1. K-State home
  2. »KSCFWRU
  3. »research
  4. »projects
  5. »Patterns of Greenness (NDVI) in the Southern Great Plains and Their Influence on the Habitat Quality and Reproduction of a Declining Prairie Grouse

Kansas Cooperative Fish and Wildlife Research Unit

Patterns of Greenness (NDVI) in the Southern Great Plains and Their Influence on the Habitat Quality and Reproduction of a Declining Prairie Grouse

Investigators:
Ashley Messier, M.S. Student

Project Supervisors:
Dr. Dan Sullins
Dr. David Haukos

Cooperators
Kansas Ecological Services Office
U.S. Fish and Wildlife Service

Funding:
U.S. Fish and Wildlife Service

Start Date: 1 January 2021 (30 months)

End Date: 30 June 2023

Introduction/Background
Grassland birds are declining more steeply than any other avian guild and we have known this for over 20 years (Peterjohn and Sauer 1999, Rosenberg et al. 2019). Declines have accrued over the past 4 decades despite limited contemporary conversion of grassland to row-crop agriculture in many areas (Cunfer 2005, Spencer et al. 2017). Many remaining grasslands have been degraded through the alteration of natural ecological drivers including the reduction and removal of fire from landscapes and shifts to unnatural grazing regimes (Askins et al. 2007). In part, the consequences of the continued decline in total grassland area and altered ecological drivers is reflected in an increasing number of grassland wildlife petitioned for listing under the US Endangered Species Act including pollinators such as the regal fritillary (Speyeria idalia), and monarch butterfly (Danaus plexippus plexippus; see U.S. Fish and Wildlife Service’s National Listing Workplan).  Furthermore, many grassland nesting migratory birds continue to decline, including the grasshopper sparrow (Ammodramus savannarum), Sprague’s pipit (Anthus spragueii), upland sandpiper (Bartramia longicauda), lark bunting (Calamospiza melanocorys) and chestnut-collared longspur (Calcarius ornatus). 

To better inform conservation efforts targeted at reversing grassland bird declines and ensure self-sustaining populations of at-risk species such as the lesser prairie-chicken (Tympanuchus pallidicinctus), a better understanding of broad-scale habitat availability and quality for grassland birds is needed. Unfortunately, our potential to effectively monitor the quality of remaining grassland habitat at relevant spatial scales is limited because >90% of the Great Plains is privately owned (Becerra et al. 2017). Furthermore, subtle changes in vegetation structure (i.e., 1 dm change in grass height) are not easily detected but may be exponentially important for the persistence of ground-nesting birds (Lautenbach et al. 2019).

Recent improvements in remote sensing technology and analytics make monitoring grassland bird habitat feasible over broad scales with high resolution. Sensors from aerial and satellite imagery can now collect data at pixel sizes <1 m in resolution in the visible light spectrum, <2 m for multispectral data, and at a temporal frequency as often as 1/day (Table 1). One particularly useful tool that can be derived from imagery is the Normalized Difference in Vegetation Index (NDVI) which estimates greenness (index of productivity). Vegetation greenness has been a strong predictor of habitat availability for several species of wildlife and may be a particularly useful tool in grasslands (Pettorelli et al. 2011, Hoagland et al. 2018). Ecological disturbances in grasslands have been associated with temporal variation in greenness (Goodin and Henebry 1997, Hutchinson et al. 2015). A common issue in NDVI analyses are acquiring images that lack cloud cover or other atmospheric interference. However, data collection at near daily frequencies makes acquiring quality data at weekly intervals possible and analytical improvements now allow for atmospheric corrections and the use of images from multiple sensors (van Leeuwen et al. 2006, Albarakat and Lakshmi 2019).  The utility of NDVI is that it can be estimated from any multi-spectral image whether collected from a camera on an aircraft or from a satellite.

Temporal patterns in greenness values that describe site-specific plant phenology could also be useful for evaluating habitat quality in grasslands (Appendix Table 1; available upon request). Several grassland bird species, including lesser prairie-chickens, need both residual grassland cover and disturbed areas to successfully reproduce (Hagen et al. 2013). Combining phenology-based metrics with field-collected data could be extremely useful for identifying high quality reproductive habitat and monitoring habitat using strategic habitat conservation in the future.

