Home » Research » Research Symposium » Research Symposium – 2020

Postgraduate Research Symposium 2020

Impactful Research Through Interdisciplinary Approach

Postgraduate Research Symposium 2020

THEME: Impactful Research Through Interdisciplinary Approach

The technical sessions of the Faculty of Science, which was apart of the 10th Annual Research Symposium of the University of Colombo, was held on the 17th of December 2020. The Faculty of Science, that reaches its centenary milestone in 2021 is in the forefront of Sri Lanka’s Science & Technology research. At any given time, approximately 500 postgraduate students pursue their studies in the Faculty under the guidance of well-qualified academic staff members. The research conducted at the faculty is supported by the Colombo Science and Technology Cell which promotes its commercialization.

This year a total of 36 abstracts from a range of disciplines (Chemistry, Mathematics, Nuclear Science, Physics, Plant Sciences, Statistics and Zoology & Environment Sciences) were presented on a virtual platform which allowed a wider audience to join in and share knowledge and valuable research insights, not-withstanding the restrictions posed by the Covid 19 pandemic.

 
Six presenters were adjudged the winners from the four tracks
  • Physical Sciences Session – M. Iddangoda
  • Biological Sciences Session – Sahan Siriwardana
  • Chemistry: Session 1 – Hasitha Weeratunga & Shanika Fernando
  • Chemistry: Session 2 – Vimarshi Liyanarachchi
Invited Talks
Seminar Talk
The Tiger Beetles of Sri Lanka: Our island biodiversity
Professor. Chandima Dangalle Department of Mathematics University of Colombo, Sri Lanka
Tiger beetles are predatory insects of order Coleoptera, family Carabidae, subfamily Cicindelinae. 2,822 species are found from many countries of the world with the exception of Antarctica, Tasmania and a few oceanic islands such as Hawaii and Maldives. They are important as biological indicators of environmental health and diversity of other taxa, bio control agents of pests of agricultural crops and predators of ecological ecosystems. Six tribes are found in the world of which two tribes are found in Sri Lanka – Cicindelini (ground-dwelling tiger beetles), Collyridini (arboreal tiger beetles). These two tribes are represented by 70 species of which 36 are endemic. Due to the high species number, Sri Lanka is ranked amongst the top 30 countries with the highest number of tiger beetle species. Further, the country is ranked number four amongst the top ten countries with the highest number of endemic tiger beetle species, and number one amongst the top ten countries with the highest number of species per square kilometer. However, tiger beetles of the island have not been studied for more than a 100 years, and the studies available have not been systematically conducted and are far outdated. Therefore, recent investigations have been conducted by the author and co-workers on the diversity, distribution and habitat types of this biologically diverse, charismatic insect group. The investigations have resulted in disturbing revelations of decline in species numbers, decline and change in distributional ranges and loss of their habitats which is significantly detrimental to habitat specific species such as the tiger beetles. Nearly twenty years (2002 – to present) of recent investigations have recorded only 23 species of tiger beetles of which 12 species were endemic. Fourteen species were ground-dwelling tiger beetles, while nine were arboreal tiger beetles. The ground-dwelling tiger beetles have short, broad bodies with colourful elytral patterns, while the arboreal tiger beetles have elongate, narrow bodies that are uniformly dark in colour with no elytral patterns. The ground-dwelling tiger beetles were found from open habitats with sparse vegetation such as beaches and coastal areas; water edges of rivers, streams and reservoirs and human-modified habitats such as gardens, lawns and other urban settings. Arboreal tiger beetles occupied vegetated habitats such as forests, woodlands, grasslands and agro-ecosystems. However, endemic tiger beetle species of both tribes mostly occupied locations of the wet zone as is evident for many other animal and plant taxa of Sri Lanka. Endemic species were found on the banks of the Kuru River, Bopath Ella; banks of Maha Oya, Dehi Owita; Sinharaja rainforest; Bodhinagala forest, Ingiriya; Thummodara, Avissawella, Yagirala forest, Kalutara; Kanneliya rainforest, Galle; Makandawa rainforest, Kitulgala; Rambukwella, Kandy and locations in Ratnapura. Studies revealed that the endemic species are at present limited in distribution and their distributional ranges have significantly declined. For example, the endemic ground-dwelling tiger beetle, Cicindela (Ifasina) willeyi Horn, 1904 is now limited to a single location in Avissawella and 99.5% of its distributional range has declined. Most tiger beetle species have exploited new habitats due to habitat loss and competition by other insect species, which was evident by a mtDNA analysis and haplotype networking study conducted by the author. Recent 4 studies have revealed that the coastal tiger beetle, Hypaetha biramosa, is an efficient biological indicator of environmental health of the beach habitats of Sri Lanka. The arboreal tiger beetle, Derocrania scitiscabra, has the potential of being