This pilot project aims to demonstrate success of the AquaWatch system concept on a pilot scale for Australian coastal water bodies. Through development of new partnerships with regional water managers in two locations, the pilot study provides ‘testbeds’ for new approaches to integrate and visualise multiple space-based observations (maps of sediment, dissolved carbon, chlorophyll-a and diffuse attenuation of light) and ground-based sensor water quality data (temperature, salinity, fluorescence, backscattering, turbidity and hyperspectral reflectance) streams, combined with hydrodynamic model outputs (salinity, temperature, stratification and current
vectors).

The result will be a demonstrator ground-to-space water quality monitoring tool with a prototype front-end visualisation tool (web/mobile-based dashboard) to display spatial maps of water quality parameters, time series of hydrodynamic, biogeochemical and optical parameters and statistical analysis to aid in decision making. In the longer-term, outcomes from the project can be in the form of new sensor technologies, new software platforms and/or packaged products. The ultimate aim is the demonstrated value of the integrated technologies, their benefit to end-users and their potential adoption. These are required to inform the development of the AquaWatch mission. Results will be available to inform the relevant industry sectors for commercial development to the benefit of the Australian economy.

This project aims to demonstrate the potential of machine learning (ML)-based algorithms and models in decoupling the complex optical signature in coastal waters and deriving water quality parameters in coastal waters from satellite observations. Through collaborating with CSIRO researchers and our industry partners, this project will investigate the potential of new machine learning approaches in predicting water quality parameters through fusing data from insitu water quality sensors and satellite observational data. We will develop a machine learning model to invert the remote sensing reflectance signal and to derive WQ parameters.

The result will be a machine learning-based modelling tool to improve the water quality products and thus enable better coastal ecosystem management. New ML-based remote sensing products from this project will help end users in the Cockburn Sound and Moreton Bay regions improve their coastal monitoring and management practices.

Blue Carbon, carbon that is stored in aquatic ecosystems, has been identified as an essential component of the global carbon cycle with the power to reduce inequities amongst nations through valuation of carbon sequestration in a global Blue Carbon economy. Foundational works discussing Blue Carbon include only three component ecosystems of Blue Carbon, namely mangroves, seagrasses, and salt marshes. Kelp is however capable of very high rates of primary production, with past work recording kelp fixing above 3000 g of carbon per square meter per year. Recent work by Filbee-Dexter and Wernberg (2020) estimated the carbon content of kelp forests in the Australian Great Southern Reef alone to be roughly equivalent to 3% of global Blue Carbon. Kelp forests have additionally been shown to sequester much of the carbon they fix through debris exports to deeper waters.

Comprehensive and accurate Blue Carbon estimates must therefore include kelp forests. Further study is however needed regarding how to include kelp forests into Blue Carbon budgets. Space-based remote sensing, for example multispectral imagery publicly available through the Landsat and Sentinel programs, provides the high quality, periodic data needed to effectively study kelp at a large scale. Considering that past research into Blue Carbon has had an outsized focus on the Global North (e.g., seagrasses, though distributed throughout both hemispheres and all inhabited continents, have been most heavily studied in the Mediterranean and North Atlantic), relying on open data sources that provide planetary coverage, such as those mentioned above, to develop a kelp detection workflow would avoid reinforcing pre-existing geographic biases. If successful, using satellite imagery to map and monitor kelp forests could thus create a global, standardized inventory methodology upon which conservation policies and a
robust Blue Carbon economy could be based.

This proposed doctoral research aims to supply the scientific knowledge required to develop and assess a system using earth observation data to map and monitor kelp forests and their biophysical properties, specifically in Australian waters, to enable their inclusion in Carbon Accounting Systems and a Blue Carbon economy. There are four proposed objectives:

1. Evaluate the utility of space-based Earth Observation (EO) in monitoring kelp forests at the regional to national scale, considering ecosystem characteristics that
   may affect monitoring success (eg canopy density, water depth, species composition);
2. Develop an Earth Observation system for mapping and monitoring kelp forest extent and biomass in Australian waters;
3. Quantify the sources and values of uncertainties present throughout kelp extent and biomass mapping processes in the project; and
4. Produce a national-scale kelp inventory using publicly available satellite imagery and estimate Australia’s kelp carbon stock and sequestration values with their associated uncertainties.

