Assessing Rangeland Forage Productivity & Quality

The Use of Remote Sensing to Inform Collaboratively Managed Landscapes on Colorado's Northern Front Range

Elin Binck, Matthew Collins, Anna Clare Monlezun, Rae Nickerson

December 3, 2020

Introduction

Grazing Systems

For millennia the ecology of western grasslands and rangelands, characterized by high quality forage and profuse biodiversity, has supported large herds of herbivores, and in turn herbivory has contributed to the maintenance of those ecosystems over time. However, a debate over the impacts of cattle grazing on government-owned rangelands is ongoing at local, national, and global levels.

Over 30% of land in Colorado is owned and managed by the federal government, 95% of which is used for domestic livestock grazing and only 10% of which is classified as “protected”. Local governments along the Northern Front Range are known for their large investments in protecting remaining open space to offset conversion to urban development and maintain natural working lands for local agriculture and recreation.

Cattle are the top agricultural commodity in Colorado, and evaluating their role in rangeland ecosystem sustainability is crucial. Forage productivity and nutritive quality in rangeland ecosystems is of particular interest to livestock producers and wildlife managers alike. Understanding the potential of working lands to provide wildlife habitat and resilient ecosystems while also providing quality forage for domestic livestock is essential for both ranching livelihoods and natural resource conservation.

Traditional research methods to assess these factors, such as clipping and weighing (productivity) and chemical lab analyses (forage quality), are extremely laborious, and the timing of results may not align with the timing of important decisions such as those involving stocking rates or pasture rotation schedules. Therefore, our study utilized remote sensing technology and ArcGIS Pro geospatial analysis tools to evaluate Net Primary Productivity (NPP) and Normalized Difference Vegetation Index (NDVI) as alternatives to on-the-ground methods. We used these tools to investigate the following question on Colorado's Northern Front Range.

Research Question

Are there differences in spatial patterns of (a) forage productivity and (b) forage nutritive quality, among select government-owned lands that have been historically managed by grazing compared to areas excluded from grazing?

Study Region

The study was conducted in the Northern Front Range region of the Colorado Rocky Mountains, USA.

GIS Concepts

Variable (a):

Net Primary Productivity (NPP)

NPP is an important aspect of the global carbon budget and a useful indicator of ecosystem function. NPP is a component of Primary Productivity, which is the rate at which organic biomass accumulates within plants. There are two types of Primary Productivity: Gross Primary Productivity (GPP), the rate of solar capture by photosynthesis, and NPP, the remaining fraction of biomass produced after accounting for energy lost due to cellular respiration and plant tissue maintenance. Thus, NPP may be calculated as:

NPP = GPP - Respiration

NPP can be assessed by measuring plant characteristics or harvesting plant materials. Yet, these practices are both difficult and costly over large areas. At large spatial scales, remotely sensed images can be used to estimate NPP.

In this context, NPP is often calculated as a product of fPAR (fraction of photosynthetically active radiation) and light use efficiency (otherwise known as radiation use efficiency). Then, a model informed by either linear modeling of reflectance data, or physical modelling of light energy absorption/reflection is used to produce maps of vegetation biomass. (Image reference: Hong et al., 2020)

NPP estimates are particularly useful for assessing ecosystem function, estimating crop yields and stocking rates, monitoring changes in productivity over time, and monitoring vegetation health. As summarized in recent scientific literature, NPP derived from remote sensing promises a reliable and economically feasible measure of rangeland vigor and growth capacity. For this study we specifically used NPP as a measure of forage productivity.

Variable (b):

Normalized Difference Vegetation Index (NDVI)

NDVI is a graphical indicator of reflected or re-emitted radiation used to assess vegetation presence and/or health. NDVI data is collected by satellite sensors as the level of radiation reflected back from the earth's surface. The light-spectrum of radiation reflected back to the satellite sensor depends on the presence and/or health of the vegetation below. When chlorophyll is present, the amount of light absorption in the near-infrared region is limited. When chlorophyll is not present, for example in more woody plants, dead plants or unhealthy plants, cellular structure dictates light absorption in the near-infrared region.

NDVI is calculated from individual pixel characteristics of near infra-red (NIR) and red (R) bands. In ArcGIS, these wavelengths are characterized by Band-4, and Band-3 respectively. Thus, the equation for calculating NDVI is:

NDVI = (NIR - red) / (NIR + red)

The resulting index is expressed as a value between -1 and 1. Within this range, -1.0 - 0 represents clouds, snow, water, etc., 0 - 0.1 represents bare soil and rock, and 0.2 - 1.0 represents vegetation of various greenness.

