Reports

Contents

Title: TZ adjustment to NDVI850_out
Date:2019-04-30
Data File: TZ_SRS_Raw_20190510.csv
Refers to:TZ,1066703767

WARNING: The conclusion of this report is still in question.  The abrupt change in reflected NIR may only appear abrupt due to the week of missing data.

There was a power outage on the Tonzi Tower from 2019-04-23 19:00 to 2019-04-30 11:30.  This power outage was either caused by the Licor LI-7550 box or damaged the LI-7550 box.  The box was returned to Licor for repair.  At the same time there was a drop in the data from the SRS NDVI_850out_all channel.  To detect and correct the change this channel was compared with the NR01_SWout channel on the tower.

Figure 1.  No obvious change in the incoming channels

Figure 2. An obvious change in the NDVI_850out_all as compared to the other outgoing channels.  Notice the change in the green vs pink lines before and after the gap.  The NR01_SWin has been scaled to make it easier to compare to.

Which channel to use to correct the NDVI_850out_all?  The NDVI_850out_grass sensor has a narrow field-of-view and the NDVI_850out sensor has a hemispheric field-of-view.  This causes the NDVI_850out_grass signal to be a little smaller it sees more shadows from the trees causing some excursions in the scatter plot.  NDVI_650out_all is a different wave band and does not over lap with the NDVI_850out_all wave band.  NR01_SWout has a similar field-of-view and even though its spectral range is much larger, it does overlap the the 850 band.

Regression Data

Residuals

Figure 3.  A relationship with the NR01_SWin before and after the power outage further shows the change and gives us a relationship that can be used to correct the affected data.

To make the post-gap data look like pre-gap data we need a multiplier so that pre = post * m.  If we substitute to slopes pre and post and solve for m:

m = 0.22923/0.19112 = 1.1994

Figure 4. Raw NDVI_850out and corrected NDVI_850out.

Conclusion: The multiplier of 1.1994 should be applied to the NDVI_850out_all channel for all times greater than 2019-04-30 11:30.