link to publication

In 2018, I led a field crew in southwestern Kansas to examine the impacts of the Starbuck fire on Lesser Prairie-chickens. While monitoring nest success, I also developed an independent study surveying mortality and post-fire recovery of trees.

I was particularly curious if the fire suppressed colonization of invasive saltcedar.

In the false color image below, I set 2015 NAIP bands to the spectral peaks of trained data based on the Random Forest Model (RFM) results. Manually delineated saltcedar boundaries produced plots with species composition of 97 ± 2% saltcedar.

National Agriculture Imagery Program (NAIP) imagery (4 bands, June 2015)

Data

randomForest package used in R

1000 random sample points of the NAIP bands within saltcedar, trees, and grassland each

R: glm(y~x, data=df, family="gaussian")

Exploration: Which burn indices or presence/absence of fire at what distance from leks best predicts change in lek counts of lesser-prairie chicken pre- to postfire?

CORRECTED AKAIKE INFORMATION CRITERION (AICc )

Parameters
dNBR and MIRBI burn index (Figure 1) and present/absence (PA) fire values at distances 10-30,000 meters from each lek (n=18).

Leks had an average of 11.86 ± 7.95 SD birds per lek (116 total) before the fire (2015), and 5.44 ± 6.1 birds per lek (45) after the fire (2018).

AIC is corrected for the the very small sample size.
Very large spread in data.

Analysis: AICcmodavg package used in R

Mid Infrared Burn Index (MIRBI, left) from March 8, 2017, and differenced Normalized Burn Ratio (dNBR, right) from March 1 and 8, 2017, in Clark County, Kansas. Indices were calculated using Landsat 8 imagery accessed from https://earthexplorer.usgs.gov/.

10 Top Models

Generalized Linear Model (GLM)
for change in lek counts against
dNBR values at 50 ft from lek