Using big data and AI to model tree root blockage patterns in urban sewer systems

Alessandro Ossola, Johannes J. Le Roux and Michelle Leishman, Biological Sciences, Macquarie University, Sydney, Australia, Alessandro Ossola, Plant Sciences, University of California, Davis, Davis, CA, Heri Bustamante and Luther Uthayakumaran, Sydney Water, Sydney, Australia

Background/Question/Methods
Tree root intrusion is the most significant factor damaging sewer pipes across urban landscapes. In Sydney, Australia, the removal of about 15,000 root blockages costs more than $10M/year, without accounting for potential public health issues, environmental and reputational costs. With urgent calls to plant more trees across cities for climate change adaptation, the future occurrence of sewer blockages are expected to increase when the urban forest-infrastructure-environment system is not properly managed.
Sydney’s sewerage infrastructure is large and complex with its pipes consisting of a variety of materials, sizes, and functions. This infrastructure further sits in a diverse urban environment, soil types and microclimate gradients. Sydney’s urban forest, composed by ~650 tree species planted along streetscapes, interacts with the infrastructure-environment system causing unpredictable root blockage patterns. Understanding which tree species disrupt urban sewers is an urgent yet challenging task.
By using big data and an artificial intelligence (AI) approach, based on distributed random forest models (DRFs), we modelled the urban forest-infrastructure-environment system (90,858 tree stems along 22,192 pipe segments) to determine which tree species were most likely to block sewers over the decade following Australia’s Millennium Drought (2010-2019) to avoid effects due to limited water availability in the urban landscape.

Results/Conclusions
Higher numbers of tree stems along pipe infrastructure and the presence of the strangler fig (Ficus macrocarpa), in particular, were the most important variables, explaining ~50% of the variability in root blockage patterns across the study area. However, infrastructure and environmental variables were more important at explaining tree root intrusion than tree species identity, stem size or useful life expectancy.
Among the top 25 species identified as problematic, most are known to have invasive root systems (e.g., Ficus obliqua, F. benjamina, Schefflera actinophylla, Washingtonia filifera). Large trees (DBH > 100cm) were more likely to block pipes than smaller trees, as well as native evergreen species. Our findings suggest that these trees species should be discouraged for planting along urban water infrastructures and from their inclusion in urban greening and forestry strategies.
Future research should use DNA barcode root material from pipe blockages and above-ground trees to validate the most problematic species. New data and insights are needed to model how sewer blockage patterns caused by urban trees might be affected by climate change, as more frequent and intense droughts and heatwaves hit Australian cities can further limit water availability for urban trees.