Progress report for GW19-196
Rangelands (e.g., grasslands, savanna, shrubland and woodland ecosystems) constitute ~50% of the Earth’s land surface and support the majority of the world’s livestock production. Rangelands also play integral roles in the global carbon, water, and nitrogen cycles and human health due to their extensive coverage. Many rangelands have undergone a shift from herbaceous plants to shrubs resulting in a decline of valuable livestock forage (e.g., perennial grasses). This phenomenon, known as “woody plant encroachment” or “shrub encroachment”, represents ecosystem degradation in arid and semi-arid regions and is typically accompanied by increases in bare soil and soil erosion. Efforts to limit shrub encroachment into rangelands include a variety of reactive brush management practices that remove or reduce the woody species via prescribed fire, mechanical, and/or herbicidal treatments. However, results of such actions are often short-lived and are seldom cost-effective or sustainable.
Instead of a reactive approach, preventative actions might be more ecologically sound and cost-effective for limiting shrub encroachment. I propose to develop SEEDS (Shrub Encroachment Early Detection System), a user-friendly tool (e.g., online application) that can be used as a first-step to promote preventative and proactive alternatives to reactive brush management. SEEDS will combine data from the literature and active research projects designed to understand how shrubs establish in grasslands with remote sensing and geospatial landscape-scale data sets to predict areas at risk for shrub encroachment. The goals for SEEDS are to (1) alert land managers, livestock producers, and citizen scientists to critical times and locations for inventory and monitoring for shrub invasion, (2) identify windows of opportunity for early intervention/brush management, and (3) assist in prioritization of shrub encroachment mitigation efforts at the scale of allotments/pastures.
The goal is to develop a SEEDS (Shrub Encroachment Early Detection System) framework for federal, state, and county land managers/rangeland specialists, ranchers, and the general public as a proactive and preventative measure to woody plant encroachment (WPE) to maximize the effectiveness of brush management strategies. Specific objectives are to:
(1) Compile published research on WPE to identify the primary drivers of and controls over woody plant establishment.
(2) Utilize existing remote sensing products and geospatial data (e.g., ecological site, soil type, depth to restrictive layer, precipitation, temperature, elevation, slope, etc.) to model WPE risk.
(3) Develop the SEEDS application.
- Combining models from (2), perform statistical analyses and validation of spatially explicit WPE risk.
- Create functions that enable spatially explicit user input (e.g., land use, stocking rates, management unit boundaries, etc.) within a GIS (Geographic Information System).
- Link Objectives 1-3 within the SEEDS framework to predict when, where, and under what conditions a given pasture/allotment will be at risk for WPE.
(4) Test and refine the SEEDS algorithms to maximize their accuracy for predicting WPE risk.
Areas identified as being at high risk for WPE could then be subject to monitoring and plans made for preventative measures (e.g., prescribed fire) to be used before encroachment progresses to the point where more expensive, less effective approaches are necessitated.
Abbreviations used in this report are listed alphabetically in Appendix A
A comprehensive timeline is presented in a Gantt Chart. An Advisory Board meeting in Summer 2019 will consolidate, finalize, and develop SEEDS partnerships with federal and private land managers. The woody plant encroachment literature review will begin in Summer 2019 and run through Fall 2019 (Objective 1 Methods). DroughtView modifications, geospatial modeling, and user function construction for SEEDS (Objectives 2&3 Methods) will run from Fall 2019 through Fall 2020. As finalization of SEEDS infrastructure and design approaches, testing and refinement (Objective 4 Methods) will commence Summer 2020 through Winter 2020. Spring 2021 will be focused on incorporation of SEEDS into existing rangeland monitoring protocols, workshop material development and distribution, and scientific article drafting and submissions.
- - Technical Advisor (Researcher)
- - Technical Advisor (Educator and Researcher)
- - Technical Advisor
- - Technical Advisor (Educator and Researcher)
Objective 1: Compile published research on Woody Plant Encroachment (WPE) to identify the primary drivers of and controls over WPE.