NDVI demonstrates well-supported accuracy in the scientific literature and provides an efficient means to assess vegetation quality and land cover at diverse spatiotemporal scales. Applications of NDVI are abundant and include global vegetation trends, grazing selectivity, wildlife migration, wildfire ecology, plant vigor, and livestock stocking rates. For this study we used NDVI as a proxy for forage nutritive quality.

Data Schema

This schematic illustrates our hypothesis: the relationship between environmental and management variables and our response variables, NPP and NVDI.

Site History & Map

Soapstone Prairie Natural Area was acquired by the City of Fort Collins Natural Areas Department in 2004 and has been managed collaboratively with a local grazing association since then. The same grazing association had also been leasing the land from the previous owners, prior to the area's acquisition by the City of Fort Collins.
Coyote Ridge Natural Area was obtained by the City of Fort Collins Natural Areas Department in 2017 and has been collaboratively managed with a private rancher/producer since 2019. Prior to 2017, this area was part of a private cattle and wheat ranch, homesteaded in 1959. The ungrazed area is officially a portion of Larimer County's Rimrock Open Space, which has been excluded from grazing for over a decade.
Coalton Trail Open Space is a large contiguous property that was acquired by Boulder County Parks and Open Space in small parcels over time. Due to its vicinity to highly populated urban areas, Coalton Trail is a popular recreation site for hiking and biking. It has been managed collaboratively with the same local rancher/producer since 1995.
Lowry Ranch is a property of the Colorado State Land Board acquired in three separate transactions in 1964, 1966, and 1991. It has a long diverse history as an army airfield and experimental bombing range, which was then converted back into ranch land and managed exclusively by a continuous grazing regime until 2007, when cattle use was discontinued for the next 7 years. In 2014, with a new sustainability initiative, Lowry's management transitioned to a holistic, rotational grazing framework. Although Lowry Ranch is a government-owned property, whose revenue supports public interests like education, it is not open to the public for recreation.

"When we try to pick out anything by itself, we find it hitched to everything else in the universe." John Muir

GIS Analyses

01 / 08

Data Sources for Variable (a): NPP

NPP data was accessed from Google Earth Engine , which provides access to free, downloadable remote sensing data. The data used for this project was 30x30m resolution Landsat NPP CONUS which uses Landsat Surface Reflectance. NPP data for each year across a 10 year time period (2009-2019) was entered into analysis.

Data Sources for Variable (b): NDVI

NDVI data was downloaded from Climate Engine , a website that offers free, on-demand remote sensing data powered by Google Earth Engine. NDVI tiff files were obtained using a concatenation of 30x30m resolution Landsat 4, 5, 7, and 8 Surface Reflectance data. The means of data collected between May 21 and June 21 for a 10 year period (2009 to 2019) were calculated in Climate Engine and subsequently downloaded for analyses in ArcGIS Pro.

Trace Clip and Clip Overlying Polygon Tools

In order to delineate Areas of Interest (AOI) for spatial analyses in Soapstone Prairie Natural Area, Coyote Ridge Natural Area, Coalton Trail Open-space, and Lowry Ranch, we added a Colorado Ownership, Management, and Protection (COMaP) data layer and used the trace feature of the Clip tool to manually clip the government lands boundaries to the (AOIs) grazed areas and ungrazed exclosures (Figure 1). The clip overlying polygon feature of the Clip tool was then used to remove the overlying ungrazed exclosures from the grazed AOI’s.

Figure 1. Coyote Ridge Natural Area grazed area boundary and exclosure boundary

Union Tool

Next, soil data from the USDA-NRCS Soil Survey Geographic (SSURGO) Database were imported for Boulder, Arapaho, and Larimer Counties. The three distinct layers were merged using the Union tool and reclassified using the Reclassify tool to create consistent syntax across counties. We then used a Select By Attribute process to select only clay-loam and sandy-loam soil units within the AOIs to ensure consistency among soil types and reduce biological variability for analysis (Table 1). We then created site-specific AOI grazed layers (AOI_Grazed_Select_Soils) and exclosure layers (AOI_Ungrazed_Select_Soils)(Figure 2).

Table 2. Soil units used to define grazed and ungrazed areas of interest. Soil units were selected based on consistency in primarily loam soil textures.

