# Understanding why MSAVI II values are outside the theoretical range

I am trying to understand why certain vegetation indices do not fall into the -1, 1 theoretical range.

I generated Modified Soil Adjusted Vegetation Index II values in Google Earth Engine using surface reflectance images. My MSAVI II values are all below 1, but can drop down to -7 (-288 in one extreme case). When I calculate the zonal statistics for a given polygon in QGIS 2.18, the extreme negative values impact the mean.

Based on a response by Jeffery Evans (Soil Adjusted Vegetation Index Values (MSAVI2) are Not Within Expected Range), I understand that MSAVI II is one of the slope corrected indices that will have values greater than 1 or less than -1; however, I cannot find information with a full explanation even after reading the two cited articles.

Could anyone provide documentation that explains this phenomenon, or suggestions on how to correct this issue? If negative values are theoretically acceptable, do I just have to ignore questionable means and go with the median?

• `I am trying to understand why certain vegetation indices do not fall into the -1, 1 theoretical range` that's because this range is valid just for normalized difference indices (like NDVI, NDSI, NDWI, and so on), other indices don't the same expected behavior. For example REP index has a common range of 690-750 approx. – aldo_tapia May 10 at 15:33
• And for the zonal statistics issue, I would go for median value instead of mean. Mean is always tricky as is affected by outliers – aldo_tapia May 10 at 15:36
• Thank you @aldo_tapia! Is it normal for the indices to remain below 1 (around 0.9), but then drop to -7? I generated a three month average over 15 years and the ranges tend to always be right below 1, but the negative values vary dramatically over the years. – Sarah May 10 at 18:30
• are you using surface reflectance products in GEE? If yes, even NDVI could retrieve weird values, especially in water bodies – aldo_tapia May 10 at 19:37