GIC has completed data collection and analysis in Ubon Ratchathani province to support its efforts in mapping plastic waste pathways in the Lower Mekong River Basin.

GIC partnered with Ubon Ratchathani Rajabhat University to handle local data collection efforts. The methodology used was developed by GIC and involves capturing high resolution roadside video with a camera attached to a passenger vehicle. 

The Ubon Team collected data in June and July 2021. In total, 558 km were covered in Ubon Ratchathani’s Muang and Warin Chamrap districts. These areas were selected based on GIC’s spatial analysis for plastic leakage pathways completed the previous year during Phase 1 of the UNEP CounterMEASURE project. Weather conditions played a critical role in data collection progress as testing demonstrated that precipitation negatively affected data quality. Data collection was frequently postponed due to inclement weather conditions as this period coincided with Thailand’s rainy season. Despite these delays, Ubon Ratchathani data collection was completed in nine days for 5-6 hours each day. 

Quality control was established at the onset of the data collection campaign by staging test runs with the Ubon Ratchathani field team. GIC staff evaluated the test data and provided recommendations for improvement to achieve optimal data quality. The final data volume was nearly 900 GB. 

Following data cleaning, several university students from around Southeast Asia were recruited as volunteers to annotate instances of plastic waste in the Ubon Ratchathani dataset. Annotations were performed using pLitter, an online plastic waste annotation platform developed by GIC. The annotated data was used to train a Neural Network to identify plastic waste in roadside images. A partnership with Google led the GIC team to use Google’s Auto ML platform for deep learning analysis. An advantage of using AutoML is that it subjects datasets to  numerous deep learning models to find the one with the best output, reducing some of the effort for the analyst. 

After training, the deep learning model can detect roadside plastic waste from the Ubon Ratchathani roadside images. GIC then used the locations of plastic waste to create a heat map depicting areas in Ubon Ratchathani with the most plastic litter that poses a risk to enter local waterways. These hotspots are being shared with UNEP and Ubon Ratchathani municipality to design a plan of action to work towards reducing the plastic problem in the Lower Mekong River Basin. 

The next step for GIC in CounterMEASURE Phase 2 is to reproduce this work in Chiang Rai province. Data collection in Chiang Rai ran from September 26 through October 3, 2021 for approximately 600 km of roads in the Chiang Saen District. Analysis will soon be underway
and a plastic waste heat map will be developed for the Chiang Rai province study area.