# [R-sig-Geo] Landsat soil moisture index

Sarah Goslee sarah.goslee at gmail.com
Fri Nov 8 22:30:15 CET 2013

```Hi,

Well, you need to start at the beginning.
Where did you get your Landsat images?
What format are they in?
Has someone already calculated NDVI and surface temperature for you,
or do you need to do that? If so, what level of processing have the
How have you imported the images into R?

Do you already know how to perform a linear regression in R?

It's not a difficult task programmatically, but there are a number of
steps that you need to consider, and things you must understand before
you can get good results. Remote sensing isn't necessarily an easy
field.

Sarah

On Fri, Nov 8, 2013 at 8:52 AM, Allen McIlwee <amcilwee at westnet.com.au> wrote:
> Dear list
>
>
>
> WANG et al. 2009 "Satellite remote sensing applications for surface soil
> moisture monitoring" derived a Landsat soil moisture index:
>
>
>
> SMI = (Tsmax - Ts) / (Tsmax - Tsmin)
>
>
>
> where Tsmax, Tsmin are the maximum and minimum surface temperatures for a
> given NDVI value, and Ts is the surface temperature at a given pixel for a
> given NDVI. Values near 1 (low surface temp & low greenness) have the
> highest estimated soil moisture, while values near 0 (high surface temp &
> greenness) have the lowest.
>
>
>
> Could anyone please advise me how I can create an image matrix that will
> extract the min and max surface temperatures for each unique NDVI value. I
> then need to apply a linear regression to the max temp vs NDVI & min temp vs
> NDVI scatter plots, and use the corresponding slopes and intercepts to
> calculate Tsmax and Tsmin
>
>
>
> Tsmax = (a1 x NDVI) + b1
>
> TSmin = (a2 x NDVI) + b2
>
>
>
> Hoping this is not such a difficult task for someone mathematically inclined
>
>
>
> Thanks
>
> Allen
>
>

--
Sarah Goslee
http://www.functionaldiversity.org

```