Published research was compiled using Clarivate Analytics Web of Science and Google Scholar. Search keywords included combinations of: “brush”, “shrub”, “manag*”, “encroach*”, “woody”, “control”, and “ecol*”, where information and variables pertaining to setting (vegetation, soil, climate, topography), management history (AUMs, grazing rotation), focal species (native status, demographics, encroachment history), and brush management undertakings (type, timing, success) were examined. Articles were stored and organized in a Mendeley Reference Manager database. Thus far, ~1400 articles have been compiled. The leading biophysical drivers of and constraints to woody cover potential are hypothesized to be climate (seasonal and annual mean precipitation/temperature), soils (clay content and depth to restrictive layer), topography (slope, aspect, elevation), and proximity to washes/arroyos. In addition to the literature searches, multiple discussions during Field Days and Tours (see Educational & Outreach Activities) with rancher/producers occurred regarding shrub encroachment patterns across their lands. Interesting patterns emerged from these discussions on shrub cover/recruitment differences between ecological sites, soil conditions, wet/drought years, and grazing management history and reinforce/refine findings from literature searches
Objective 2: Utilize existing remote sensing products and geospatial data (e.g., ecological site, soil type, depth to restrictive layer, precipitation, temperature, elevation, slope, etc.) to model WPE risk.
Geospatial Data Compiling and Pre-processing
A total of 31 remote sensing and geospatial products relating to biophysical attributes (e.g., ecological sites, maximum clay content, depth to maximum clay content, seasonal precipitation and temperature, slope aspect/inclination, etc.) have been compiled, georeferenced, and clipped to the ~21,000-hectare Santa Rita Experimental Range (SRER) using ESRI ArcGIS Pro (version 2.4.1). The compiled biophysical variables include those related to climate (temperature and precipitation), soils, topography, and hydrology. A select subset of model variables are presented in Figs. 1-3 as examples. All variables were resampled to 30 meters to match Landsat 8 resolution for use in final SEEDS model development as independent variables.
Figure 1. Percent clay (A) and depth to maximum clay percentage (B) rasters (30-m resolution) developed using the NRCS gSSURGO database for the Santa Rita Experimental Range. Areas in white are sites lacking clay (e.g., sandy wash beds and rock outcrops).
Figure 2. Ecological sites on the Santa Rita Experimental Range (30-m raster; based on Breckenfeld and Robinett 1996).
Figure 3. Distance from wash/arroyo raster (30-m) calculated using the USGS National Hydrography Dataset for the Santa Rita Experimental Range. Distance calculations were based on Euclidean Distance from each wash/arroyo and rasterized.
Shrub Cover Imagery Classification
Landsat 8 Vegetation Indices (VI). By relating shrub cover to Landsat 8 VI, changes in shrub cover can be quantified across time and space using historical Landsat imagery. Thus, Landsat-based woody cover values are used as the SEEDS model dependent variable. An Iso Cluster Unsupervised Classification (Memarsadeghi et al., 2007; Schowengerdt, 2007) was completed using USDA NAIP (National Agricultural Imagery Program) from July 2, 2015 with a spatial resolution of 1-m. The classification was performed using red, green, blue, and near infrared bands and a Canopy Height Model for shrub/woody plant detection. The classified image was evaluated using the true color NAIP imagery to detect the woody cover classes; remaining classes represented non-woody land cover. A majority filter was applied to the classified image to remove small and isolated pixels with a sensitivity of four neighboring cells. An accuracy assessment of the classified image was performed using independent orthorectified high-resolution (10-cm) imagery obtained from the National Ecological Observation Network (NEON) Airborne Observation Platform (AOP) August 20-28, 2017 mission. A total of 200 points (100 per woody and 100 per non-woody classes) were used for binomial classification accuracy assessment following McCoy (2005).
Landsat 8 Vegetation Indices and Shrub Cover Modeling
Shrub cover (%) was calculated using the classified 1-m NAIP image with a moving window (30 X 30 pixel) average to achieve a 30-m resolution and match the spatial resolution of the Landsat 8 imagery. The shrub cover on NAIP-derived 30-m raster was binned in 10% increments. The relationship between shrub cover and Landsat 8 VI was then used to identify the biophysical variables most important for predicting a site-specific woody cover. The VIs chosen for analysis were MSAVI (Modified Soil-Adjusted Vegetation Index) and NDVI (Normalized Difference Vegetation Index). The NAIP-derived % woody cover and each VI product was clipped to the SRER boundary and sampled (20 points per class for a total of 140 points) using an equally stratified random design. Point samples were split (i.e., 80% for training the models, 20% for testing and assessing model accuracy) in a cross-validation procedure. The relationship (e.g., linear, quadratic, polynomial, etc.) between NAIP-derived woody cover and Landsat 8 VI were explored for the best model fit. The resulting best-fit relationship was then used to quantify woody cover across the SRER.
