Data Science Applications in Environmental Monitoring Using Remote Sensing Data, Machine Learning Algorithms, and IoT-Based Sensors

Abstract

This paper tested how information technology methods, which include satellite tv for pc faraway sensing, system studying algorithms, and sensor integration the usage of the Internet of Things (IoT), may be used to display the surroundings in Lahore, Pakistan, one of the maximum environmentally pressured metropolitan areas in South Asia. The ordinary objectives had been to acquire and pre-method multi-supply multi-environmental information on Lahore, make use of system gaining knowledge of algorithms consisting of Random Forest (RF), Support Vector Machine (SVM) and Artificial Neural Networks (ANN) to categorise air fine and are expecting land floor temperature (LST), and compare the consequences of those fashions the use of preferred statistical metrics and to try to examine the 3 algorithms in phrases in their capacity to help in early detection of environmental anomalies. The satellite tv for pc structures owned through the US Geological Survey (USGS) and European Space Agency (ESA) had been used to collect environmental statistics, which had been complemented via way of means of readings of IoT-primarily based totally air excellent sensors and statistics of the Pakistan Meteorological Department. The preprocessing strategies worried cleansing of the facts, normalization of the records, and geospatial characteristic extraction with Python programs which include scikit-learn, GeoPandas, and Rasterio. However, out of the 3 algorithms considered, Random Forest become the maximum suitable in phrases of typical overall performance in air nice category with a excessive accuracy of 91.4, precision of 90.8, don't forget of 89.7 and F1-rating of 90.2. In predicting the land floor temperature, the ANN version gave the minimal root imply rectangular error (RMSE) at 1.84degC. The evaluation found out that the 3 maximum widespread functions in all of the fashions had been the flora index (NDVI), the awareness of PM 2.5, and the class of land use. The consequences imply that unified facts technological know-how structures related to far flung sensing, gadget mastering, and IoT sensor facts can provide a sturdy and scalable platform to real-time city environmental monitoring, and the direct packages to the safety of civic fitness and environmental coverage to Lahore and different South Asian cities.