Figure 2. Selected soil units clipped to the Soapstone grazed area AOI.

Raster Calculator

To further reduce confounding environmental variables, we grouped years according to precipitation. This created 2 classes: above average (wet) and below average (dry) precipitation years (Table 2). This was determined by Arapaho, Larimer, and Boulder County rainfall averages retrieved from the Colorado Climate Center (Table 3). The Raster Calculator was used to calculate mean NPP (Landsat Net Primary Production CONUS) and NDVI (Landsat 4/5/7/8 Surface Reflectance) for above average and below average precipitation year classes. Further, the Raster Calculator tool was used to rescale NPP values from the ArcGIS default unit to its reportable unit: KgC/m^2

Table 2. Summary of above average (wet) and below average (dry) classes by county for each year (2010-2019)

Table 3. County-specific precipitation data for Boulder, Larimer, and Arapaho Counties

Batch Raster Clip Tool

The Batch Raster Clip tool is useful for clipping large numbers of raster layers to the boundaries of a singular shape-file. We used the Batch Raster Clip tool to clip NPP and NDVI raster layers for above average (wet) and below average (dry) precipitation year classes to each AOI (Figure 3). We implemented several conventions in the Batch Clip tool: the output extent was set to each respective AOI, the no data value was set to -999, 32 bit float was selected, and boxes were checked for both "input features for clipping geometry" and "maintain clipping extent."

Figure 3. The batch clip analyses resulted in 4 layers for each AOI, including grazed-wet, grazed-dry, ungrazed-wet, and ungrazed-dry layers. This image shows NDVI for grazed areas of Lowry Ranch during wet years.

Raster to Multipoint Tool

In order to create a sample point layer for statistical analyses, an example grazed and ungrazed raster for each AOI were entered in the Raster to Multipoint tool. To be consistent, we used the NDVI wet class raster for each AOI. This tool created a layer with a point for every pixel centroid that could potentially be included in the final sample set.

Create Random Points Tool

Our statistical analyses required that NPP and NDVI values were derived from a sample of data points on grazed and ungrazed rasters. In order to sample points across grazed and ungrazed AOIs while minimizing potential for spatial autocorrelation, we used the Create Random Points tool. We sampled 50 points from each grazed AOI and 10 points from each exclosure to account for the smaller spatial extent of the ungrazed exclosures. We constrained the tool to select points with a minimum distance of 25 m between any pair of points.

Figure 5. Random sampled points in Coalton Trailhead Open Space grazed area and ungrazed exclosure

Extract Multi Values to Point Tool

To attribute NPP and NDVI values of each AOI raster to corresponding points created by the Create Random Points tool, we used the Extract Multi-Value to Points tool. For this analysis, the input point features were “%SiteName%_Grazed_25_50RandPts” and “%SiteName%_Ungrazed_25_10RandPts.” Input rasters included all of the associated above average and below average precipitation class NPP and NDVI rasters. Output field names were abbreviated for column headings. These names and associated values were added to attribute tables of input point feature layers (Figure 6). The 8 resulting tables were combined using the Merge tool and exported to a .csv file for statistical analysis in R.

Figure 6. Attribute table indicating the NPP and NDVI values in "wet" and "dry" years corresponding with distinct points (OID) determined by the Create Random Points Tool.

Naming Conventions

Flowchart

Visualizing the analysis process.

Let's take a closer look.

Initial steps taken were to create areas of interest (AOIs). A combination of public lands polygons, on the ground transects, and ArcGIS Pro satellite imagery were used to manually outline the study areas. The AOIs were then clipped to a Natural Resources Conservation Service (NRCS) Soil Unit layer to reduce noise in the analyses by removing soil types that were not of primarily loam soil textures.

The NPP and NDVI data were obtained and added to our project. NPP values needed to be rescaled to the reportable range of 0 to 1. Colorado Climate Center precipitation data was used to determine the threshold between "wet" and "dry" years classification. NPP and NDVI data were averaged according to this determination.

Grazed and Ungrazed soil layers were used to clip the "wet" and "dry" NPP and NDVI layers and create rasters incorporating only those select soil types within our AOIs.

All "wet" year NDVI rasters were entered in the Batch Raster to Multipoint Tool to create a point layer specifying a point for every pixel of the raster. This point layer was subsequently used to create a layer of random points within our AOIs. The Extract Multi-Value to points tool was used to assign the values from all NPP and NDVI rasters to the randomly sampled points layers for use in the statistical analyses.