Objective 3: Develop the SEEDS application.
a) Combining models from (2), perform statistical analyses and validation of spatially explicit WPE risk.
The Landsat 8-based woody cover estimates (dependent variable) and the 31 biophysical rasters (independent variables) were sampled (n = 237,659 pixels) over the entire SRER. Stepwise regression, Support Vector Machine, and Random Forest modeling were used to examine independent variable importance and assess the model accuracy (e.g., Root Mean Square Error, R2, and Mean Absolute Error [excluding stepwise regression]) in predicting woody cover. Based on its superior accuracy, Random Forest model-based shrub cover was used for calculating woody cover potential and subsequently, WPE risk (i.e., the ratio of shrub cover estimated by the Enhanced Vegetation Index to potential woody cover). The WPE risk estimates were binned into 5 classes using Natural Breaks (Jenks, 1967), where areas with the lowest risk have a class of 1 and highest-risk areas are classified as 5.
b) Create functions that enable spatially explicit user input (e.g., land use, stocking rates, management unit boundaries, etc.) within a GIS (Geographic Information System).
Additional collaborators (e.g., UArizona Communications and Cyber Technologies [CCT] and Rangelands Partnership have recently joined the project to provide guidance, expertise and resources to better integrate the modeling work and development of the geospatial web application into a comprehensive extension and outreach framework. Their addition to the project stems from a USDA-NIFA-RREA grant obtained in Fall 2020 (see Participation Summary section below). This collaboration will enable broader user/stakeholder-guided involvement in web application design and development (specifically web product mock-ups/wire-framing, testing, and creation) and help ensure the web product meets user expectations and decision-making needs. Our collaborators at CCT are experts in designing and developing web- and mobile-based GIS user app functions (e.g., user-provided data/shapefile uploading; shape and polygon drawing; map data extraction, graphing, and reporting) used in natural resource management and inventorying purposes. For app examples, see myRAINgeLog and Desert Landscaping.
With the additional collaboration and design/development processes, finalization of the web-based application will take longer than initially planned. Accordingly, we will be postponing the proposed outreach activities (e.g., workshops/web application demonstrations to stakeholders and user guides/technical documents) until Spring 2022 (see Table 2 below).
c) Link Objectives 1-3 within the SEEDS framework to predict when, where, and under what conditions a given pasture/allotment will be at risk for WPE.
Random Forest model sensitivity was initially assessed by running the 2015 woody cover model on 2005 and 1995 Landsat NDVI data. This allowed for the comparison of the differences in potential woody cover values between multiple years. These years were chosen to allow for time-steps long enough for detectable changes in woody cover to occur and to include a range of weather conditions (e.g., drought periods vs wet periods). Additionally, the 1995 imagery contained effects of a wildfire that occurred in May/June 1994, which allowed for the assessment of model performance and modeled woody plant encroachment risk responses to wildfire (à la brush management). Next steps include evaluation of model sensitivity forward in time relative to our 2015 start year with 2020 Landsat imagery. The Sawmill Fire of 2017 on the SRER will provide an additional test of the SEEDS/WPE risk model sensitivity to fire as a brush management treatment.
Objective 4: Test and refine the SEEDS algorithms to maximize their accuracy for predicting WPE risk.
The SEEDS (Shrub Encroachment Early Detection System) application user functionalities, including the testing and refinement of the model for predicting WPE risk, are the principal goals for project Year 3. The current progress in Year 1 and Year 2 will allow for timely finalization of Objectives 3 and 4 during project Year 3, and project progress is currently in line with our newly proposed timeline (Table 1).
Table 1. Updated project #GW19-196 timeline from approved (as of 6/28/2021) No Cost Extension.