We propose to evaluate the utility of NDVI and phenology-based metrics in estimating lesser prairie-chicken reproductive habitat quality remotely over broad spatial scales.  Upon completion of research, we will provide data, information and technical assistance to the U.S Fish and Wildlife Service (Service) and conservation community through communication and coordination of the research findings and their utility to Service work.

Utility
Our proposed research is directly related to advancing the principles of Strategic Habitat Conservation, providing information that supports reversing the declining trend of grassland nesting migratory birds, and ensuring the self-sustaining populations of an at-risk grassland species.

Monitoring is a critical component of Strategic Habitat Conservation and our research seeks to advance monitoring functionality of grasslands over broad spatial scales. If successful at estimating habitat quality remotely, targeted conservation efforts can be implemented using spatially explicit habitat map products. Habitat products will inform conservation design and delivery for grassland initiatives focused on species currently protected under the Endangered Species Act, Migratory Bird Treaty Act, and other declining grassland species in need of conservation.

The proposed research will also advance our understanding of habitat availability for the lesser prairie-chicken. It is well known that lesser prairie-chickens need large grassland landscapes; however, researchers have struggled to quantify habitat quality within remaining grasslands (Sullins et al. 2019). Our proposed research would advance our understanding of quality grassland habitat availability for lesser prairie-chickens, and associated grassland nesting migratory birds.

Description of Analyses/Service
Principal Investigators and Masters of Science student will conduct original applied research that will evaluate the use of remotely sensed NDVI (Normalized Difference Vegetation Index) and NDVI phenology metrics to monitor high quality grasslands used by lesser prairie-chickens, and specifically addressing the following objectives:

Objective 1. Evaluate the influence of grazing, fire, and precipitation on NDVI-related metrics in the Central and Southern Great Plains.

Objective 2. Identify NDVI-related metrics that best distinguish lesser prairie-chicken reproductive habitat from non-habitat.

Objective 3. Relate NDVI variation within season and among breeding seasons to the timing of nest incubation among successful and unsuccessful lesser prairie-chicken nests.

Methods, Protocols and/or Scientific Standards
The proposed research will leverage field data previously collected at study sites in Kansas, Colorado, New Mexico, and Texas. The study areas within these states encompass the four ecoregions within the occupied range of the lesser prairie-chicken: Mixed-Grass Prairie, Sand Sagebrush Prairie, Shinnery Oak Prairie, and Short-Grass Prairie/CRP Mosaic. Over 800 lesser prairie-chickens have been captured and marked with VHF or GPS transmitters from 1989-2018, including nest and brood location data from >300 lesser prairie-chickens in Kansas and Colorado. These latter data will largely be used to evaluate fine-scale NDVI relationships with reproductive habitat. Data from Texas and New Mexico will be combined with the Kansas and Colorado locations to address nest phenology questions over a large latitudinal gradient.

Remotely Sensed Data Collection

For pursuit of all objectives, we will compile and organize remotely sensed data to estimate NDVI and NDVI-related metrics at multiple scales. We will use a variety of freely available data from MODIS, NAIP, and LANDSAT sensors as well as commercial products including data from Worldview or GeoEye, which are available at finer resolutions (Table 1).

Table 1. Remotely sensed multispectral data, data resolution, frequency, and availability for use in evaluating patterns of greenness.

Sensor

Multispectral Resolutions

Temporal Frequency

Cost Y/N

MODIS

250 m

Daily

N

NAIP

1 m

Typically once a year

N

LANDSAT

30 m

Every 16 days

N

WorldView

1.24 m multispectral resolution

Daily

Y

GeoEye

1.65 m multispectral resolution

Every 3 days

Y

 

Specific tasks include:
Objective 1: Evaluate the influence of grazing, fire, and precipitation on NDVI-related metrics.

We will examine linear and quadratic relationships of NDVI, NDVI heterogeneity, and phenology related metrics with grazing, fire, and drought data that have been collected at >20,000 locations in the study area including used sites, nest sites, brood locations, and random site locations, which includes lower quality grasslands. At each site, grass height, visual obstruction, litter depth, and percent cover of grass, forbs, litter, and bare ground were collected. We also conducted 250-m point-step transects in the Kansas and Colorado portion of the study area following Evans and Love (1957). 

