MFMichael J. Franklin

Michael J. Franklin

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Michael J. Franklin has a total of 74 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Few large tree hollows in an endangered peri-urban ecosystem highlights the need to restore suitable habitat for dependent woodland fauna

    by Michael J. M. Franklin, Christopher E. Gordon, Emma J. Gorrod, Zahra Emlaei, Jeremy Bendall, Dylan Bristol, Paul D. Rymer, Uffe N. Nielsen

    Context: Tree hollows are essential habitat resources used by fauna in woody ecosystems globally. In addition to extensive historical and ongoing clearing, human use of woodlands has left remnants with structurally modified tree assemblages, including reductions in large hollow-bearing trees. Aims: We aimed to assess tree hollow resources for dependent, arboreal vertebrates in fragmented, peri-urban woodlands. The objectives were to quantify the extant tree size distribution, estimate densities of hollow-bearing trees with different sized hollows, and determine the effect of tree size on both hollow occurrence and the number of hollows per tree. Methods: The study was conducted in 100 plots located in remnant stands across the distribution of critically endangered Cumberland Plain Woodland, south-eastern Australia. Tree size was assessed using diameter at breast height (DBH). Hollows were classed as small, medium or large during systematic searches. Bayesian hurdle models were used to analyse effects of tree size on hollows. Key results: There were 5566 live trees recorded (median DBH 10.2 cm), but only 105 were hollow-bearing. The mean density of trees bearing any size hollow was 15.6 per ha, but those bearing large hollows were scarce (3.1 per ha). Hollows were unlikely to occur in live trees with a DBH of <50 cm. Conclusions: Overall hollow availability was diminished and comparable to other modified woodlands. Large hollow availability was disproportionately low. Implications: Management of large, hollow-bearing trees in these systems should include the identification and protection of remaining individuals, actions to increase the number of individuals through time, and addition of artificial hollows.

  • Open Access

    Urban Geochemical Contamination of Highland Peat Wetlands of Very High Ecological and First Nations Cultural Value

    by Ian A. Wright, Holly Nettle, Uncle David King, Michael J. M. Franklin, Amy-Marie Gilpin

    Temperate Highland Peat Swamps on Sandstone (THPSS) are wetlands in the Blue Mountains, south-eastern Australia. The wetlands have legislative protection as endangered ecological communities. They have long-standing cultural significance for Gundungurra Traditional Custodians. Previous studies document their degradation by urban development and vulnerability to extreme weather. Water quality in our study was assessed at wetlands in protected areas and compared with others exposed to urban development. We derived water quality guidelines that are intended to help future water quality assessment at THPSS and, in particular, to detect any impact from urban development on these wetland systems. Water quality in urban swamps was consistent with the freshwater salinisation syndrome despite all the swamps having relatively low electrical conductance (−1). Urban swamp water had salinity (mean 87.3 µS cm−1) three times that of non-urban swamps (mean 28 µS cm−1). The ionic composition of urban swamp water was dominated by calcium and bicarbonate, consistent with urban alkalisation syndrome. Our guidelines instead recommend limits for pH, salinity, turbidity, dissolved oxygen, and metals detected in greater concentrations that were found in urban swamps (iron, manganese, barium, and strontium). Our results support the theory that the dissolution of urban concrete materials is a degradation process that contributes to the impairment of urban swamp water quality.

  • Open Access

    Initial Assessment of Salinity Tolerance in the Invasive Aquatic Weed Frogbit ( Hydrocharis laevigata )