Landsat 8 Vegetation Indices and Woody Cover Modeling
The 2015 NAIP imagery woody vs non-woody classification accuracy and error matrix is provided in Table 2. A total of 140 sample points across 7 cover classes (20 points per class) were used to investigate the relationship between NAIP-derived woody cover (true color imagery example = Fig. 4A; Fig. 4B = derived woody cover) and Landsat 8 NDVI and MSAVI (Fig. 5). These indices were chosen due to their documented performance and application within arid/semi-arid ecosystems including the SRER (Huete & Jackson, 1987; Hueng et al., 2018; Holifield-Collins et al., 2015). The best-fit lines for both NDVI (RMSE = 7.89%, R2 = 0.8, p ≤ 0.001) and MSAVI (RMSE = 10.8%, R2 = 0.72, p ≤ 0.001) were quadradic, wherein shrub cover becomes asymptotic at ~70%. The high cover class areas are concentrated near washes and the mountain foothills. Landsat 8 NDVI was used for developing Landsat-based woody cover estimations for further modeling given a lower RMSE compared to MSAVI (Fig. 6). Furthermore, high-resolution NAIP imagery (example, Fig. 4A) availability is limited in time and space with a relatively short collection history compared to the moderate-resolution Landsat, which has been collected every 8 or 16 days dating back to 1982. This flexibility in the Landsat dataset allows for broader applications of the final SEEDS model beyond the SRER and over longer time-periods. Provided the woody cover error rate of 7.9%, the Landsat-based woody cover estimates are elevated slightly compared to the classified imagery (Fig. 4B), but the high versus low woody cover areas were accurately modeled (e.g., high cover near washes and in higher elevation zones) using only the Landsat 8 NDVI values.
Figure 4. Natural color 1-m 2015 NAIP imagery (A) and resulting 30-m shrub cover percent (B) raster-based shrub/non-shrub classifications. The shrub cover raster was used in conjunction with Landsat 8 vegetation indices (e.g., NDVI and MSAVI) to develop algorithms to quantify the amount and pattern of shrub cover on historical Landsat imagery.
Table 2. Error matrix to assess the accuracy of the classification performed on July 2, 2015 NAIP images of the Santa Rita Experimental Range. The reference data for the accuracy assessment was the 2017 NEON Airborne Observation Platform at a 10-cm spatial resolution.
Figure 5. Classified USDA NAIP (National Agricultural Imagery Program) and Landsat 8 NDVI (Normalized Difference Vegetation Index) (A) and MSAVI (Modified Soil Adjusted Vegetation Index) quadratic regression. Points (n = 140) were equally stratified along 7 classes in 10% shrub cover increments for model development. This relationship was used for modeling woody cover across the Santa Rita Experimental Range (see Fig. 6).
Figure 6. Classified woody cover in 2015 (9 cover classes in 10% increments) modeled using NAIP (National Agricultural Imagery Program) woody cover and Landsat 8 OLI NDVI (Normalized Difference Vegetation Index). The maximum and minimum percent woody cover values were 88% and 0%, respectively. The densely vegetated areas (e.g., washes and foothills) appear orange to dark red due to their high shrub cover. Blue areas are bare soil patches and/or Pleistocene aged surfaces with high clay content and low shrub cover (see also Fig. 1A).
Modeling Woody Cover Based on Biophysical Variable Importance
Landsat 8 NDVI values show patterns of shrub cover but do not provide insight into the biophysical/ecological variables that determine cover site potential. A total of 31 independent variables pertaining to seasonal and annual mean precipitation (PPT) and temperature (T), topography, soil characteristics, and ecological site were used in a stepwise regression, Support Vector Machine (SVM), and Random Forest (RF) modeling procedures for determining 1) the best/most accurate variables governing woody cover potential; 2) the relative importance of the 31 biophysical variables potentially influencing maximum woody cover; and 3) the SEEDS model best suited to represent woody plant encroachment risk. A graphical comparison of the three models is shown in Fig. 7.
Figure 7. Stepwise regression (A), Support Vector Machine (SVM; B), and Random Forest (C) woody cover model predictions versus 2015 Landsat NDVI-based woody cover for the Santa Rita Experimental Range. Per-pixel cross-validation (n = 45,839) was used for assessing each model’s prediction performance. Blue lines represent a 1:1 fit; red line is the linear regression of the actual fit.