Finally, the attribute tables of the random point layers were exported in .csv format, and the data was transposed and imported to R. T-tests and two-way ANOVAs were performed.

“Ranching is now the only livelihood that is based on human adaptation to wild biotic communities.” Jim Corbett

Results

Results of the multi-layered ArcGIS analysis illustrated spatial differences of (a) forage productivity (Net Primary Productivity) and (b) forage nutritive quality (Normalized Difference Vegetation Index) between historically grazed and ungrazed parcels of select government-owned landscapes on the Northern Front Range of Colorado. To statistically examine these differences, 2 types of tests were performed: Analysis of Variance and t-tests.

Analysis of Variance (ANOVA)

A Two-Way ANOVA was performed to test differences in the means of each variable as a function of both Study Site and Treatment (grazed versus ungrazed) based on the significant level alpha = 0.05. Assumptions of equal variance and normally distributed data were confirmed by examining Residual vs Fitted plots, Normal Q-Q plots, and Histograms.

An ANOVA for each variable in above average and below average precipitation classes resulted in statistical differences between grazed and ungrazed areas.

Results of Two-Way ANOVA for Net Primary Productivity (NPP) and Normalized Difference Vegetation Index (NDVI). * indicates a significant difference between grazed and ungrazed areas.

A Tukey's Honest Significant Difference (HSD) test was also performed for each ANOVA. Results underscored statistical differences between all grazed and ungrazed AOIs.

Results of Tukey's HSD test for Net Primary Productivity (NPP) and Normalized Difference Vegetation Index (NDVI). * indicates a significant difference between grazed and ungrazed areas.

T-TESTS

T-tests were used to test the hypothesis that there were actual differences in the means of grazed versus ungrazed areas for each of the variables: NPP in above and below average precipitation years and NDVI in above and below average precipitation years. T-tests were conducted using a 95% confident interval and an alpha = 0.05. A Fisher's test for homoscedasticity was performed to account for unequal sample sizes, and appropriate steps were taken (i.e. coding for equal or unequal variances) based on the result.

T-tests comparing the means of NPP and NDVI in above average and below average precipitation classes for all study sites combined resulted in statistically significant differences.

Summary table for t-tests used to compare grazed and ungrazed areas of combined study sites. * indicates a significant difference between grazed and ungrazed areas.

T-tests for Soapstone Prairie indicated that there were statistically significant differences in the means of NPP and NDVI in above average and below average precipitation classes for grazed and ungrazed areas.

Summary table for t-tests used to compare grazed and ungrazed areas on Soapstone Prairie Natural Area. * indicates a significant difference between grazed and ungrazed areas.

A Closer Look at Soapstone Prairie...

Coyote Ridge Natural Area

T-tests for Coyote Ridge indicated that there were statistically significant differences in the means of NPP and NDVI in above average and below average precipitation classes for grazed and ungrazed areas.

A Closer Look at Coyote Ridge...

Coalton Trail Open Space

T-tests for Coalton Trail indicated that there were NOT statistically significant differences in the means of NPP and NDVI in above average and below average precipitation classes for grazed and ungrazed areas. While raw means for grazed areas were consistently higher than for ungrazed areas, a statistical difference was not found.

A Closer Look at Coalton Trail...

Lowry Ranch

T-tests for Lowry Ranch indicated that there were statistically significant differences in the means of NPP in above average and both NPP and NDVI in below average precipitation classes. However, NDVI for above average precipitation years was not significantly different between grazed and ungrazed areas.

A Closer Look at Lowry Ranch...

Conclusions

Discussion of statistical results:

The purpose of this study was to compare grazed and ungrazed areas of select government-owned lands using remote sensing data. This is not to be confused with a comparison between different grazing regimes. While all of these sites have been subject to slightly varying grazing strategies, they all share the common thread of long-term, collaborative grazing management. Intricate management plans for these sites have been created on varying scales of collaboration with partnering ranchers/producers and prioritize the conservation of natural resources, like biodiversity, soil health, and wildlife habitat.

ANOVA and t-tests across all study sites indicated an overall trend toward higher levels of NPP and NDVI in areas that have been historically managed by cattle grazing compared to areas that have been excluded from grazing.

