VEGETATION SURVEY AND MAPPING OF THE FORT RILEY MILITARY RESERVATION, KANSAS

Jennifer M. Delisle1, Craig C. Freeman2, and Dana L. Peterson3

1Kansas Biological Survey University of Kansas, 2101 Constant Avenue, Lawrence, KS 66047-3759 [email protected]

2R. L. McGregor Herbarium & Kansas Biological Survey University of Kansas, 2045 Constant Avenue, Lawrence, KS 66047-3729 [email protected]

3Kansas Applied Remote Sensing Program, Kansas Biological Survey University of Kansas, 2101 Constant Avenue, Lawrence, KS 66047-3759 [email protected]

VEGETATION SURVEY AND MAPPING OF THE FORT RILEY MILITARY RESERVATION, KANSAS

Submitted December 31, 2012

Citation: Delisle, J. M., C. C. Freeman, and D. L. Peterson. 2012. Vegetation Survey and Mapping of the Fort Riley Military Reservation, Kansas. Open-file Report No. 174. Kansas Biological Survey. Lawrence, KS. 78 pp.

Appendix A may be cited as follows: Peterson, Dana L. 2012. Testing a new method for mapping encroachment of woody vegetation on the Fort Riley Military Reservation using an object-based image analysis approach. Pp. 50–77. In Delisle, J. M., C. C. Freeman, and D. L. Peterson. 2012. Vegetation Survey and Mapping of the Fort Riley Military Reservation, Kansas. Open-file Report No. 174. Kansas Biological Survey. Lawrence, KS. 78 pp.

Cover Photo: Prairie landscape at Fort Riley Military Reservation, Riley County, Kansas. Photo by Craig C. Freeman, 2011.

Table of Contents

LIST OF FIGURES ...... v

LIST OF TABLES ...... vi

ACKNOWLEDGMENTS ...... vii

ABSTRACT ...... viii

CHAPTER 1: INTRODUCTION ...... 1

1.1. PROJECT BACKGROUND ...... 1 1.2. STUDY OBJECTIVES ...... 1

CHAPTER 2: CURRENT VEGETATION CONDITIONS ...... 5

2.1. INTRODUCTION ...... 5 2.2. METHODS ...... 5 2.3. RESULTS AND DISCUSSION ...... 6 2.4. CONCLUSIONS ...... 6

CHAPTER 3: PRAIRIE ASSESSMENTS ...... 9

3.1. INTRODUCTION ...... 9 3.2. METHODS ...... 14 3.3. RESULTS AND DISCUSSION ...... 15 3.3.1. PRAIRIE ASSESSMENT RESULTS FOR 2011/2012 ...... 15 3.3.2. COMPARISON OF 2011/2012 AND 2002/2003 PRAIRIE ASSESSMENTS ...... 16 3.4. CONCLUSIONS ...... 18

CHAPTER 4: WEED SURVEYS ...... 31

4.1. INTRODUCTION ...... 31 4.2. METHODS ...... 32 4.3. RESULTS AND DISCUSSION ...... 33 4.3.1. WEED SURVEY RESULTS FOR 2011/2012 ...... 33 4.3.2. COMPARISON OF 2011/2012 AND 2002/2003 WEED SURVEY RESULTS ...... 33 4.3.3. COMPARISON OF 2007 AND 2008 SURVEY DATA FROM FRMR WITH 2011/2012 SURVEY DATA...... 36 4.3.4. ASSESSMENT OF POTENTIAL IMPACT OF SPRAYING ON SERICEA LESPEDEZA ...... 38 4.4. SPECIES OF POTENTIAL CONCERN...... 40 4.5. CONCLUSIONS ...... 41

CHAPTER 5: RARE SPECIES ...... 43

5.1. INTRODUCTION ...... 43 5.2. METHODS ...... 44 5.3. RESULTS AND DISCUSSION ...... 44 5.4. CONCLUSIONS ...... 44

LITERATURE CITED...... 47

APPENDIX A. MAPPING ENCROACHMENT OF WOODY VEGETATION ...... 51

A.1. INTRODUCTION AND BACKGROUND ...... 52 A.2. METHODS ...... 53

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE iii

A.2.1. MAPPING APPROACH ...... 53 A.2.2. DATA PREPROCESSING ...... 54 A.2.3. RULE-SET DEVELOPMENT ...... 58 A.2.4. ACCURACY ASSESSMENT ...... 60 A.3. RESULTS ...... 60 A.3.1. SUBSET MAPPING...... 60 A.3.2. ENTIRE FRMR MAPPED USING RULE-SET 1 ...... 70 A.4. DISCUSSION ...... 72 A.4.1. CHANGE DETECTION OF WOODY ENCROACHMENT ...... 72 A.4.2. LIMITATIONS & FUTURE DIRECTIONS ...... 74