used as a suitable bio-control agent for pests of coconut, tea, betle, pepper, cinnamon and fruit cultivations. At present, the vast amount of data on tiger beetles collected during the two decades of study have been utilized to develop a machine learning application for the identification of tiger beetles. It is intended that this development will facilitate the identification of other insects as well with future modifications. Insect studies are often hindered by the difficulties in identification. Morphological identification has been weakened by the close similarities evident between species of the same genus, similarities evident between species of different genera, excessive time consumption and lack of skill; while molecular identification has been set back mainly due to cost of equipment and other resources. Morphometric and habitat data of tiger beetles collected during the 20 years of study and images captured in the field excursions have been utilized to develop two machine learning models – Ensemble Extra Tree Classifier method based on morphometric and habitat data, and the Revised SqueezeNet Tranfer Learning Approach based on image data. The two models can be integrated to develop an approach that is more accurate, efficient and user-friendly. Insects are the Earth’s most diverse organisms, and Sri Lanka’s insect fauna represents a large part of its biodiversity. Unfortunately, despite this dominant position they are rarely included in the accounts of biodiversity on the Island. The usual excuse for this negligence is lack of information on Sri Lankan insects due to the inadequacy of taxonomic studies on the group. The recent studies have succeeded in providing information on a very important insect group on the island. Diversity of tiger beetles, their applications as biological indicators and bio control agents, and utilization of their information in developing a machine learning approach to identify insects has provided a foundation of vast knowledge and data for future studies. However, it is imperative that the necessary resources to study our rich insect fauna are made available sooner while some of the past diversity is still with us.
Seminar Talk
Environmental research dealing with climate change and its link to the carbon cycle
Professor Erandathie Lokupitiya Department of Zoology and Environment Sciences University of Colombo Sri Lanka
Professor Erandathie Lokupitiya
Climate change has become one of the most serious environmental issues of the century. Global temperature has increased over time, and the last decade included the majority of the years with the highest recorded global temperatures. The observed warming on Earth has been attributed to the increased levels of carbon dioxide (CO2) and other greenhouse gases due to various anthropogenic activities including fossil fuel burning and land use changes. The focus on the issue of climate change has increased during the last few decades due to the irreversible impacts it has caused across the world. Atmospheric CO2 concentration has increased from 280 ppmv in 1850 to ~410 ppmv in 2019 (NOAA 2020). Therefore, research dealing with human impacts on climate change and perturbations of the carbon cycle has become an important area in environmental research. Climate change related research could aim at both adaptation (i.e. action to reduce the impacts of climate change) and mitigation (i.e. action to reduce the sources or greenhouse gas emissions associated with climate change). Sri Lanka has been an island vulnerable to a variety of impacts of climate change, including impacts from increased frequency of extreme weather events such as droughts and floods, and sea level rise, etc. Such impacts can negatively affect the absorption of atmospheric CO2 by various ecosystems. Therefore, research focusing on the climate change impacts, including those on the carbon cycle, and the associated adaptation and mitigation aspects, has become a timely need. The research led by the author at the Department of Zoology and Environment Sciences has focused on a variety of environmental concerns/issues including air pollution, climate change and its link with the global carbon cycle. Specific research areas include the analyses of observed climatic trends and their impacts on various ecosystems, potential adaptation measures, evaluation of the existing carbon stocks and stock changes in specific terrestrial and coastal/marine ecosystems to identify their role in mitigation of climate change and atmospheric CO2, measurements and modeling of the greenhouse gas emissions, etc. These research studies have incorporated multidisciplinary approaches and analyses involving measurements, modeling, GIS, and remote sensing, etc. and have been funded through both local and foreign grants, published in indexed journals, and presented at various scientific meetings. In conducting the research, the author has collaborated with both local and foreign scientists and research institutes.
Seminar Talk
Statistical methods for detecting disease outbreaks
Professor C.D. Tilakaratne Department of Statistics University of Colombo, Sri Lanka
Savini Bandaranayake
