8

I'm trying to build an air pollution model. So far I have built one for point sources. Now I'm looking at monitoring air pollution with satellite images, based on the idea that every basic pollutant somehow reacts with EM.

What is a sufficient number of sampling points?

Yes, I know, it depends on the size of area. But right now I have only 10 sampling points (monitoring stations) for a Landsat image (a really big area). The Landsat image covers part of central Slovakia, ca. 23,000 km2.

I'm afraid I can't complete my master's thesis without adding my own measurements, but that's difficult to implement. Maybe a possible way is to add data from measuring stations that are located in neighboring countries. But that would still only make 28 sampling points..

ABCD
  • 213
  • 1
  • 5
  • What is EM in your field ? –  Sep 10 '15 at 12:18
  • @gansub gases and particles may absorb and reflect at different wavelengths( same as ca. plants reflect in the infrared) I have correlation with measured pm and landsat TIR band..and Im looking for mor to build regression models and try to predict pm concetration. – ABCD Sep 10 '15 at 12:31
  • why not use an existing model - why build your own ? –  Sep 10 '15 at 12:37
  • @gansub I dont know If there really is some model. I have seen some papers where they build own models. I have NDVI, NDBI( building index), and humidity, temperature, elevation etc... also there is difference between measured temp. and temp. captured with satellite. So far yet with 20 sampling and 8 control points my predicted pm conc. differs from hundredths to tenths. – ABCD Sep 10 '15 at 12:46
  • If you want the interaction of pollution with the atmosphere and sun... you need a model like WRF-CHEM or CAMx or CMAQ, etc. You could work on revising your question to something specific and get a better answer. – f.thorpe Sep 10 '15 at 17:55
  • @farrenthorpe and interaction between pollution and satellite scaning? – ABCD Sep 11 '15 at 08:41
  • @Fina are you referring to "averaging kernels" that are used to compare model data with satellite data? Every satellite product is different and so without specific information about what satellite data you are using it is tough to get specific. You mention LandSat imagery... but LandSat doesn't collect data about air pollution. – f.thorpe Sep 11 '15 at 18:54
  • @farrenthorpe and finding correlation between measured pollution and ndvi, tepmperature etc... Than build linear regression model and predict pollution is it possible? – ABCD Sep 11 '15 at 19:41
  • 1
    @Fina if you could get PBL height from airplane data or from a good weather model... you could create a simple model that predicts PM for a specific location (e.g. based on prior measurements). If you are looking to correlate PM measurements with satellite data... you should be looking at an instrument that does aerosol measurements like MODIS or VIIRS. – f.thorpe Sep 12 '15 at 04:49
  • thanks @farrenthorpe I didnt consider about PBL and for the MODIS...I have seen their products but spatial ressolution is 250m my goal is something like this link – ABCD Sep 12 '15 at 11:05
  • @Fina In my opinion that paper is not well written and doesn't prove much. It's a model that has been tested for one day in one region that performs poorly. And even if it did perform well... I don't see how it would be used to help inform people about air pollution... especially with so many other better sources of air pollution data. – f.thorpe Sep 12 '15 at 16:05
  • @farrenthorpe so you think it is unhelpful / useless to do it lthis way? I will look at MODIS but isnt spatial ressolution too big for Slovakia or part of country? Also data which I can collect are only coordinates of stations (30p for 49,000km2 ) and sources with emissions. Or you think that it would be better(harder) to try to write a prediction algorithm for a set of pollution data? – ABCD Sep 12 '15 at 16:56
  • @Fina what timescale are you trying to predict for? If you want a daily or hourly predictive model you need wind fields. If you are just trying to do something seasonal or annual, then you could probably do something with as little as you describe... though I still don't understand what your end-product would be used for. – f.thorpe Sep 12 '15 at 19:50
  • Hi @Fina, OpenAQ is a non profit service providing lots of pollution data from all over the world through a single API. You should be able to cross reference their data with landsat. – Michał Karpacki Oct 14 '18 at 11:58

0 Answers0