Novel R-CNN and transformer models for pollution impacts and land cover changes around iconic heritage sites in developing countries: a case study

Abstract

Abstract Irreversible Land Use Land Cover changes around iconic heritage sites in south Asian countries, precipitated by an ever-increasing urban density build up, is a serious cause of concern. Some of the monuments are UNESCO world heritage sites and are currently under threat. Heavy vehicular traffic and unregulated encroachments over several south Asian cities, including Indian cities have led to particulate pollution build up around iconic heritage structures. This study shows that emissions from diesel engines operating around a network of roads around the 2000-year-old Meenakshi Amman Temple complex, emit soot and black carbon particles in the accumulation mode (0.2–0.5 μm) which remain airborne for over a week and eventually soil and mar the beauty of the temple’s iconic tower gates (Gopurams). Over this size range the scavenging efficiency from natural rain showers is ~ 20%. Efficient scavenging depends not only on the rain drop size distribution, but also on the collected aerosol particle size distribution. The region of Madurai remains largely dry for 9 months in a year and municipal authorities must look for the use of pressure calibrated sprinklers optimised with the drop-size distribution of the airborne particles. Quantifying particle size distribution requires the use of in situ scanning mobility particle sizer spectrometers-SMPS -which is prohibitively expensive for developing nations. Creating artificial rain showers is an even more expensive process and is resorted to only in dire situations. From an analysis of European data repositories and with the use of the state-of-the-art Weather Research and Forecasting (WRF) model we show (i) a progressive loss of vegetative cover (and hence a decrease in the leaf area index, LAI) (ii) decrease in precipitation (iii) decrease in soil moisture and (iv) an increase in fractions of non-precipitating clouds over the region over a time period of 1940–2050. It is therefore clear that any economically affordable engineered solution must involve smart roadside sprinklers yielding a fine spray mist which will be able to capture these particles through efficient collision and coalescence until non-polluting vehicles are used. It is shown that when vehicles are in an idling mode (as is very often the case) the median size is at its broadest and progressively decreases with increasing vehicular operating loads. This study proposes a quick AI mediated method to specify the range of Sauter Mean Radii (SMR) for varying vehicular operations, completely circumventing the use of SMPS and geoengineered artificial rain showers. The drop size distribution (DSD) range will help in the choice of sprinkler attributes—a sprinkler yielding an assortment of millimetre sized droplets should scavenge much of the soot particles with moderate spray durations of the order of a few hours. Mask Recurrent-Convolutional Neural Networks(R-CNN) modelling was used to ascertain particle SMR. Future predictions spanning the years 2020–2050 using the Stacked Transformer model show a progressive decrease in the LAI and soil moisture and an aggravated increase in heat stress during the Representation Concentration Pathway (RCP) 8.5 high emissions scenario. Simulations are contrasted for business-as-usual scenarios (using polluting vehicles), with scenarios using a phased-out transition to electric vehicles (EVs) and it was ascertained that if municipal agencies replace diesel vehicles with EVs, then the above indicators significantly improved. These quick protocols can be used by government officials and other stakeholders to provide smart forecasts for Impact analyses (EIA) and cost savings assessments for calculating paybacks resulting from a partial transition to cleaner vehicular transport in the interim period. Estimates suggest benefits to the tune of millions of USD over a 5-year period vis-à-vis facade restoration ravaged by soot deposition.