Site-specific results largely reflect the same conclusion. One exception was Coalton Trailhead Open Space, which did not show statistically significant differences between grazed and ungrazed areas, although raw averages of NPP and NDVI in grazed areas were higher than in the exclosures. The other exception was Lowry Ranch, which demonstrated slightly higher NPP in ungrazed areas, and no signifiant difference in NDVI in above average precipitation years. However, NDVI in below average precipitation years was statistically higher in grazed areas. Further investigation and improvements in the model will aid interpretation of these results.

Overall, it can be concluded that in this region, select government-owned landscapes that are collaboratively managed with local ranchers may exhibit higher forage productivity and nutritive quality compared to areas that have not been managed, at least in-part, by long term grazing leases. This is a significant finding in an industry with a long history of controversy regarding the use of government lands for cattle production. It is noted that interpretation of results must be site-specific and take into account the unique management history and timeline of each site.

Scroll/click to the right for more -->

Challenges and shortcomings of the analysis:

Sourcing the appropriate data was the first significant hurdle in this analysis. Ensuring that the data source was of quality, highly regarded by the scientific community, and utilized in comparable studies was important. We took careful steps to confirm that data was available for our study's time frame, and that the resolution was consistent across variables.

Working with compromised or imperfect imagery from remote sensing satellites was also a challenge, but fortunately the issue wasn't widespread. Specifically, the Landsat NDVI imagery exhibited "striping" and "holes" in the data. Complex destriping or stripe noise reduction methods exist, but were not implemented since the level of striping on only a few datasets was faint, and not numerically pronounced. To correct small holes over the Lowry Ranch AOI for the year 2011 and 2015, we used the Set Null tool in ArcGIS Pro to eliminate areas that were missing values.

Choosing a point sampling method was the most difficult challenge in this study. The difficulty arose from the fact that each AOI had a different land area and therefore a different pixel count, which made a "balanced" sampling design an ambiguous concept. We also wanted to account for spatial autocorrelation and ensure that points entered into statistical analysis were truly independent samples and not pseudo-replicates. In on-the-ground rangeland science, time, geography, and economic constraints dictate sample size and location, but with GIS, the available data is almost overwhelming. For our study, we chose a combination of methods to reduce biological variability and improve sample independence among AOIs, like controlling for consistent soil types and lumping data into above average and below average precipitation classes. Within this reduced biologically variable space, we then chose to take a random sample of points with a minimum distance constraint. This decision was a result of scrutinizing each AOI and identifying the maximum number of potential points across grazed and ungrazed AOIs within a minimum distance constraint. The goal was to ensure ample grazed and ungrazed AOI sample sizes, while reducing potential autocorrelation and increasing sample independence.

Improving the model would incorporate increased effort to isolate grassland and pasture lands from even small areas of tree cover and manmade disturbance, like pavement or hiking trails, which may create unnecessary noise in the data and results. This could be accomplished by incorporating a GIS layer from the National Land Cover Database and masking-out areas of land cover that were not of the "grassland/herbaceous" type. Integrating local knowledge from ranchers or land agency partners could further identify and eliminate parts of the AOI that have experienced heavy disturbance (i.e. picnic areas, gravel pits) and therefore skew the results.

Looking Ahead:

As a result of our study we recommend several paths for future research. Using our methods, one could investigate further into specific grazing methods and effects on NPP and NDVI by spatially and/or temporally isolating specific allotments or pastures by grazing strategy. We also recommend further investigation into user-friendly destriping or noise correction methods for remotely-sensed raster datasets. Another area of important future study involves determining appropriate sampling methods for highly variable natural landscapes. This is a fascinating issue that exists only because of the plethora of possibilities in working with GIS and remotely-sensed data. We feel this is a vital issue as the scientific community continues to increasingly embrace GIS and remote sensing applications.

Finally, we conclude and reiterate that remote sensing is a powerful alternative tool for spatially and/or temporally extensive ecological research. With relatively minor labor and cost inputs, land stewards may harvest critical patch and landscape scale information with proven efficacy and accuracy. It is important that the data is not used incorrectly, as in the case of over-interpolation or over-extrapolation. When used correctly, data derived from remote sensing, like NPP and NDVI, can provide valuable information for the collaborative management of government-owned lands, whose social-ecological objectives are multifaceted and pose complex challenges for decision-making and sustainability science.