The stepwise regression modeling (RMSE = 5.6%, R2 = 0.457) revealed 12 variables to be the most important: aspect; depth to restrictive layer/bedrock; soil clay percent; depth to maximum clay percent; maximum 1980-2010 T; distance to the nearest wash; mean winter PPT; minimum spring and fall T and maximum spring and summer T. Although the stepwise error was relatively low (RMSE = 5.6%), the prediction utility was poor (Fig. 7A) and the temperature variable/model variability was unconstrained as reflected in a low R2 value (R2 = 0.46). The SVM model for woody cover improved the overall model prediction errors (Mean Absolute Error [MAE] = 4.2%; RMSE = 5.5%) and accounted for more of the variation (R2 = 0.63) compared to the stepwise regression model. The best modeling technique was found to be the 2015 RF with the lowest modeling errors (MAE = 3.22%; RMSE = 4.2%), best fit (R2 = 0.70), and strongest prediction utility (Fig. 7C). The most important biophysical variables/predictors for determining woody cover potential, as determined via Gini Index of Impurity, were elevation, % soil clay, slope aspect, fall PPT, winter PPT, slope inclination, and distance to the nearest wash (Fig. 8).
Figure 8. Woody cover model predictor variable importance for the Santa Rita Experimental Range as calculated by the Gini Index of Impurity.
Model Testing Back-in-Time
Using the same biophysical predictor variables as the 2015 RF model (e.g., elevation, % soil clay, slope aspect, fall PPT, winter PPT, slope inclination, and distance to the nearest wash), the 1995 RF regression model (Fig. 9A) had strong predictive power with a low error (RMSE = 3.93%) and model variability constraint (R2 = 0.74). The variable importance of the 1995 RF model vs the 2015 RF model ranked elevation higher than fall PPT and the importance of % soil clay was reduced compared to topographic variables (e.g., slope aspect and slope inclination). The differences in predictor variable importance are possibly due to PPT patterns (e.g., drought vs wet vs normal periods) during the fall and winter months across the SRER, where further examination of the historical climate on the SRER is needed. Similarly, the 2005 RF model (Fig. 9B) ranked topographic variables more important in the model compared to 2015, but model performance was still quite robust (RMSE = 2.64%, R2 = 0.76). The consistency in low error rates and amount of explained variance of the multiple RF models based solely on biophysical variables provides support for use in woody cover modeling at the SRER as compared to SVMs or stepwise regression techniques.
Figure 9. Random Forest models of woody cover potentials on the Santa Rita Experimental Range in (A) 1995 (max = 82.6%, min = 0%), (B) 2005 (max = 80%, min = 0%), and (C) 2015 (max = 77%, min = 0%). The densely vegetated areas (e.g., washes and foothills) appear orange to dark brown due to their high shrub cover. Yellow areas are bare soil patches and/or Pleistocene-aged soils with high clay content and low shrub cover.
The Random Forest model predictions for maximum potential woody cover were used in finalizing the proposed SEEDS models (Fig. 10A-C) for determining a site’s woody plant encroachment (WPE) risk. Risk was calculated as the ratio of a site’s shrub cover on a given date (estimated by the Enhanced Vegetation Index for a given year: 1995, 2005, 2015) to its potential shrub cover. The wildfire that occurred in 1994 in the center of the SRER was detectable in the 1995 potential woody cover model (Fig. 9A, light yellow) and in the resulting WPE risk map (Fig. 10A, dark brown). Much of the burned region was classified as moderately high to high WPE risk in the 1995 due to the potential for velvet mesquite, a vigorous sprouter tolerant of top-kill, to recover or establish post-fire. The burned moderately high to high risk regions in 1995 then decreased in risk level slightly by 2005 (Fig.10B), whereas many of the high risk burned areas transitioned to moderate/moderately high. By 2015 (Fig. 10C), the woody plant cover increased and began approaching the site’s woody cover potential, as reflected by a decrease in risk levels as compared to 2005. The regions of the SRER dominated by bare soil patches and/or Pleistocene aged soils with high clay content and low shrub cover at the SRER remained at low to moderately low risk of shrub encroachment over the 20-year period (Fig.10).
Figure 10. Classified woody plant encroachment risk calculated as the ratio of a site’s shrub cover on a given date (estimated by the Enhanced Vegetation Index for a given year: 1995, 2005, 2015) to its potential shrub cover. The five risk classes range from Low to High and were defined by Natural Breaks (Jenks, 1967). The areas at most risk appear orange to dark brown due to their high potential to increase shrub cover. Yellow to blue areas are low to moderate risk areas reflective of low shrub cover potential and/or areas that have reached their potential woody cover.
