Kenneth,

Yes, if you have them boots on the ground trumps theory and is the way to go when available.  Of course, things that have to do with plume size, concentration levels, transport direction and pattern, and frequency are things more difficult to do by just looking out the window.  As we know remote sensing will never replace field work but it can help focus and optimize the field work (e.g., it can tell us where and when to go out) and add to any analyses of the impact of fires and the smoke emitted by those fires.

I have heard on the news as to how bad things are in your neck of the woods; good luck and I hope the fires settle down soon.  The smoke plumes are quite extensive when looked at on the MODIS images.

Pat

The Agricultural residue burning is during specific harvesting periods of certain crops like Rice and wheat. so they are in typical agricultural landscape as compared to forest fires. The simple land use map of study areas can also help to identify whether the burning is in  agricultural landscape or fires in the forest. 

MODIS Active Fire Product (MOD/MYD14),a regionally-tuned 8-day cropland differenced Normalized Burn Ratio product can also be used to differentiate between forest fires and biomass burning in agricultural landscape.

Long time-series NDVI/LAI products would be helpful to identify the burning crop residue or forest fires.

I suppose author want separate agro and forest when they are neighbors inside satellite pixel footprint. That is some problem - but forest fires are longterm and agro are shortterm. Try this. I have no any idea on physical reason to separate. Use time duaration of fire

Kenneth - as approvement of burned area - we may use more accurate resolution - SPOT, LANDSAT,SENTINEL, but data from satellite may be delayd few days or weeks... and over agriculture fields you will find new grass ))) and no burns

Sorry - we here in Russia work with Siberia wild fires with nobody (no humanbodies 1000 km around) How you imagine visit it? It is not joke. Look Googlearth over Yakutia region. If we speak about urban territory - what we may call "wildfire"  - all fires are counted and fire service work to fix?  )))  Another - agriculture... one fall a sigaret and dry grass fires immediately. but does fire go to forest - by coarse resolution of 1 km  (Modis, NPP, Avhrr) we do not know.

Are you attempting to detect and map the aerosols / haze using satellite image data or trying to identify what cover type is actively burning or the cover type that burnt after the fact?  This will help focus on the type of image data and method to help with your application.

Pat

Hi

You can use multi temporal data set . also use hyper spectral data.its provide a broad range of object spectral. Envi software have a section with name spectral that present vegetation analysis its might help you.

One simple visual inspection of the shape of the burned areas can help: agricultural areas tend to have rectangular shapes whereas wild fires have irregular shapes.

Land cover map will be helpful to define origin of the burnig fires. Global land cover maps are available from the online or offline organizations. For example, http://landcover.usgs.gov/globallandcover.php

Dear Kenneth! the most of forest fires in Russia also are "manual". But at some spring/automn time (in our region) fires come to forest from grass (agriculture) fires.

I would still like to know if the interest is in classifying the smoke type of active fires by remote sensing 'smoke information' or determine after the fact the type of land cover than burnt and generated the smoke.  If the application is not identifying the type of smoke directly then the use of multi-temporal Landsat TM type image data could be used to not only extract spectral information but also spatial information. 

Local to moderate spatial variability information can be used to help identify the type of vegetation that burnt.  Also  topographic data (DEMs from SRTM) can be used to easily identify if the fires are in relatively flat or in non-flat areas; if non-flat terrain then more than likely this would mean it is a forested not ag area that burnt.

My suggestion would be to use the SRTM as a first cut to classify the burnt area into flat vs non-flat terrain; if non-flat then it is probably a forest area that burnt.  If it is relatively flat terrain then continue the analyses by using Landsat TM type data to help determine if ag, grass, or forest area that is in relatively flat terrain burnt.  Spatial variability analyses can be used to help determine if forest area or ag / grass areas burnt because the local to moderate textural information will allow the cover types to be separated.

Pat

Kenneth,

Yes, there are flat areas within non-flat areas.  However, if the variance in slope and/or aspect within a given moderate to regional 'pixel size window' (say 51 to 501 pixel window) indicates something other than flat then a good first cut is to say the fire is not in an ag fields and perhaps not in a grass only type landscape.

As you know there are various options as to how to attempt to do some 1st level automatic classifications of landscapes and terrain and using DEMs is just one possible option.  After the 1st level classification (based on topography) bring in multi-spectral (such as Landsat) and multi-temporal data into the analyses.

Pat

Kenneth,

Yes, if you have them boots on the ground trumps theory and is the way to go when available.  Of course, things that have to do with plume size, concentration levels, transport direction and pattern, and frequency are things more difficult to do by just looking out the window.  As we know remote sensing will never replace field work but it can help focus and optimize the field work (e.g., it can tell us where and when to go out) and add to any analyses of the impact of fires and the smoke emitted by those fires.

I have heard on the news as to how bad things are in your neck of the woods; good luck and I hope the fires settle down soon.  The smoke plumes are quite extensive when looked at on the MODIS images.

Pat

 Any chemist can describe the difference in chemistry of smoke produced. If chemistry is different then the respective spectral reflectance should be different.

Dear colleagues!

just yesterday we finish Russian forest fire season 2016!

What can I say - to simpliphy  remote sensing usage - we use coarse (1km) resolution to operative fire monitoring to assist forest fire service. And later use Landsat, Sentinel, KMSS  data for burnt area qualification. If one need a CO2 calculation  - just use area and fuel map (accumulated plants  without fires )