References

Abdalla, M., Hastings, A., Chadwick, D. R., Jones, D. L., Evans, C. D., Jones, M. B., . . . Smith, P. (2018). Critical review of the impacts of grazing intensity on soil organic carbon storage and other soil quality indicators in extensively managed grasslands. Agriculture, ecosystems & environment, 253, 62-81.

Barry, Joshua M., et al. (2019). Pumas as ecosystem engineers: ungulate carcasses support beetle assemblages in the Greater Yellowstone Ecosystem. Oecologia 189.3, 577-586.

Bell, R. H. (1971). A grazing ecosystem in the Serengeti. Scientific American, 225(1), 86-93.

Briske, D., Derner, J., Brown, J., Fuhlendorf, S., Teague, R., Gillen, B., . . . Willms, W. (2008). Benefits of rotational grazing on rangelands: an evaluation of the experimental evidence. Rangeland Ecology and Management, 61(2), 3-17.

Briske, D. D., Sayre, N. F., Huntsinger, L., Fernandez-Gimenez, M., Budd, B., & Derner, J. D. (2011). Origin, Persistence, and Resolution of the Rotational Grazing Debate: Integrating Human Dimensions Into Rangeland Research. Rangeland Ecology & Management, 64(4), 325-334.

Brunson, M. W., & Huntsinger, L. (2008). Ranching as a conservation strategy: can old ranchers save the new west? Rangeland Ecology & Management, 61(2), 137-147.

Colorado Climate Center, Department of Atmospheric Science, Colorado State University (2020). Fort Collins, CO. https://climate.colostate.edu/general.html. (Accessed November 1, 2020).

Curtin, C. G., Sayre, N. F., & Lane, B. D. (2002). Transformations of the Chihuahuan Borderlands: grazing, fragmentation, and biodiversity conservation in desert grasslands. Environmental Science & Policy, 5(1), 55-68.

Daniel, J., Potter, K., Altom, W., Aljoe, H., & Stevens, R. (2002). Long–term grazing density impacts on soil compaction. Transactions of the ASAE, 45(6), 1911.

Derner, J. D., Boutton, T. W., & Briske, D. D. (2006). Grazing and ecosystem carbon storage in the North American Great Plains. Plant and Soil, 280(1-2), 77-90.

Derner, J. D., Augustine, D. J., & Frank, D. A. (2018). Does Grazing Matter for Soil Organic Carbon Sequestration in the Western North American Great Plains? Ecosystems, 1-7.

Donahue, D. L. (1999). The western range revisited: removing livestock from public lands to conserve native biodiversity (Vol. 5): University of Oklahoma Press.

Elbroch, L. Mark, et al. (2017). Vertebrate diversity benefiting from carrion provided by pumas and other subordinate, apex felids. Biological Conservation, 215, 123-131.

Environment Colorado Research and Policy Center. (2006). Losing Ground: Colorado’s Vanishing Agricultural Landscape.

ESRI. (2019). NDVI Function. ESRI Press. https://desktop.arcgis.com/en/arcmap/10.3/manage- data/raster-and-images/ndvi-function.html

Fensholt, R., et al. (2012). Greenness in semi-arid areas across the globe 1981–2007—An Earth Observing Satellite based analysis of trends and drivers. Remote Sensing of Environment, 121, 144–158.

Gates, David M. (1980) Biophysical Ecology, Springer-Verlag, New York, 611.

Gibson, D. J. (2009). Grasses and grassland ecology: Oxford University Press.

Lubell, M. N., Cutts, B. B., Roche, L. M., Hamilton, M., Derner, J. D., Kachergis, E., & Tate, K. W. (2013). Conservation program participation and adaptive rangeland decision-making. Rangeland Ecology & Management, 66(6), 609-620.

Mladinich, C. S. (2006). Regional Landscape Change in Northern Colorado Front Range. US Geological Survey professional paper(1726), 139-152.

Montgomery, B., Dragićević, S., Dujmović, J., & Schmidt, M. (2016). A GIS-based Logic Scoring of Preference method for evaluation of land capability and suitability for agriculture. Computers and Electronics in Agriculture, 124, 340-353.

Pietola, L., Horn, R., & Yli-Halla, M. (2005). Effects of trampling by cattle on the hydraulic and mechanical properties of soil. Soil and tillage research, 82(1), 99-108.

Pizon, J. (2005). Satellite time series correction of orbital drift artifacts using empirical mode decomposition. Hilbert-Huang transform: introduction and applications, 167-186.