The percent area for each WPE risk level for 1995, 2005, and 2015 is summarized in Table 3. Areas at the most risk have been decreasing over time ostensibly due to sites across the SRER experiencing an increase in woody cover over time as they progress toward their maximum potentials on the various topoedaphic settings. Accordingly, the low to moderately low levels have been increasing in area. It should be noted that the 1995 high risk percent area (5.6% of the SRER) is likely inflated due to the 1994 wildfire, where the moderately high risk areas remained rather stable between 1995 and 2005 (24.6% vs 24.9%, respectively). The largest change in WPE risk occurred between 1995 and 2005, where the moderately low risk areas increased by 7.5% or ~3,940 acres (Table 4) reflective of the reduction in moderate to high risk sites. Percent area calculations and acreage estimates for each WPE risk level can be used by stakeholders to estimate when and where to implement brush management treatments on a landscape or watershed. These calculations will be provided once all modeling activities have been finalized in project Year 3.
Table 3. Percent area of the ~21,000-hectare Santa Rita Experimental Range for a given WPE risk level in 1995, 2005, 2015.
Table 4. Absolute percent change in WPE risk level area between the three evaluated time steps (e.g., 2015 vs 2005, 2015 vs 1995, and 2005 vs 1995). Calculated from percentages provided in Table 3.
Additional validations and statistical testing of woody cover potential models and the resulting risk calculations are in progress at the time of this writing and will be used to finalize project Objectives 3.2, 3.3 and 4 in Year 3.
This project does not contain a stakeholder academic interview component for use in analysis per se, but instead includes general participation to guide research applicability and development of the SEEDS application. Ranchers/producers and agriculture professionals have participated in local meetings and field days/tours, where this project was highlighted and discussed, including the 2020 and 2021 Altar Valley Conservation Alliance Community Meetings and the 2019 (also at the upcoming Fall 2021) Las Cienegas National Conservation Area Bio Planning Field Days (see Educational and Outreach Activities section). Targeted workshops of the SEEDS tool and brush management with quantifiable participation and surveys are slated for project Year 3 with the collaborations with CCT and the Rangelands Partnership initiated this year continuing into next year (see Materials and Methods Objective 3b). Workshops will be conducted in collaboration with University of Arizona Cooperative Extension personnel and when the SEEDS tool goes public Spring 2022.
To create opportunities for more active stakeholder involvement, we collaborated with Dr. Elise Gornish, UA Cooperative Extension, and Sheila Merrigan from the Rangelands Partnership to develop a proposal to the USDA Renewable Resources Extension Act (RREA)-National Focus Fund Project. The proposal, entitled “An Online Toolkit for Managing Shrub Encroachment” has been funded (Award No. 2020-46401-32795, effective September 2020) and will enable me to refine, elaborate and extend the activities and online presence initiated in this project. Its objectives are to compile local/regional stakeholder knowledge and input for use in decision support tool design and broaden existing brush management online resources available to stakeholders via the Rangelands Partnership. The Rangelands Partnership is a long-term initiative among 19 western and Great Plains land-grant universities that facilitates a unique collaboration between rangeland specialists, librarians, and technology experts. Thus far, this new grant has enabled us to complete potential user interviews of the SEEDS model and application (via Zoom) in Spring/Summer 2021 and gain valuable insights for tailoring web resources to meet stakeholder expectations and support their specific management needs. We have also recently partnered with CCAST (Collaborative Conservation and Adaptation Strategy Toolbox) and their Grassland Restoration Community of Practice for creating a case study on the SEEDS model and brush management geospatial application.
Educational & Outreach Activities
Instructor for the Institute for Tribal Environment Professionals Tribal Exchange Network Fall 2020 Virtual Data Academy and Conference. Taught Tribal Natural Resource Professionals on how to apply the R programming language to their own unique data manipulation, organization, and visualization tasks.
Osher Lifelong Learning Institute Lecturer on an introduction to Rangelands, Range Science, and personal research summary.
