Hi... There are two MODIS sensors aboard two different satellites i.e. Terra and Aqua. Please check if the data is not available from both these satellite sensors. If data is available from any of the sensors, you can use it.

If above doesn't work, try to get data from VIIRS sensor aboard S-NPP satellite.

Please note that before using data from other satellites for the missing data period, you need to understand the relationship between two different AOD datasets.

I hope this helps.

Also, please mention the time period in the question, so that people can help quickly.

Thank you.

Thanks, Yogesh Sathe sir I have tried both sensors. I will look into data from the VIIRS sensor.

If you have long term data before and after gap

you can just estimate by root interpolation method

i.e. Newton Raphson method for missing data

Thanks, Nimit Godhani I have long term data and will try this method. If you have a document of how can I apply this method please attach it, it would be really helpful.

If the AOD measurements from the satellite during Jul to Sep are not available, then it is possible that clouds might have prevailed during the overpass. You can try Level 3 data that is available from Giovanni (https://giovanni.gsfc.nasa.gov/giovanni/).

In addition to VIIRS, you can try MISR and DSCOVR at their original spatiotemporal resolution. You can also double check MAIAC; even though it is derived from MODIS, it might be possible it provides retrievals even though the ordinary MODIS retrievals return no data. Finally, you can use MERRA2 at hourly resolution, or CAMS-rad at 1-min resolution (use the expert option from http://www.soda-pro.com/web-services/radiation/cams-radiation-service).

Interpolation from monthly data is highly inaccurate and should not be used!

If the sky is covered with clouds during the period that you want to capture the AOD, one way is to first develop a model to predict that missing values and then use the results, if the model performance is acceptable. I would say machine learning tools such as random forest model or ANN time series would be a good option. You can also find an association for example between the brightness temperature of Landsat8 and MODIS AOD, and then landsat8 BT might be able to predict the missing AODs.

However, I am not sure you want to develop a model to predict the missing AOD value in the meanwhile.

I hope it helps

I forgot to mention the MERRA2 AOD, which is more accurate than MODIS:

Article Worldwide validation of CAMS and MERRA-2 reanalysis aerosol ...