14 April 2012 - Unique applications of NDVI and GOES dryness products

Bob Rabin from NSSL/UW-CIMSS provided me with a couple interesting images late last week in preparation for the upcoming events on Friday/Saturday. I thought I would share them here with you all to get a new perspective on a product that we used extensively within the Fire Weather Experiment last Fall to monitor vegetation and surface moisture. Bob Rabin gathers and generates NDVI composite imagery as well as GOES derived surface dryness values that we provide within SPC operations in support of their fire weather forecast desk. The products were mainly intended to monitor for dry vegetation and anomalously dry surface conditions that would be a potential hazard for fires. In these examples, Bob pointed out that the NDVI and GOES surface dryness products were both picking up on a very well defined dry / moist boundary across much of KS, OK and TX (see images below).

Continental US NDVI composite from 9 April 2012. Green areas indicate regions with significant green vegetation cover, while yellow and brown regions indicate decreased green vegetation cover. Note the pronounced gradient in green vegetation along a line extending from western KS and OK into central TX.

Continental US GOES surface dryness composite from 11 April 2012. Green areas indicate regions with significant surface moisture mesurements, while yellow and red regions indicate dry surface measurements. Note the pronounced gradient in surface moisture along a line extending from western OK into central TX. It would be interesting to examine how these surface moisture boundaries interacted to potentially enhance convective initiation set up by circulations due to differential heating. In fact, one SPC forecaster noted this in a Mesoscale Discussion issued on the 14th prior to convective initiation along the TX/OK border. In addition, it would be interesting to examine the interaction these surface moisture boundaries have on the evolution of the dryline. It is exciting to see applications of satellite imagery used to detect these features not easily observable from other systems in such a constant and relatively highly spatial manner.

Relying on GOES observed dryness

GOES surface dryness product from 30 August 2011 with the PSA dryness grids overlaid from 31 August 2011. Areas in red and yellow indicate significant surface dryness, with areas in green indicating relative moist surface areas. We see that the PSA dryness grids drop off sharply at the OK/KS/CO border, which is not reflected in the GOES surface dryness product.
Today during our fire weather forecast we attempted to analyze the burnable fuel threat and the relative dryness over our forecast area covering much of TX, OK and KS. Apparently there has been a data problem with the operational PSA dryness grids over KS today which forced us to rely on the GOES observed surface dryness and NDVI products, which gave us a unique 'data denial' experiment to determine the availability of dry fuels (see image above). We see that GOES observed surface dryness is very high (reds) over much of KS, so we had to include this area within our burnable fuels threat today.


7-day NDVI composite (left) and 28-day NDVI change (right) from 29 August 2011. Areas of green indicate regions of increased (or increasing) vegetation, while areas of brown indicate regions of decreased (or decreasing) vegetation.
We also examined the NDVI and NDVI change composites in our analysis to determine the amount of vegetation available over these extremely dry areas. If we look at the NDVI composite, much of this area is shown as not containing a lot of green vegetation (leftmost image above). However, examining the experimental NDVI change product, we do see that most of this area is showing signs of 'greening' (rightmost image above). From this we determined that the amount of vegetation is slightly increasing, but still relatively dry and burnable (from our observations using the GOES dryness and NDVI composite products).

NDVI and GOES surface dryness comparisons

GOES observed surface dryness with the PSA dryness product overlaid for 29 August 2011. Areas of low dryness in green with areas of extremem dryness in red.
Today we examined the GOES surface dryness and dryness anomaly fields to help outline our fuel threat area and noticed that there were one significant mismatch between it and the official PSA dryness product provided within SPC operations (see image above). An area extending across much of western and central OR where we saw a good amount of rain over the past week is being described as extremely dry and extremely anomalous within the GOES dryness products, whereas the PSA dryness products have this area outlined as not a threat for dry fuels, which is what we were expecting. We are not sure why the GOES surface dryness products are not picking up on this, so we tended to put more faith on the PSA dryness product today over that area. otherwise, the dryness and dryness anomalies tend to match up fairly well with the PSA dryness values.

14-day NDVI change with PSA dryness product overlaid. Areas of green indicate areas of increased "greenness" in the NDVI change.
We did also compare the 7- and 14-day NDVI composite and NDVI change products with the official PSA dryness product and we were seeing similar features as the PSA dryness product, especially with this area over OR that shows up as increased "greenness" in the NDVI change (see image above). It suggests that a combination of the NDVI and surface dryness measurements might provide a more accurate analysis of the fuels, with a higher resolution to that of some of official dryness products currently provided within operations.

NDVI and satellite-based dryness observations versus operational data

PSADryness product for 22 August 2011. This product is routinely available within SPC operations. Areas of yellow and red indicate significantly dry surface measurements
SPC forecasters routinely use a product originally developed by the NWS Salt Lake City and the Eastern Great Basin Predictive Service Office called the Predictive Service Area Dryness (psadryness) product (see above) to help make their day 1-8 fire weather outlooks. This product provides the forecasters with an idea of the dryness of burnable fuels near the surface. In addition, the forecasters use a high-resolution 'land-use' product that attempts to simulate the NDVI product, but is rarely, if ever, updated. One of the goals we wanted to accomplish from this experiment is to see how the higher resolution datasets of observed NDVI and NDVI change (below), as well as the GOES surface dryness and dryness anomaly products (also below), compare to products currently in operations, such as the psadryness product (above).

14-day composite NDVI (top) and 28-day NDVI change (bottom) from 15 August 2011. Areas of green indicate regions where increased 'greenness' is observed.
During our first day, 5 SPC fire weather forecasters participated and examined these products to make an experimental "update" for their day-1 fire weather outlook, or out to 12 UTC the next day. In particular, forecasters were asked to make a forecast graphic depicting the areas of high threat for burnable fuels. When comparing the psadryness product and the satellite-based products, we noticed that there was a noticeable discrepancy over some areas, specifically over central ID (see above). While the psadryness product said that the area was extremely dry, the NDVI and NDVI change depicted areas of increasing 'greenness'. In addition, the GOES surface dryness and dryness anomaly products indicated no significant drying over the area.

GOES 14-day composite surface dryness (top) and 5-year average dryness anomaly (bottom) for 22 August 2011. Areas of yellow and red indicate increased surface dryness.
So what gives? Well we explained to the forecasters that the satellite-based products are limited to sensing the canopy of the location they are observing. this means that if there is any forest in the area, we cannot see the undergrowth, which could be dry and only measurable from surface instruments or observers. Unfortunately, this is a limitation we have to deal with, particularly from geostationary satellite-based instruments. However, the forecasters were impressed by the spatial resolution and relative rapid updates of the products, which is not provided to them from the psadryness or land-use products. It may be useful to combine these datasets to get a more detailed picture of what may actually be going on at the surface when it comes to burnable fuels.