Supplementary Code and Video for "Local PM2.5 Hotspot Detector at 300 m Resolution: A Random Forest-Convolutional Neural Network Joint Model Jointly Trained on Satellite Images and Meteorology"

In the previous version, I forgot to include the trained random forest model. This version adds that in the Code zip file. The code and supplementary material for the paper "Local PM2.5 Hotspot Detector at 300 m Resolution: A Random Forest-Convolutional Neural Network Joint Model Jointly Trained on Satellite Images and Meteorology" at https://www.mdpi.com/2072-4292/13/7/1356. If you ever find the code helpful for your research, please cite this paper (https://www.mdpi.com/2072-4292/13/7/1356). We wanted to give you the best coding experience, but unfortunately Planet's Data Use Agreement prohibits us from sharing the Planet Imagery used in this paper. So you will not be able to run the Part A and Part B of the Jupyter Notebook files. But we did upload the PM2.5 predictions made by the RF-CNN Joint Model on the Planet Imagery. So you can play around with the Part C of the Jupyter Notebook files. Video S1. Temporal evolution of all the images used in the model training at an example training site #18 Mandir Marg (the training site that has the most samples) in chronological order. Video S2. Temporal evolution of all the images used in the model testing at an example test site #37 NSIT Dwarka (the test site that has the most samples) in chronological order. Video S3. A video that shows the RF-CNN-predicted daily PM2.5 maps of Delhi on the most trustworthy 43 days in Figure F1 frame by frame. Each day’s prediction map saved in a single file can be found in 43PredictionMaps.zip