    by Adrian Renshaw, Michael J. M. Franklin

    Frogbit ( Hydrocharis laevigata ) is an aquatic invasive species with a high capacity to disperse, which can lead to rapid colonisation and potential domination of the littoral zone of affected freshwater systems. Little is known about the tolerance of this species to saline waters, or whether it has the potential to impact estuaries or tidal sections of rivers. On receipt of a communication that Frogbit had been observed in an estuary in south‐eastern Australia, we conducted a glasshouse experiment with the aim of making an initial assessment of the salinity tolerance of Frogbit. Mature and juvenile Frogbit ramets were placed in open 1‐L opaque, cylindrical containers with water and six sodium chloride concentrations (control (0 g/L), 1, 2, 5, 10 and 20 g/L) replicated 12 times each. Ramets were assessed as alive or dead at seven, 14 and 21 days. Bayesian logistic regression was used to model Frogbit responses to salinity. At 21 days, all except two ramets remained alive in the control, 1 g/L and 2 g/L NaCl treatments. At 7 days, the probability of mature and juvenile survival began to decline sharply at 5 g/L and was approximately zero at 15 g/L on average. The probability of survival reduced further with time at these intermediate concentrations. All ramets were dead in the 20 g/L solution at day 21. Our results show that while Frogbit may tolerate slightly elevated salinity for short periods, it is unlikely to persist in parts of estuaries where seawater prevails. However, further research is urgently required to understand the extent to which Frogbit can occupy tidal waterways with regular influxes of fresh water on receding tides.

  • Open Access

    State of knowledge on the effectiveness of management interventions to restore degraded eucalypt woodlands

    by Michael J. M. Franklin, Emma J. Gorrod, Laura A. White, Ian Oliver, Paul D. Rymer, Uffe N. Nielsen

    To support the persistence of Australian eucalypt woodlands, conservation of remnant vegetation must be augmented with ecological restoration of degraded ecosystems. Certainty about the effectiveness of restoration interventions is urgently required to consistently transition degraded woodlands to reference states. The aim of this meta‐analysis was to quantify the effectiveness of restoration interventions to improve plant and edaphic attributes in degraded temperate and semiarid woodlands of Australia. Our structured literature search retrieved 35 studies that were suitable for analysis, which enabled assessment of six types of restoration interventions and 11 ecosystem response metrics. Effectiveness was quantified using estimates of the probability and magnitude of responses generated from Bayesian multi‐level models. We found consistent increases with varying average levels for carbon (via sugar) addition (43%) and burning (27%) on native plants, burning on cryptogams (91%), and woody debris addition on soil moisture (35%) and carbon (21%). Native plants had a low probability of benefitting from slashing (0.33) or herbicide application (0.09). Slashing had a high probability of increasing introduced plants (0.83). Planting almost always failed to achieve reference levels for native plant communities, introduced plants, or soil phosphorus. A very high level of uncertainty was evident for the outcomes of herbicide and sugar addition on introduced plants. Overall, we found a paucity of adequate studies, including insufficient quantitative information on combinations of interventions, and a lack of effectiveness in common interventions. Our results indicate an urgent need for experiments to be embedded in restoration programs to improve certainty in restoration effectiveness.


  • Open Access

    Implications of altered fire regimes for birds of dry sclerophyll forest under climate change

    by Michael Franklin, Richard Major, Ross Bradstock

    Climate change is driving increases in forest fire activity around the world. Consequently, recent and future research into the implications for affected biota, and its translation into policy and management, is of vital importance for the conservation of forest biodiversity. The dry sclerophyll forests of south-eastern Australia are affected by changing fire regimes, including having recently been extensively burnt. In this review, we synthesise contemporary research into the responses of dry sclerophyll forest birds to fire regimes, with a focus on contrasts in levels of time since fire, fire severity, fire frequency, and their interactions. We identify and discuss high-priority knowledge gaps and management issues in this domain. Recent studies have gained insights into the complexity of the effects of fire on birds by, for example, showing how avian fire responses are influenced by climate variability. In dry sclerophyll forests, most species that responded to time since fire were less likely to occur where it was short. There were mixed positive and negative responses to increasing fire severity and/or frequency among species in these forest bird assemblages. Given that quantification of the risk of extinction of rare and uncommon birds under increased fire activity is challenging, options for investigating their fire response are proposed. Availability of habitat for birds that flourish where fire is recent, severe and/or regular is unlikely to decline given current trends. We present a management approach that conceptualises large blocks of biologically connected forest with longer fire ages, but otherwise differing fire histories, as biodiversity reservoirs.