LITERATURE CITED...... 77

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE iv

List of Figures

FIGURE 1.1 Location of the Ft. Riley Military Reservation (FRMR), and landscape and cultural features referenced in this report...... 3 FIGURE 1.2 Training areas boundaries on FRMR in 2011/2012...... 4 FIGURE 2.1 Vegetation of FRMR based on surveys conducted in 2011/2012...... 8 FIGURE 3.1 Summary of evaluation process for estimating viability of conservation targets...... 10 FIGURE 3.2 Histogram of size for 120 Flint Hills Tallgrass Prairies assessed on FRMR in 2011/2012...... 26 FIGURE 3.3 Histogram of floristic quality index values for 120 Flint Hills Tallgrass Prairies assessed on FRMR in 2011/2012...... 26 FIGURE 3.4 Scatterplot of prairie size vs. floristic quality index for 119 Flint Hills Tallgrass Prairies assessed on FRMR in 2011/2012...... 27 FIGURE 3.5 Scatterplot of prairie size vs. percent non-native species for 120 Flint Hills Tallgrass Prairies...... 27 FIGURE 3.6 Locations and grades of Flint Hills Tallgrass Prairies on FRMR in 2011/2012...... 28 FIGURE 3.7 Scatterplot of floristic quality indices for prairies assessed both in 2002/2003 and 2011/2012...... 29 FIGURE 3.8 Condition grade changes among prairies assessed in 2011/2012 as compared to 2002/2003...... 29 FIGURE 3.9 Overall grade changes among prairies in 2011/2012 as compared to 2002/2003...... 30 FIGURE 4.1 Populations of black locust documented on FRMR in 2011/2012...... 34 FIGURE 4.2 Populations of sericea lespedeza documented on FRMR in 2002/2003 and 2011/2012...... 35 FIGURE 4.3 Locations of prairies and infestations of sericea lespedeza on FRMR in 2011/2012...... 37 FIGURE 4.4 Number of years for which some type of spraying (spot, broadcast, aerial) has been recorded in training areas on FRMR...... 42 FIGURE 4.5 Scatterplot of estimated acres of sericea lespedeza in training areas in 2011/2012 and 2002/2003...... 42 FIGURE 5.1 Locations of Henslow’s Sparrow sightings on FRMR in 2011...... 46 FIGURE A.1 A subset of FRMR showing coarse delineations of woody and forest classes from 2002/2003...... 52 FIGURE A.2 Subsets in FRMR used to develop and test rule-sets using an object-based image analysis (OBIA) classification approach...... 54 FIGURE A.3 a. Natural-color composite of the 2010 aerial imagery for a subset of the FRMR to use as visual reference for other layers calculated; b. Example of the median filter of the nDSM derived from 2010 LiDAR data...... 55 FIGURE A.4 a. Example of the normalized difference vegetation index (NDVI); b. Example of the Sobel operator applied the near-infrared (NIR) band from 2010 aerial imagery...... 56 FIGURE A.5 False-color composite (NIR, Red and Green bands) of the South subset in FRMR...... 61 FIGURE A.6 Rule-Set 1 mapping results for the South subset of the FRMR...... 62 FIGURE A.7 Rule-Set 2 mapping results for the South subset of the FRMR...... 62 FIGURE A.8 Rule-Set 3 mapping results for the South subset of the FRMR...... 63 FIGURE A.9 False-color composite (NIR, Red and Green bands) of the Central subset in FRMR...... 64 FIGURE A.10 Rule-Set 1 mapping results for the Central subset of the FRMR...... 65 FIGURE A.11 Rule-Set 2 mapping results for the Central subset of the FRMR...... 66 FIGURE A.12 Rule-Set 3 mapping results for the Central subset of the FRMR...... 67 FIGURE A.13 Rule-Set 4 mapping results for the South subset of the FRMR...... 70 FIGURE A.14 Rule-Set 1 mapping results for the entire FRMR...... 71 FIGURE A.15 Woody encroachment mapped in the South subset for 2006...... 73 FIGURE A.16 Woody encroachment in the South subset for 2010...... 73 FIGURE A.17 The change in woody encroachment from 2006 to 2010 in the South subset of FRMR...... 74 FIGURE A.18 a. False color composite of 2010 aerial imagery for area where data were collected in February 2012; b. The map shows how the vegetation types are mapped relatively well in this area...... 75 FIGURE A.19 a. False color composite of 2010 aerial imagery for area where data were collected in August, 2012; b. Woody encroachment is overestimated in this area...... 76

