Reports

Contents

Title: Vaira NDVI Data
Date:2020-12-07
Data File: VR_NDVI_all.csv
Refers to:VR_met

Vaira NDVI 2000-01-01 through 2020-12-31

NDVI has been measured for many years at Vaira using different methods including tower based sensors, manual measurements and satellites. 

Name First Day Frequency Area Comment
MODIS_250m 2000-02-18 16 days 250m pixel NDVI product, mostly grass
MODIS_500m 2000-02-18 8 days 500m pixel Red and NIR surface reflectance, larger area
PAR-SW 2004-07-22 Daily Tower Broadband NDVI = [(SW-PA ) - PAR]/[(SW-PAR) + PAR]
LED 2004-07-24 Daily Tower Homebuilt sensor, reflectances correct as Ryu et al
Spec 2004-11-20 Site visit/seasonal multiple spots Ocean Optics USB2000 or Jaz hand held spectrometer
SRS 2017-08-18 Dailyq Tower Decago/METER Group SRS NSVI sensor
PL_Center 2018-01-01 Daily -gap filled 3m pixel Gap filled and corrected to Landsat and Sentinal
PL_All 2018-01-01 Daily -gap filled 729m  Average of 59049 3m pixles centered on the Tower
Arable 2019-05-01 Daily Tower Red and NIR from Arable Mark I/II

MODIS_500m, MODIS_250m: MODIS satellite data is available from 2000-02-18 to present.  Here we have two products: 16day 250m NDVI and NDVI calculated from 8day 500m Red and NIR reflectance. This data is available through Ameriflux and ORNL DAAC. 

https://modis.ornl.gov/cgi-bin/sites/site/?network=AMERIFLUX&network_siteid=US-Var&product=MOD13Q1

The MODIS 500m pixel:

MODIS_8day_500m.png

The MODIS 250m pixel:

MODIS_16day_250m.png

PAR-SW: Broadband NDVI can be calculated from PAR and SW-PAR (Huemmrich et al., 1999).  Here we are converting PAR to W/m2 with a conversion factor of 4.6 (µmol m-2 s-1) / (W m-2)

LED: Our homebuilt LED sensor has been on the tower from 2004-07-24 to present and is described here:

Ryu Y, Baldocchi DD, Verfaillie J, Ma S, Falk M, Ruiz-Mercado I, Hehn T, Sonnentag O (2010) Testing the performance of a novel spectral reflectance sensor, built with light emitting diodes (LEDs), to monitor ecosystem metabolism, structure and function. Agricultural and Forest Meteorology. 150, 1597-1606.

Spec: Handheld spectral measurements are made at most field visits with an Ocean Optics USB2000-VIS-NIR (and most recently the Jaz model) sensor.  NDVI is calculated from these spectra using a Red band of 620-670nm and a NIR band of 841-876nm.  The LED sensor was calibrated with this spectrometer so these measurements should match at least at the beginning. 

PL_Center, PL_All: Recently we received R, G, B and NIR imagery at 3m pixel resolution from Planet Labs.  This data has been corrected using Landsat and Sentinel data and gap fill to generate daily images.  An area 243x243 pixels was pulled to match the tower's foot print.  Both the average of this total area and the NDVI of the center pixel are presented here.

PlanetLabs_243x243.png

SRS: We also had a Decagon (now METER) SRS NDVI sensor on the tower for a few months 2017-09-07 through 2018-02-07 and now again since 2019-02-20.  The information about the previous inter-comparison can be found in the report here:

https://nature.berkeley.edu/biometlab/bmetdata/rtreport.php?screen=view&id=5

Arable: And finally, we have had a number of Arable Mark units on the tower since 2019-05-01.  The data here is from a first generation Mark I, a prototype Mark II and finally a production Mark II.

For all of the Tower based data (LED, PAR-SW, SRS, Arable) an average of 30min midday values (11:00 - 13:00) were used to calculated daily NDVI.

All of this data is presented in the first figure.

Figure 1. Vaira NDVI for all sensors from 2000-02-18 to 2020-12-31

Figure 2. Satelite data.  The two MODIS values and the Planet Labs NDVI averaged over a similar sized area match very well but this is no surprise as Planet Labs uses MODIS data to correct their reflectances. 

Figure 3. Ground based data.  The handheld spectrometer has some problems with low NDVI values especially the older USB2000 model that was used until mid 2019 when the Jaz model began to be used mostly.  The LED sensor was largely calibrated using the USB2000 spectrometer so their high NDVI values match well.  The early LED data from 2004-5 has some issues and should not be used without correction.  In the early years there seemed to be some drift in the LED data compared to PAR-SW especially in the low NDVI dry season but that seems to have stabilized in the last eight or nine years.  The SRS values are lowest overall and the Arable values are similarly low in the dry season but do tend to agree with the Planet Labs center pixel values in Figure 4.

 

Regression Data

Residuals

Figure 4. Scatter

Â