This work will thus produce not only scientific knowledge and a reproducible methodology, but also a data product that can be used in management and policy decisions, addressing a significant Australian and global knowledge gap.

Space is becoming increasingly congested, contested, and competitive due to increased economic access to orbit, launches of distributed satellite systems, and conjunction events due to debris and conflict. For future satellite missions, on-board algorithms that can detect, track, identify, and characterise events and hazards that occur on-orbit will be required for space activities. One activity, autonomous maneuvering, is yet to be fully realised with the latest developments in artificial intelligence. Ability to test these
algorithms on the ground is limited and the capability does not exist in Australia. This project will commence with the development of a facility, and develop novel machine perception, navigation, guidance, and machine learning algorithms on space-based hardware.

Natural hazards, including bushfires and floods, are a constant threat to world’s ecosystem. The occurrence of fire events, for example, depends on multiple factors such as the type of fuel present in the vegetation, temperature of land etc.

Data gathered by different satellite systems can be a key in predicting future fire events and damage assessment. With the availability of large amounts of data, Artificial Intelligence (AI) systems can be developed to help minimize the damage caused by these natural hazards. With data growing exponentially, annotating it is becoming a next big challenge. However, due to the heterogeneity of remote sensing measurements and tasks, there is no single go-to dataset that could serve the purpose of benchmarking. Moreover, annotating this type of data is also a challenge.

To deal with this issue the development of supervised learning algorithm is being shifted towards self supervised learning. These systems operate with high precision and make real time decisions with little to no human intervention. However, the black-box nature of these models gives rise to another set of questions such as which classifier is best for a given condition, what factors are controlling the accuracy of the model. we aim to develop an explainable and interpretable selfsupervised deep learning model that can infuse multi-modal data gathered from satellites and open data sources.

Development of an interpretable decision support system can help organizations take timely actions to prevent the spread of a natural disaster such as forest fires. Our approach would involve training an interpretable vision transformer in a self-supervised manner that could fuse information from multi-modal satellite and geospatial data. For this purpose, Data can be acquired from a range of satellite data available. Datasets of PRISMA, MODIS, Sentinel-2 and Landsat can contribute to a great deal in the development of AI based systems for hazard detection, area mapping and damage assessment.

Australia has a total of 134 million hectares of forest, which is equivalent to 17% of Australia’s land area. Of this total forest area, determined as at 2016, 132 million hectares (98%) are ‘Native forests’, 1.95 million hectares are ‘Commercial plantations’, and 0.47 million hectares are ‘other forest’. Australia has about 3% of the world’s forest area and globally is the country with the seventh-largest forest area. The Australian ecosystem is shaped by fires for over 70 000 years and each ecosystem has its fire regimes. Additionally, fuel management and predicting flammable areas are the key to managing wildfires. These factors play a vital role in resource allocation, mitigation and recovery efforts. Forest fire is a major ecological disaster, which has economic, social and environmental impacts on humans and also causes the loss of biodiversity. Therefore, it is important to know and understand the behaviour of fire ignition and spread so that fire management agencies can prevent and mitigate wildfires. This project aims to develop a tool to predict Forest Fire Spread using Machine Learning approach and Weather Research and Forecasting with fire spread model (WRF-SFIRE).

In this project, fire risk probability mapping, Prediction of fire points, and fire spread modelling will be carried out for the flammable forest areas in Australia. The fire risk probability model will be prepared by using the two-step Analytic Hierarchy Process (AHP) approach. The fire risk probability model will also be used for cross-validation of Support Vector Machine (SVM) model outputs. Then the Prediction of fire points will be done using SVM model (Linear kernel, Polynomial kernel, Radial kernel, sigmoid kernel) taking elevation, slope, aspect, Soil Moisture Layer (SML), Land Surface Temperature (LST), vegetation type layers for training the data set. Then the Weather Research and Forecasting (WRF) model will be used for obtaining meteorological data for the input of fire spread model using the Global Forecast System (GFS) as source data. After that, the fire spread path will be traced for various recent past fire event. The additional variables will be used for spread modelling other than SVM are, canopy cover, wind, temperature. The Cellular Automata (CA) fire spread model will setup in any two different locations of Australia.