X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 27 Jul 2010 03:16 AM by  anon
Why the NDVI value is over -1 and 1 after atmospheric correction?
 3 Replies
Sort:
You are not authorized to post a reply.
Author Messages

anon



New Member


Posts:17
New Member


--
27 Jul 2010 03:16 AM
    I am try to calculate NDVI value after atmospheric correction in ENVI. The atmospheric correction process is as describe below. http://www.ittvis.com/Use...e/posts/Default.aspx If calculating NDVI by ENVI vegetation analysis module, the value range is within -1 and 1, however, if I calculate the NDVI by the band manth tools :(float(b4)-b3)/(float(b4)+b3), the histogram value shown that there are many value are far greater than 1 or below -1. Basic Stats Min Max Mean Stdev Band 1 -867.000000 1049.000000 0.151136 2.237053 Part of the histrogram is here: -108.113725 33 1382 0.0001 0.0035 -100.600000 403 1785 0.0010 0.0045 -93.086275 42 1827 0.0001 0.0046 -85.572549 158 1985 0.0004 0.0050 -78.058824 358 2343 0.0009 0.0060 -70.545098 148 2491 0.0004 0.0063 -63.031373 300 2791 0.0008 0.0071 -55.517647 272 3063 0.0007 0.0078 -48.003922 177 3240 0.0004 0.0082 -40.490196 720 3960 0.0018 0.0101 -32.976471 483 4443 0.0012 0.0113 -25.462745 728 5171 0.0018 0.0131 -17.949020 2951 8122 0.0075 0.0206 -10.435294 20165 28287 0.0512 0.0719 -2.921569 39297578 39325865 99.8458 99.9176 4.592157 23078 39348943 0.0586 99.9763 12.105882 5332 39354275 0.0135 99.9898 19.619608 1222 39355497 0.0031 99.9929 27.133333 371 39355868 0.0009 99.9939 34.647059 599 39356467 0.0015 99.9954 42.160784 108 39356575 0.0003 99.9957 49.674510 177 39356752 0.0004 99.9961 57.188235 41 39356793 0.0001 99.9962 64.701961 66 39356859 0.0002 99.9964 72.215686 373 39357232 0.0009 99.9973 79.729412 42 39357274 0.0001 99.9974 87.243137 27 39357301 0.0001 99.9975 94.756863 56 39357357 0.0001 99.9977 102.270588 5 39357362 0.0000 99.9977 109.784314 26 39357388 0.0001 99.9977 117.298039 18 39357406 0.0000 99.9978 124.811765 61 39357467 0.0002 99.9979 132.325490 24 39357491 0.0001 99.9980 139.839216 23 39357514 0.0001 99.9981 147.352941 5 39357519 0.0000 99.9981 I also write a IDL function to compute the NDVI, the results are also over -1 and 1. ;NDVI Function FUNCTION TM_NDVI,B3,B4,CHECK=check Den = float(B3)+B4 IF(KEYWORD_SET(check)) THEN $ Ptr = WHERE(Den EQ 0.,Count)$ ELSE Count=0 IF(Count GT 0) THEN Den[ptr] = 1.0 Result = (float(B4) - B3)/Den IF(Count GT 0) THEN Result[ptr]=0.0 RETURN, Result END

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    27 Jul 2010 07:24 AM
    The VI Calculator by default applies a biophysical mask to values that are unreasonable. It is likely that the atmospherically corrected data has some values that are outside a reasonable range (because some pixels are not modelled well) and thus return unreasonable NDVI values. These should be masked.

    Deleted User



    New Member


    Posts:17
    New Member


    --
    27 Jul 2010 06:48 PM
    I am very confusing about this. Because I want do a batch computation of NDVI by IDL program. I am also try the envi_dot function to calculate NDVI. the code is here. out_name = 'H:\Image\Output\Classfication\TempResults\Pre-Processing\Hainan_NDVI.img'; bandname=['TM band3','TM band4','TM NDVI'] ENVI_DOIT, 'NDVI_DOIT', FID=fid, POS=pos, DIMS=dims,/CHECK, $ OUT_DT=4,OUT_BNAME=bandname[2], OUT_NAME=out_name, R_FID=r_fid However, the results is same as the results of TM_NDVI function or (float(b4)-b3)/(float(b4)+b3) band math results, all of them are out of range. What should I do if I want to do a batch process? I think this problem will be existing in other vegetation indices calculation, especially for vegetation indices that not included in ENVI soft, which should be writen IDL program myself for calculation. It's necessary to do atmospheric correction before calculation of vegetation indices?

    MariM



    Veteran Member


    Posts:2396
    Veteran Member


    --
    28 Jul 2010 07:22 AM
    NDVI is a very simple equation. If your intial reflectance values are outside the range that is reasonable, then so will your NDVI values. While atmospheric correction is not necessary, if you are performing any sort of comparisons with other imagery or temporal analysis then it is highly recommended. You can convert your atmospherically corrected data to floating point reflectance that should range from 0-1. Any pixels outside of this range should be masked because they were not modelled well.
    You are not authorized to post a reply.

    Forums Products ENVI