  • Open Access

    Data Station: Delegated, Trustworthy, and Auditable Computation to Enable Data-Sharing Consortia with a Data Escrow

    by Siyuan Xia, Zhiru Zhu, Chris Zhu, Jinjin Zhao, Kyle Chard, Aaron J. Elmore, Ian Foster, Michael Franklin, Sanjay Krishnan, Raul Castro Fernandez

    Pooling and sharing data increases and distributes its value. But since data cannot be revoked once shared, scenarios that require controlled release of data for regulatory, privacy, and legal reasons default to not sharing. Because selectively controlling what data to release is difficult, the few data-sharing consortia that exist are often built around data-sharing agreements resulting from long and tedious one-off negotiations. We introduce Data Station, a data escrow designed to enable the formation of data-sharing consortia. Data owners share data with the escrow knowing it will not be released without their consent. Data users delegate their computation to the escrow. The data escrow relies on delegated computation to execute queries without releasing the data first. Data Station leverages hardware enclaves to generate trust among participants, and exploits the centralization of data and computation to generate an audit log. We evaluate Data Station on machine learning and data-sharing applications while running on an untrusted intermediary. In addition to important qualitative advantages, we show that Data Station: i) outperforms federated learning baselines in accuracy and runtime for the machine learning application; ii) is orders of magnitude faster than alternative secure data-sharing frameworks; and iii) introduces small overhead on the critical path.

  • Open Access

    Recirculating Water through Concrete Aggregates Rapidly Produced Ecologically Hazardous Water Quality

    by Ian A. Wright, Holly Nettle, Michael J. M. Franklin, Jason K. Reynolds

    The use of recycled concrete aggregates as a construction material is growing and this study was conducted to investigate the potential water quality issues that arise when concrete aggregates are exposed to water. The water used in the study was from a high-conservation-value wetland. It was dilute (17.8 µS cm−1), acidic (pH 5.97) and poorly buffered. The ionic composition comprised sodium, bicarbonate and chloride ions. Water was recirculated for 60 min through a control treatment and three treatments containing recycled concrete aggregates (RCAs) of different fragment sizes (10 mm, 20 mm and 60 mm). The fragment size influenced the final water quality, but the response patterns varied between the attributes tested. Post-recirculation, the RCA treatments increased the electrical conductivity by 6 to 12 times; pH by 2.3 to 3.8 pH units; and concentrations of calcium, potassium, bicarbonate and sulphate. The water exposed to RCA materials also increased the concentrations of several metals (aluminium, arsenic, copper, lead and zinc), resulting in hazardous concentrations for aquatic species according to ecological water quality guidelines. Strontium concentrations in water exposed to RCAs increased by 30 to 120 times background levels. The results from this study added further support to a growing body of evidence that the exposure of concrete materials to water can produce environmentally hazardous water quality.


  • Fast Adaptive Similarity Search through Variance-Aware Quantization

    by John Paparrizos, Ikraduya Edian, Chunwei Liu, Aaron J. Elmore, Michael J. Franklin