The World Health Organization defines a disease outbreak as the occurrence of disease cases in excess of normal expectancy (https://www.who.int/). An infection, transmitted from one human to another, animals to humans, from the environment or other sources, can result in a disease outbreak. It is worth to take precautionary actions before a disease becomes widespread rather than handling large numbers of patients within a short period of time. Public health authorities of a country/government needs to identify the exact outbreak location in advance in order to take necessary steps to control the spread of a disease. Health Indicator Surveillance, Symptom based Surveillance and Internet based Bio surveillance are some of the methods used for disease surveillance (Buckeridge, 2007). However, these surveillance systems are time consuming and cost ineffective. With the development of rapid data collection tools, shareable database platforms and modern data analytical tools, computerized disease surveillance systems have today become very popular. These systems are cost effective and has the ability to detect a diseases outbreak location very quickly. Data related to emergency department visits, pharmacy sales, hospital hotline calls, ambulance dispatches has to be effectively analysed to detect disease outbreak areas. Statistical methods play a prominent role in analysing such data. In Syndromic surveillance medical data are analysed to detect or forecast disease outbreaks (https://en.wikipedia.org/wiki/Public_health_surveillance). A variety of statistical methods are applied in past studies to detect disease outbreaks. A widely used statistical method to detect a shift in the number incidences (infected cases or deaths) by epidemiologists is the CUSUM charts with the assumption of Poisson distribution. However, this method has limitations due to the size, the structure of the at‐risk population and the baseline rate which may not be constant during the period of surveillance. An improvement to the CUSUM procedure, based on the normal approximation to a Poisson process was proposed to overcome these limitations (Rossi, Lampugnani & Marchi, 1999). Forecasting the number of patients/deaths using time series approaches, such as ARIMA models, which is also a popular statistical method applied in disease surveillance. Studies done by Reis and Mandl (2003) and Burkom, Murphy and Shmueli (2007) are a few examples for such applications. The aforesaid statistical methods take only the time dimension into account when detecting a disease outbreak. However, in general, we say that there is a disease outbreak if a significant number of incidents is evident in a comparatively small area within a short period of time. Kleinman, Lazarus and Platt (2004) pointed out that the ignorance of spatial variation is a weakness of some widely used statistical methods in detecting a disease outbreak. They applied an approach incorporating generalized linear mixed models to detect incident clusters of a disease in small areas. In addition to generalized linear mixed models, multiple linear 7 regression, Poisson regression and logistic regression have been also used to detect disease outbreaks (Wieland, Brownstein, Berge, & Mandl, 2007). Scan Statistic is a density-based clustering technique which is used to detect clusters in a series of randomly positioned points (Tilakaratne & Liyanage-Hansen, 2018). This is the most popular statistical method used in detecting disease outbreaks. Spatial Scan Statistics (Kulldorff & Nagarwalla,1995) is the first version of scan statistic that was applied for detecting spatial clusters of a disease. Ignorance of the time dimension is a major drawback of this method. Then space-time scan statistic (Kuldroff, 2001) which accounts for both space and time dimensions in detecting disease outbreaks was introduced. Thereafter, several versions of space-time scan statistics were developed from time to time and those versions are currently very popular. However, all these scan statistics assumes Poisson distribution to identify significant disease clusters and this assumption has limitations as the occurrences of an infectious disease are not independent of each other. Other statistical methods which were mentioned beforehand, are also associated with distributional assumptions. Failure of distributional assumptions will lead to false alarms or non-detection of disease outbreaks. Hence, it is worth to consider distributional free statistical methods to detect disease outbreaks. Mann Whitney scan statistic (Cucala, 2016) allows the detection of clusters in continuous data indexed by time or by space, without using any distributional assumptions. However, inability of taking into account both space and time simultaneously prevents the application of this method in detecting disease outbreaks. Under the author’s guidance, Hettige (2019) developed the Space-time Mann Whitney scan statistic by extending the Mann Whitney scan statistic which was applied successfully to detect disease outbreaks using a simulated data set. High computational time taken for calculating space-time Euclidian distances between individual data points is a drawback in this statistic. Future research directions must focus on the reduction of the computation time of the Space time Mann Whitney test. Application of an improved clustering algorithm to identify disease clusters is one option. Instead of ARIMA models, the spatial-temporal aspect can also be incorporated by applying the Spatio-temporal Autoregressive models to forecast the number of infected persons.