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE v

List of Tables

TABLE 2.1 Classification of Natural/Semi-natural and Cultural vegetation types known to occur on FRMR...... 7 TABLE 3.1 Generalized evaluation matrix for landscape context rating × size grade...... 11 TABLE 3.2 Generalized evaluation matrix for landscape context/size rating × condition grade...... 11 TABLE 3.3 Floristic quality assessment data for all sites evaluated on FRMR in 2011/2012...... 20 TABLE 3.4 Grades for landscape context, size, and condition of Flint Hills Tallgrass Prairies evaluated on FRMR in 2011/2012...... 23 TABLE 4.1 Number of training areas (including the MPRC) in which sericea lespedeza and black locust were documented, and total area infested (in acres), in 2002/2003 and 2011/2012...... 33 TABLE 4.2 Number of infested acres and percent of cover in 10 cover classes for sericea lespedeza on FRMR in 2002/2003 and 2011/2012...... 36 TABLE 5.1 Species tracked by the Kansas Natural Heritage Inventory documented on FRMR within the last 25 years...... 43 TABLE A.1 Accuracy assessment results for Rule-Set 1–3 for the two subset areas combined...... 68 TABLE A.2 Error matrix for Rule-Set 1...... 68 TABLE A.3 Error matrix for Rule-Set 2...... 69 TABLE A.4 Error matrix for Rule-Set 3...... 69 TABLE A.5 Accuracy assessment results for Rule-Set 1 for the entire FRMR...... 72 TABLE A.6 Error matrix for Rule-Set 1 applied to the entire FRMR...... 72

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE vi

Acknowledgments

This work was carried out under cooperative agreement W9126G-10-2-0033, “Vegetation Survey and Mapping of Ft. Riley, Kansas,” between the U.S. Army and the University of Kansas Center for Research, Inc. The Kansas Biological Survey was the administrative unit responsible for the work.

We are especially grateful to Jeff Keating, Directorate of Public Works, Ft. Riley, for his support of the work, for assistance setting up the agreement, and for logistical assistance throughout the study. Also at Ft. Riley, we wish to thank Chris Jeffrey, Jerold Spohn, and Shawn White, all of whom provided assistance during the study. Hillary Loring and Frank Norman assisted with fieldwork in 2011 and/or 2012. The project could not have been completed without their expertise, endurance, and good humor. Katie Clemens and Amanda Sevcik assisted with data processing for the floristic quality assessment. Ryan Surface, Aoesta Khalid Mohammed, Alex Gareis, and Kevin Dobbs contributed to the effort to map woody encroachment on the installation.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE vii

Abstract

In 2010, the Kansas Biological Survey initiated a 2-year project to re-examine the vegetation of the Ft. Riley Military Reservation in northeast Kansas. The focus of this work was to document changes on the installation since 2002/2003, when baseline vegetation studies were conducted. Specific objectives of the project were to: 1) survey and assess the current condition of the vegetation in an 800-acre tract (Training Areas 102 and 103) formerly in the Impact Area and consequently not surveyed in 2002/2003; 2) compare coverage of woody vegetation across the installation between 2002 and the current period; 3) conduct follow-up assessments of floristic quality on prairies identified in 2002/2003 and examine trends in condition; 4) compare current coverage of Lespedeza cuneata (Dum. Cours) G. Don (sericea lespedeza) with prior surveys; 5) update, as needed, the general vegetation map prepared in 2002/2003; 6) document locations of protected and rare animal and plant species; and 7) compare current coverage of Robinia pseudoacacia L. (black locust) with prior surveys.

Field surveys identified 120 Flint Hills Tallgrass Prairies ranging in size from 12–2,172 acres. Training Areas 47S (an extension of TA 47), 102, and 103 were assessed using floristic quality assessment for the first time in 2011. Using assessment criteria that considered landscape context, size, and condition, 78.7% of the prairies were found to be A-grade or B-grade in 2011/2012, a significant increase from 33.6% in 2002/2003. These prairies are least impacted by humans compared to the remaining 21.3%, which are C-grade or D-grade. As a consequence of receiving higher condition grades, 47 prairies (43.5%) received higher overall grades in 2011/2012 as compared to 2002/2003.