Vegetation data were collected among a range of weather and drought severity. The effects of fire on NDVI will be derived using vegetation data from south-central Kansas where we have before and after data following a large, intense wildfire in 2017 and where patch-burn grazing was implemented. To evaluate the influence of grazing on NDVI we will subset data from 3 large (>5,000 ha) ranches where we have cattle stocking rate data. We will use generalized linear models to examine relationships between disturbances and NDVI metrics using a multimodel inference approach (Burnham and Anderson 2002). We plan to provide predictive habitat monitoring tools that will link interactions of grazing, fire, and precipitation with remotely sensed NDVI, and on-the-ground vegetative characteristics. Python and/or R script used to predict vegetation cover will be provided to the Service and other stakeholders after peer review.

Objective 2: Identifying quality reproductive habitat based on NDVI.

We will evaluate relationships of NDVI with nesting and brood-rearing habitat based on locations of >300 nest sites and corresponding successful broods throughout the study area. Mean values for NDVI, variability of NDVI, and other available metrics that include time-integrated NDVI and amplitude will be estimated at nesting and brood rearing sites. Values from reproductive locations will then be compared to values at random locations using resource selection functions (Boyce et al. 2002). 

We expect that moderate values for time-integrated NDVI and amplitude will best predict reproductive habitat for both nesting and brooding lesser prairie-chickens. A rapid pulse in greenness as expressed in amplitude, or a high value of time integrated NDVI, would likely indicate little available residual grass cover in the grassland, which is needed as nesting substrate (Lautenbach et al. 2019). In contrast, a minimal level of amplitude might indicate that either the area was not grazed and is too dense and thick for use by prairie-chicken broods. We predict that quality of nest and brood habitat will be linearly related with heterogeneity of NDVI values at a pasture scale (~3 km2). Heterogeneity is important for ground-nesting precocial birds that need areas of dense residual grass cover for nesting adjacent to more disturbed grasslands for brooding individuals (Dreitz 2009, Hovick et al. 2015). Heterogeneity of NDVI will be estimated using the coefficient of variation with a curvature metric that simultaneously accounts for variability and interspersion of NDVI.

Relationships of NDVI metrics with reproductive habitat will be facilitated among multiple spatial scales. We will first estimate mean NDVI, heterogeneity of NDVI, and phenology values for all nest sites and brood sites. Mean values will be estimated among multiple scales using focal statistics (moving window analysis) in ArcGIS 10.8. The finest scale evaluated (4-m radius) will be directly related to field-based vegetation measures of grass height, visual obstruction, percent covered by plant functional groups, and litter depth. We expect that results will establish an effective method to remotely identify high quality reproductive habitat over large spatial extents. Results will directly aid in spatially targeting conservation for lesser prairie-chickens, and for other grassland species in informing biological planning and conservation design (e.g., listed species Recovery Plans, National Wildlife Refuge Comprehensive Conservation Plans, Species Status Assessments). Targeted conservation efforts could benefit a suite of grassland-dependent wildlife (Pavlacky et al. 2018).

Objective 3: Relate NDVI variation within and among seasons to the timing of nest incubation among successful and unsuccessful nests.

Nest initiation date can strongly predict nest survival and brood survival for many avian species including lesser prairie-chickens. Birds that begin nesting earlier are typically more successful at raising a brood (Lautenbach et al. 2019). We seek to examine relationships among plant phenology and reproductive rates. To do so, we will use NDVI and phenology-based metrics collected throughout the study area including parts of Kansas, Colorado, Texas, and New Mexico. We predict that timing of greenness will be associated with optimal nest initiation timing. Optimal greenness patterns for successful nests will vary latitudinally and among wet, average, and dry years. To analyze greenness patterns, we will use generalized linear mixed models and reproductive phenology data from marked lesser prairie-chickens (n > 300) captured and monitored from the 1990s to 2018. We will use an information theoretic approach to identify competing models that may best explain reproductive success including nest initiation timing, greenness, weather values, and vegetation-based metrics (Burnham and Anderson 2002). Results from this objective will assist the Service and other conservation efforts in understanding the phenological qualities that affect prairie-grouse habitat and reproduction across a latitudinal gradient and may enable monitoring effects of future global change.