UCSD-led WiFire project and the RISER project. A research summit in San Diego is planned in August 2014 to plan collaborative research activities, in particular around the fuel moisture content and stream processing/workflow computational frameworks used across the two projects. A further enduser workshop to showcase the outcomes of the collaboration in Melbourne in 2015.

Peter have been investigating the potential of conducting various spatial analyses using the IBM InfoSphere Streams architecture. Although the current spatial extent of RISERnet is small, the project also consider the issues that arise when scaling to networks of thousands or more nodes. The current investigation considers using spatial stream processing to achieve efficient and accurate spatial interpolation. For example, Kriging is one method used extensively to interpolate spatial data. But the Kriging matrix computations required for massive, scattered sensor readings are computationally expensive, and not well-suited to rapid, online stream processing. We are testing the computational efficiency and accuracy of interpolation stream operators by decomposing the sampled area using different schemes. We will also investigate the impacts of different sampling schemes on the performance of the spatial interpolation in streams. For example, the figure (right) shows a Delaunay and Voronoi decomposition of a simulated wireless sensor networks. Efficient stream processing requires relies on such spatial “windows” which affect both the efficient and accuracy of the resulting interpolation.

Azadeh has investigated the temporal pattern of the RISERnet data by decomposing the RISERnet data to seasonality, trend, and residual. She also studied the spatial characteristics of the RISERnet data through variography analysis and prediction maps. The knowledge gained through this preliminary analysis is helping us prepare for when the full network data becomes available in the Spring.

Peter investigated algorithms for fitting dynamic models to noisy observations by minimum least mean square error estimation. The chosen algorithm was tested by simulation before using real observations from the chamber experiment. FMC observations were simulated by adding Gaussian noise to the output of an FMC model proposed by Matthews et al. The fitted model can generate predictions that match the simulated observations (see the figure, right). Once experimental FMCs are measured, these observations will be fitted to the different models using this algorithm, which allows us to compare the validity of these models.

Matt and Peter have designed the experiments for validating three selected fuel moisture content (FMC) models. Winter weather means that there is currently not enough variation to FMC for field validation. We are therefore planning a preliminary experiment on a fuel and soil sample in a growth chamber. In the growth chamber, the environmental conditions can be artifically controlled. The experiment will vary the temperature, humidity, light, and wind speed condi- tions inside the chamber. One RISERnet mote in the chamber will then monitor the environmental condi- tions. Hourly changes to FMC will be measured using the FMC meter. The chamber validation will then help supplement the subsequent field validation, giving useful information about the performance of the different models in a controlled environment.

On 5th May Matt presented a paper on behalf of RISER at the 2014 Australian and New Zealand Disaster and Emergency Management Conference (ANZDMC) in Gold Coast, Queensland. The paper was entitled “RISERnet: Situational Awareness using Environmental Sensor Networks,” and gave an over- view of the RISERnet subproject, introducing the network itself, the system architecture, and the overall motivation for RISERnet. The paper was well received and prompted several questions from the audience about next steps. The conference itself also provided an excellent opportunity to discuss the RISER project more broadly with a range of stake-holders.

On 13th May Matt, Allison, Sarah, and Peter visited CFA in Burwood East for a follow-up meeting with Daniel and David in connection with the CFA Grassland Curing Program. We discussed potential collaborations between RISER and the CFA team on developing applications and information processing systems for field observations of grassland curing. In particular, the possibility of using wireless sensor networks or UAVs to automate and extend field observations were discussed, and further actions are planned around this growing collaboration.

This week we welcome to the team Aadit Shrestha. Aadit is a new PhD student with a background in GIS, electrical engineering, and sensor networks. Aadit is starting by investigating techniques for fusion of spatially and temporally dense (but uncertain) wireless sensor network data streams (such as RISERnet data) with sparse but authoritative data (such as from the Bureau of Meteorology). His initial focus will be to investigate probabilistic approaches, such as Kalman filtering, Bayesian estimation, and kernel recursive least squares.

The second RISER Project Management Committee (PMC) meeting was held at the Office of the Fire Services Commissioner on 23rd April 2014. Matt, Sarah, Kim, and John were at the OFSC, with Jessica and Juerg participating via teleconference. After summarizing the progress in RISER over the past 4 months, the discussion opened more broadly to review and planning. Amongst the key topics discussed were included:

The spatial correlation in the RISERnet data is being analyzed by calculating the semivariogram of the data, using Moran’s I and Geary’s C autocorrelation statistics. The results show that data captured at different time of a day showed very different levels of spatial autocorrelation. Current work is applying the geostatistics methods described above (e.g., cokriging) to investigate the relationship between different environmental variables, and between these environmental variables and topography.

Peter has continued his work on the implementation of further spatial interpolation methods using the IBM Infosphere Stream Processing platform. Ordinary and sample Kriging and cokriging interpolation have now been implemented. This represents a major milestone in the MoistuRISER platform, as we now have completed key foundational stream operators. Peter is currently documenting the usage of these interpolation operators and summarizing the main steps for building stream programming language (SPL) operators with C++.

Antony Galton from University of Exeter, UK, will be a visiting academic connected with the RISER project at the University of Melbourne during his sabbatical in November 2014. Antony is a world-leading researcher in the area of the ontology of geographic information, and has worked extensively in connection with ontologies for emergency management.

Matt, Allison, Kevin, Azadeh, and Peter have had an abstract accepted to the AFAC (Australasian Fire and Emergency Service Authorities Council) 2014 conference, to be held in Wellington, New Zealand, September 2014. The abstract was accepted as a poster.

Matt Duckham, Allison Kealy, Azadeh Mousavi, Kevin Tolhurst, and Peter Zhong. High-detail fuel moisture content monitoring using the RISERnet sensor network. In: Australasian Fire and Emergency Service Authorities Council, 2014.

Recent work on RISERnet has focused on improving the battery system on the nodes. James replaced non-rechargeable batteries in about 15 motes with new rechargeable batteries, recently arrived from the supplier in Spain. Working together, Peter, James, and Simon have agreed on a hardware modification that will allow the mounting of rechargeable battery packs outside of the motes. This change will significantly reduce the time and labor required to swap batteries. In addition, the new battery packs will be capable of being charged separate from the sensor motes, not possible with the current configuration. These updates should significantly accelerate the deployment and reduce the maintenance load for the RISERnet network.

Peter implemented a generic C++ primary operator for the IBM Infosphere Stream Processing platform for spatial interpolation of RISERnet data. The incoming RISERnet point data is interpolated to a regular grid in real time by this interpolator. The new stream operator is based on tumbling window-event handlers and supports algorithms which includes inverse distance weighting (IDW), natural neighbour interpolation, and thin plane spline. In the future, further interpolation algorithms such as Kriging and GWR (geographically weighted regression) are planned.