Presentations, conferences, and workshops/field days during Year 1 (2019-2020) and Year 2 (2020-2021) are listed below. In addition to these activities, an educational tool for the Natural Resource Conservation Service Soil Health Division learning module under the “Social and Economic Considerations” topic area was updated to target Arizona- and rangeland-focused learners. In Year 1, tours and field days with collaborators occurred at the Las Cienegas National Conservation Area, Altar Valley, and the Santa Rita Experimental Range in southern Arizona. During these tours and field days I was able to learn producer/manager shrub encroachment concerns and discuss brush management projects being planned or proposed. In Year 2, one tour/field day on vegetation/brush management occurred in Fall of 2020 in Altar Valley, AZ. Unfortunately, many scheduled opportunities were cancelled due to the COVID-19 pandemic in accordance with University of Arizona guidelines. Details and methods for this research has also been presented to mixed undergraduate and graduate groups in the University of Arizona College of Agriculture and Life Sciences as part of Data Science and R-language programing community meetings. My presentations emphasized the utility and applicability of Big Data modeling practices to local and regional land management goals and research questions.
Webinars, talks, and presentations presented:
- Rutherford, WA and SR Archer. (2021) Evaluating shrub proliferation in the Sonoran Desert. Oral/Lightning Talk; International Arid Lands Consortium, Virtual Meeting.
Finalist in Graduate Student Lightning Talk Competition (1 of 2 from U.S. Institutions of Higher Education)
- Rutherford, WA, S Merrigan, SR Archer, ES Gornish. (2021) Fusing range and data science: developing online resources for managing shrub encroachment. Oral Presentation; Rangelands Partnership Annual Meeting, Virtual.
- Rutherford, WA. (2020) Instructor, Natural Resources Data in R. Institute for Tribal Environment Professionals Tribal Exchange Network 2020 Virtual Data Academy and Conference.
- Rutherford, WA. (2020) The rise of mesquite shrubs in Sonoran Desert grasslands: a story of rain, cows, ants and rats. Osher Lifelong Learning Institute Lecturer on an introduction to Rangelands, Range Science, and personal research summary.
- Rutherford, WA and SR Archer. (2020) Evaluating shrub proliferation risk on Sonoran Desert rangelands. Invited Oral/Inspire presentation; Ecological Society of America Annual Meeting, Virtual Meeting.
- Rutherford, WA, SR Archer. (2020) Predicting Woody Plant Encroachment Risk on Sonoran Desert Rangelands. Poster Presentation; 73rd Annual Society for Range Management Meeting, Denver, CO.
- Rutherford, WA. (2020) AVCA 2020 Update: Shrub Encroachment Early Detection System (SEEDS). Oral Presentation; AVCA Community Meeting, Virtual.
- Rutherford, WA. (2020) Velvet Mesquite Establishment in Sonoran Grasslands. Oral Presentation; AZ Society for Rangeland Management Winter Meeting, Wickenburg, AZ.
- Rutherford, WA, SR Archer. (2019/20) Predicting Woody Plant Encroachment Risk on Sonoran Desert Rangelands. Oral Presentations; University of Arizona R-programing Community Meetings, Tucson, AZ.
- Rutherford, WA, SR Archer. (2019) Evaluating Shrub Encroachment Risk on Sonoran Desert Rangelands. Poster Presentation; Society for Ecological Restoration-Southwest, Tucson, AZ.
- Jones, S, WA Rutherford, SR Archer. (2019) Evaluating Woody Plant Encroachment in Sonoran Grasslands for Brush Management Planning. Oral presentation (co-presented by Jones and Rutherford); Science on the Sonoita Plain Symposium, Sonoita, AZ.
Conferences/Workshops/Field Days Attended (2019):
- Science on the Sonoita Plain. June 22nd, 2019, Elgin, Arizona. Organized by Cienega Watershed Partnership. Approximately 50 participants.
- Research Insights in Semiarid Ecosystems (RISE). October 26th, 2019, Tucson, Arizona. Organized by Agricultural Research Services and College of Agriculture and Life Sciences University of Arizona. Approximately 60 participants.
- Southwestern Grasslands Research & Management Workshop. November 7th, 2019, Tucson, Arizona. Organized by Southwest Division Resources/Society for Ecological Restoration. Approximately 100 participants.
- Society for Ecological Restoration Southwest (SER SW). November 8th-10th, 2019, Tucson, Arizona. Approximately 500 participants.
- Altar Valley Conservation Alliance (ACVA) Vegetation Watershed Working Group (bi-monthly from 2019-Current)
- AVCA Field Day. November 1st, 2019, Altar Valley, AZ.