Provenza, F., Pringle, H., Revell, D., Bray, N., Hines, C., Teague, R., . . . Barnes, M. (2013). Complex creative systems: principles, processes, and practices of transformation. Rangelands, 35(5), 6-13.

Reeder, J. D., Schuman, G. E., Morgan, J. A., & LeCain, D. R. (2004). Response of organic and inorganic carbon and nitrogen to long-term grazing of the shortgrass steppe. Environmental management, 33(4), 485-495.

Reeves, M. C., Winslow, J. C., & Running, S. W. (2001). Mapping weekly rangeland vegetation productivity using MODIS algorithms. Journal of Range management, 54, A90.

Roche, L. M., Schohr, T. K., Derner, J. D., Lubell, M. N., Cutts, B. B., Kachergis, E., . . . Tate, K. W. (2015). Sustaining working rangelands: Insights from rancher decision making. Rangeland Ecology & Management, 68(5), 383-389.

Rouse, J. W., Haas, R. H., Schell, J. A., & Deering, D. W. (1974). Monitoring vegetation systems in the great plains with ERTS. In Third Earth Resources Technology Satellite-1 Symposium (Vol. 1, pp. 331–340). NASA.

Sharrow, S. H. (2007). Soil compaction by grazing livestock in silvopastures as evidenced by changes in soil physical properties. Agroforestry Systems, 71(3), 215-223.

Sherwood, K. (2010). Common Ground for Competing Issues. Colorado College State of the Rockies Report Card.

Soil conservation service. (1975). Soil Taxonomy. Agríc. Handbook(436).

Teague, W., Dowhower, S., Baker, S., Haile, N., DeLaune, P., & Conover, D. (2011). Grazing management impacts on vegetation, soil biota and soil chemical, physical and hydrological properties in tall grass prairie. Agriculture, ecosystems & environment, 141(3), 310-322.

Teague, W., Apfelbaum, S., Lal, R., Kreuter, U., Rowntree, J., Davies, C., . . . Wang, T. (2016). The role of ruminants in reducing agriculture's carbon footprint in North America. Journal of Soil and Water Conservation, 71(2), 156-164.

Theobald, D.M., Stevens, D.L., White, D. et al. (2002) Using GIS to Generate Spatially Balanced Random Survey Designs for Natural Resource Applications. Environmental Management 40, 134–146. https://doi.org/10.1007/s00267-005-0199-x.

Timmins, S. M. (2002). Impact of cattle on conservation land licensed for grazing in South Westland, New Zealand. New Zealand Journal of Ecology, 107-120.

Tucker, C. J., Pinzon, J. E., Brown, M. E., Slayback, D. A., Pak, E. W., Mahoney, R., ... & El Saleous, N. (2005). An extended AVHRR 8-km NDVI dataset compatible with MODIS and SPOT vegetation NDVI data. International Journal of Remote Sensing, 26(20), 4485-4498.

United States Department of Agriculture. (2017). 2017 Census of Agriculture (AC-12-A-6). Retrieved from

The Landscape Toolbox. (2013). Net Primary Productivity. Retrieved Sunday December, 2020 https://wiki.landscapetoolbox.org/doku.php/remote_sensing_methods:net_primary_productivity#:~:text=NPP%20%3D%20GPP%20%E2%80%93%20respiration.,an%20indicator%20of%20ecosystem%20function.

https://www.nass.usda.gov/Publications/AgCensus/2017/Full_Report/Volume_1,_Ch apter_1_State_Level/Colorado/

Western Regional Climate Center. (2010-2020). Total Precipitation. Retrieved 12 September 2019 https://wrcc.dri.edu/cgi-bin/cliMAIN.pl?co3005

Wilmer, H., Derner, J. D., Fernández-Giménez, M. E., Briske, D. D., Augustine, D. J., Porensky, L. M., & Group, C. S., a. (2018). Collaborative adaptive rangeland management fosters management-science partnerships. Rangeland Ecology and Management, 71(5), 646-657.

Wilmer, H., Augustine, D. J., Derner, J. D., Fernández-Giménez, M. E., Briske, D. D., Roche, L. M., . . . Miller, K. E., b. (2018). Diverse management strategies produce similar ecological outcomes on ranches in western great plains: Social-Ecological assessment. Rangeland Ecology & Management, 71(5), 626-636.

World Wildlife Fund. Sustainable Agriculture: Beef. Retrieved from https://www.worldwildlife.org/industries/beef