You should do a little bit of homework before posting questions here... Much of the information you request is available from USGS, the home of the Landsat program:

http://landsat.usgs.gov/Landsat8_Using_Product.php

And if you search the Internet with the keywords "landsat 8 convert dn to radiance", you'll get 26,900 results, most of which are directly relevant to your question.

Also, neither NDVI nor SAVI are probably very appropriate to look at soil erosion, so you need to do some more research on your methods and approaches too!

You should do a little bit of homework before posting questions here... Much of the information you request is available from USGS, the home of the Landsat program:

http://landsat.usgs.gov/Landsat8_Using_Product.php

And if you search the Internet with the keywords "landsat 8 convert dn to radiance", you'll get 26,900 results, most of which are directly relevant to your question.

Also, neither NDVI nor SAVI are probably very appropriate to look at soil erosion, so you need to do some more research on your methods and approaches too!

You should used the thermal bands from Landsat 8 to calculate at-satellite brightness temperature in 2 steps: Convert the raw bands into Top of Atmosphere Radiance (TOAr). Convert TOAr into degrees kelvin.

Open Raster Function Editor from the toolbar. Right click on the raster and insert a Composite Band Function. Click on the plus sign and select Add copy of selected input. You have to do this because there are two thermal bands for Landsat 8.

Conversion from DN to RadianceThe Radiance Multiplier and Radiance Add are used to convert the DN back into radiance.

Right click on Raster and insert Band Arithmetic. Set the method to user defined.

Convert Radiance into degrees kelvinFrom TOAr, we now need to convert to temperature in degrees kelvin. This is where you use the K1 and K2 inputs. The equation we will be using functions to create is:

K2 / ln(k1/TOAr + 1)

The easiest way to do this is to create the (k1/TOAr + 1) part first using the Band Arithmetic function.

Good Luck

http://landsat.usgs.gov/Landsat8_Using_Product.php

Fully agree with Dr. Verstraete about NDVI and SAVI their purpose is slightly different. 

For red./refl. my guess is to use the reflectances and indices from ESPA site. Works fine for TM and ETM+.

I recommended ATCOR 2/3 for Landsat 8 imagery processing or ENVI 5.2 software. ATCOR is available for one-month trial from http://www.rese.ch/products/atcor/atcor3/index.html.

You can find more details about image processing in ATCOR 3 manual

ftp://ftp.dfd.dlr.de/put/richter/ATCOR/atcor3_manual.pdf

Best wishes

Both the NDVI and the SAVI require red and NIR reflectance data. Thermal infrared data are irrelevant for this question.