- Las Cienegas National Conservation Area (LCNCA) Winter Planning Meeting/Field Day. December 12th, 2019, Sonoita, AZ. Organized by Bureau of Land Management and Southwest Division Resources. Approximately 50 participants.
Conferences/Workshops/Field Days Attended (2020):
- Society for Range Management Annual Meeting. February 16th-20th, 2020. Denver, CO.
- Arizona Society for Range Management Winter Meeting. January 22nd – 24th, 2020. Wickenburg, AZ.
- AVCA Community Meeting. May 20th, 2020. Virtual.
- AVCA Vegetation/Brush Management Field Day. October 16th, 2020.
Conferences/Workshops/Field Days Attended (2021):
- Society for Range Management Annual Meeting. February 15th-18th, 2021. Virtual.
- Arizona Society for Range Management Winter Meeting. February 4th, 2021. Virtual.
- Rangelands Partnership Annual Meeting. April 19th – 21st, 2021. Virtual.
- International Arid Lands Consortium. May 24th – 26th, 2021. Virtual.
- AVCA Community Virtual Meeting. May 26th, 2021. Virtual.
Summer/Fall 2021and Spring 2022 meetings and workshops will include
- Arizona Section Society for Range Management Summer Meeting
- Society for Rangeland Management Annual Meeting
- Society for Ecological Restoration Southwest
- UA Annual Research Insights in Semi-arid Ecosystems symposium
- AVCA Watershed Working Group Meetings (monthly)
- LCNCA Fall Planning Meeting/Field Day.
My dissertation research projects are being compiled and updated to be presented online via https://cals.arizona.edu/research/archer/exotic.html. As results from this project become available the site will be updated. This site will also nest other products of this research such as fact sheets, published manuscripts, and other outreach materials. In addition, an open-sourced software container image was developed and released for performing any statistical modeling activity in the R programing language using the ‘Tidymodels’ package. The container image was initially created for development of the woody plant encroachment risk models and initial decision support tool development but is now shared in a Docker Hub repository for potential community use.
This project is approaching the end of its second year and recently was granted a No Cost Extension for a third year. Accordingly, the full outcomes have yet to be realized. Objectives 1 and 2, which have been met, have identified the biophysical variables most applicable for determining woody cover and predicting site risk for woody plant encroachment in Sonoran Desert grasslands. Objectives 3 and 4 are focused on community/producer participation with validation of the utility of this project and the SEEDS (Shrub Encroachment Early Detection System) model. This will be the focus in Year 3 and those activities will be enhanced by the USDA-NIFA RREA award described earlier.
The data-driven, decision-support framework being developed in this project will formalize and streamline the management planning process for meeting conservation and livestock production goals. The SEEDS tool will increase the effectiveness and reduce the cost (money and time) of rangeland monitoring by directing it to landscapes most at-risk for shrub encroachment while simultaneously facilitating planning efforts for the location, type, and timing of brush management. An additional focus of this project is to reduce, limit, and concentrate future brush management treatments to minimize adverse environmental impacts while concurrently enhancing ecosystem quality for livestock production and native wildlife.
A primary focus of my efforts, outside of compiling research articles and performing geospatial analyses, has been making local and regional connections with ranchers and land management practitioners. Some of these connections were made by attending and presenting at local meetings (e.g., periodic meetings of the Arizona Chapter of the Society for Rangeland Management and the annual UA Research Insights in Semi-arid Ecosystems symposium). Many attendees at these meetings expressed interest and recognize the potential utility of a tool like SEEDS for their brush management planning needs. These interactions in relatively formal meetings, have been complimented by informal interactions with stakeholders at field days and ranch tours. The latter have been particularly rewarding as they occurred in the field and on the land. These multiple opportunities have greatly improved my appreciation for the challenges associated with outreach and Extension and have inspired Extension as a career goal. I plan to continue my outreach efforts in project Year 3 as COVID-19 safety guidelines allows.
Analysis for this project and tool development is still underway, but in Year 1 and 2 we have gleaned interesting results on the most important biophysical variables as well as methodologies for understanding woody cover and shrub encroachment potentials in southern Arizona. We have also employed innovative modeling techniques in geospatial analyses for the specific purpose of sustainable livestock production in rangeland ecology and management. This project will provide the required accuracy and information for brush management planning and decision-making by our rancher/producer and land management stakeholders.