Dear Usman,

What exactly do you mean typing "best NDVI range for Landsat?" As for average values, you can analyse homogenic subsets from the whole scenes.

THough, remember that Landsat 8 provide much lower temporal resolution compared with Modis

HI !

It is hard to say right off. You will need to look at your data availability in time and quality with respect to the grow season. The first idea would be to make minor adaptations the the work done with MODIS.

The approach to this question would be largely empirical. Thus you will need to develop plausible relationships and test them with regression models. If you have additional data such as rain, temperature, photosynthetic radiation, soil moisture, soil matric potential, etc, consider adding them to  the regression model. I have had success developing a linear regression of surface vegetative  biomass as a function of NDVI from LANDSAT.

As to crop yield possible approaches would be to look at the following:

1. The final crop yield as a function of change in NDVI over the growing season.

2. If you can get multiple NDVI values over the season try to estimate yield as a function the the rate of change in NDVI over the season.

3. Look at final crop yield as a function of final NDVI.

4. Look at final crop yield as a function of some threshold NDVI value at a time point in an earlier part of the season.

There are ways to estimate soil moisture from LANDSAT. Consider adding this to the regression in various ways.

Take Care

Bill

Dear Maria,

Yes i agree with you Landsat has low temporal resolution and it is big hurdle for calculating NDVI values at heading stage.For example in case of Pakistan,Wheat heading stage start from 25 February up to 15 march depending upon different areas and in case MODIS i calculate average NDVI values during this period.But this facility is not available for LANDSAT due to temporal resolution 16 days and sometimes if clouds are present then you easily imagine what will happen .

Dear Maria and Bill,

Could you please tell me is radiometric correction is necessary for LANDSAT 8 images .Actually i download images from Earth Explorer>Data Set>Landsat CDR>Landsat 8 OLI surface reflactance images.Whats your opinion about it

Surface reflectance is already atmospherically corrected data and hence don't need radiometric correction. Please read the corresponding documentation

Though in some regions there may be some inconsistencies so I prefer to use Level 1 products to make radiometric and atmospheric corrections myself

http://earthexplorer.usgs.gov/redirect/collectioninfo/?collection_id=8704

Dear Maria and Bill 

Could you please tell me how will i used parameters like soil type,temperature or soil moisture with NDVI to make linear regression model for crop yield.As Bill told me that only NDVI is not enough to make linear regression model between crop yield and NDVI

Dear Usman,

If your reference harvest data is parcel by parcel firstly you should calculate the parcel mean NDVI to compare parcel mean harvest for linear regression. However, I suggest you that use the NDWI, SAVI and Tasseled cup moisture data for multiple linear comparison. Because, NDVI only refer the chlrophyll content NDWI refer the plant weter content and moisture refers to soil surface wetness all of them is very inportant for plant growth.

Kind regards

Dear Onur,

I agree there could be used various approaches rather than NDVI. Although, NDVI doesn't refer only the chlorophyll content. And the tasseled cap moisture is not the only way to measure moisture of soil.

And by the way if you decide to use TCT wetness it would be strange if you do not use TCT Greenness.

Dear Usman,

I think you can start with NDVI incorporated with moisture conditions. May be it will be OK (this depend on crop character in the place you study and maybe something else, so you may try). And, if soil types in your study area are significantly different, you should take into account this too. And that would be fine if you could use your previous Modis data. Nevertheless remember that Modis NDVI and Landsat NDVI are not the same.

Anyway if you worked before with Modis data you may try to use Image analysis adjustable for multiple sensors (in this case Landsat and Modis) such as FAPAR. On the page 8 of pdf file listed below you can find the references to FAPAR products available for download (including data based on Modis Imagery).

Moreover FAPAR data as a rule is more related to physical plant characteristics than NDVI. You can also calculate mean FAPAR over the season.

On FAPAR web site (last reference) you can find a lot of interesting information including methods on Landsat FAPAR.

It's not as easy task but I think it might be helpful.

And finally I want to add one comment on weather conditions which of course greatly influence the productivity:

Ground meteorological station data is rarely suitable for such analysis if they are situated far away from the study area. So you may use simulated data (the first reference).

Taken from this source it usually work well.

Good luck!

Maria

http://www.cpc.ncep.noaa.gov/products/global_precip/html/wpage.cmap.html

http://fapar.jrc.ec.europa.eu/Home.php