    With the explosive growth of high-dimensional data, approximate methods emerge as promising solutions for nearest neighbor search. Among alternatives, quantization methods have gained attention due to the fast query responses and the low encoding and storage costs. Quantization methods decompose data dimensions into non-overlapping subspaces and encode data using a different dictionary per subspace. The state-of-the-art approach assigns dictionary sizes uniformly across subspaces while attempting to balance the relative importance of subspaces. Unfortunately, a uniform balance is not always achievable and may lead to unsatisfactory performance. Similarly, hardware-accelerated quantization methods may sacrifice accuracy to speed up the query execution. We propose a Variance-Aware Quantization (VAQ) method to encode data by intelligently adapting dictionary sizes to subspaces to alleviate these significant drawbacks. VAQ exploits intrinsic dimensionality reduction properties to derive the subspaces and only partially balances the importance of subspaces. Then, VAQ solves a constrained optimization problem to assign dictionary sizes proportionally to the importance of each subspace. In addition, VAQ accelerates the query execution by skipping data and subspaces through a hardware-oblivious algorithmic solution. To demonstrate the robustness of VAQ, we perform an extensive evaluation against quantization, hashing, and indexing methods using five large-scale benchmarking datasets. VAQ significantly outperforms the strongest hashing and quantization methods in accuracy while achieving up to 5× speedup. Compared to the fastest but less accurate hardware-accelerated method, VAQ achieves a speedup@recall performance up to 14×. Importantly, a rigorous statistical comparison using over one hundred datasets reveals that VAQ significantly outperforms rival methods even with a half budget. Notably, VAQ's simple data skipping solution achieves competitive or better performance against index-based methods, highlighting the need for new indices for quantization methods.


  • Open Access

    Understanding and Optimizing Packed Neural Network Training for Hyper-Parameter Tuning

    by Rui Liu, Sanjay Krishnan, Aaron J. Elmore, Michael J. Franklin

    As neural networks are increasingly employed in machine learning practice, how to efficiently share limited training resources among a diverse set of model training tasks becomes a crucial issue. To achieve better utilization of the shared resources, we explore the idea of jointly training multiple neural network models on a single GPU in this paper. We realize this idea by proposing a primitive, called pack. We further present a comprehensive empirical study of pack and end-to-end experiments that suggest significant improvements for hyperparameter tuning. The results suggest: (1) packing two models can bring up to 40% performance improvement over unpacked setups for a single training step and the improvement increases when packing more models; (2) the benefit of the pack primitive largely depends on a number of factors including memory capacity, chip architecture, neural network structure, and batch size; (3) there exists a trade-off between packing and unpacking when training multiple neural network models on limited resources; (4) a pack-aware Hyperband is up to 2.7x faster than the original Hyperband, with this improvement growing as memory size increases and subsequently the density of models packed.

  • How much survey effort is required to assess bird assemblages in fire-prone eucalypt forests using acoustic recorders?

    by Michael J. M. Franklin, Richard E. Major, Ross A. Bradstock

    Context: Forest fire activity is expected to increase in many parts of the globe over the course of the 21st century, with corresponding potential for heightened levels of proximate and ultimate threats to avian diversity. Landscape-scale investigations of the responses of birds in locations where current extreme fire regimes represent those expected in the future provide opportunities to identify potentially vulnerable species in advance. Autonomous acoustic recorders are well suited to survey birds in the typically large and remote natural areas with low accessibility required for these types of studies, because they offer cost-effective and relatively safe options for obtaining reliable data. Aims: The present study aimed to optimise survey using acoustic recorders to achieve a satisfactory assessment of montane dry sclerophyll forest bird assemblages using these devices. Survey completeness, or the number of species detected as a percentage of total species, was used as a metric to gauge survey suitability. Methods: Acoustic recorders were deployed in 10 ridge-top forest sites in the Blue Mountains, south-eastern Australia. Extensive field recordings were processed by an analyst, with species detected by their calls recorded in a series of 20-min samples. A results-based approach, incorporating a stopping rule that established when to conclude sampling at a site, was applied to the data. The results guided the target survey completeness and sampling effort levels assigned to a set of fixed-effort survey methods, which were subsequently evaluated. Key results: The optimal survey method involved using recordings from five 20-min sampling periods immediately following dawn for 2 days, achieving an average survey completeness level of 69%. Conclusions: The optimal survey method can obtain results that are suitable for many types of studies involving assessments of bird assemblages, because the method can detect all common and moderately common species in assemblages, plus a fair proportion of rare species. Implications: The present study has systematically developed an effective method of using autonomous acoustic recorders to research and monitor montane bird assemblages in fire-prone dry sclerophyll forests. This methodological approach may also be applied in systems subject to altered patterns of flood, storm or other extreme weather under climate change.