An object-based image analysis approach was tested for mapping woody encroachment across the installation. The object-based approach incorporates spectral and contextual data and groups pixels into meaningful image objects called segments. Image segments were classified as non-vegetation, grassland, woody encroachment, or forest/woodland. Multiple rule-sets were developed, tested, and compared to determine optimal variables and algorithms for separating vegetation classes. The results show that regardless of the rule-set and/or algorithm used, separating woody encroachment and herbaceous vegetation in formerly cropped areas was problematic, where woody encroachment was consistently overestimated. Woody encroachment was more accurately mapped in native grassland areas. Black locust was recorded in 39 training areas, with an estimated 142 acres infested. This did not differ significantly from the cover in 2002/2003 and the species is not a serious threat to biodiversity on the installation. Sericea lespedeza was recorded in 101 training areas, including the Multi- Purpose Range Complex, with an estimated 21,604 acres infested. The estimated infested acres increased 2.1 times since the completion of comparable surveys in 2002/2003. While the most severely impacted areas are on lands that formerly had been cultivated, intrusion into native tallgrass prairies is increasing. Sericea lespedeza occupied 19% of native prairie acres in 2011/2012, an increase from 11% in 2002/2003. Attempts to compare data from other surveys with survey data from this study were limited. The data do suggest that sericea lespedeza is increasing in distribution and abundance on FRMR in spite of efforts to control it.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE viii

Ammodramus henslowii (Audubon) [Henslow’s Sparrow] was the only state-rare animal species tracked by the Kansas Natural Heritage Inventory documented during this study. No state-rare plant species were documented. Based on data collected during the 2002/2003 study, Chenopodium pallescens Standl. [pale goosefoot] and Sporobolus heterolepis (A. Gray) A. Gray [prairie dropseed] were removed from the list of state-rare species.

The most recent version of the vegetation map of Ft. Riley (Freeman and Delisle 2004) was updated based on field surveys in 2011/2012. Revisions were minor and included adjustments due to changes in training unit boundaries, polygon boundaries within training units, and vegetation types resulting from polygon redeterminations or earlier coding errors.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE ix

Chapter 1: Introduction

1.1. Project Background

In 2010, the Kansas Biological Survey (KBS) initiated a 2-year project examining the vegetation of the roughly 101,600-acre Ft. Riley Military Reservation (FRMR), located in Geary and Riley counties, Kansas (Figure 1.1). The purpose of this study was to update information gathered by KBS in 2002 and 2003 (Freeman and Delisle 2004) and to examine general trends in vegetation patterns since 2003. The seven research goals of the project were to: 1) survey and assess the current condition of the vegetation in an 800-acre tract formerly in the Impact Area and consequently not surveyed in 2002/2003; 2) compare coverage of woody vegetation across the installation between 2002 and the current period; 3) conduct follow-up assessments of floristic quality on prairies identified in 2002/2003 and examine trends in condition; 4) compare current coverage of Lespedeza cuneata (Dum. Cours) G. Don (sericea lespedeza) with prior surveys; 5) update, as needed, the general vegetation map prepared in 2002/2003; 6) document locations of protected and rare animal and plant species; and 7) compare current coverage of Robinia pseudoacacia L. (black locust) with prior surveys. This report summarizes the methods used to achieve each objective and highlights the results of each phase of the study.

Five digital data layers were produced for this study and delivered to Fort Riley: Fort_Riley_vegetation_2012, Fort_Riley_weed_points_2012, Fort_Riley_weed_polygons_2012, Fort_Riley_Henslow’s_Sparrow_locations_2011, and RuleSet1_EntFt. These files are available for public download from the Geospatial Data portal of the Kansas Biological Survey at http://kbs.ku.edu/geodata.

1.2. Study Objectives

This report is organized into five objectives that address the seven research goals: 1) current vegetation condition (goals 1 and 5); 2) prairie assessments (goal 3); 3) weed surveys (goals 4 and 7); 4) rare species (goal 6); and 5) woody vegetation encroachment (goal 2). The section on woody vegetation is included as Appendix A.

Objective 1. Determine the current condition of vegetation on the installation. One of the products developed by KBS from the 2002/2003 surveys was a general vegetation map of the installation, excluding the Impact Area. To create this map, a new vegetation classification was developed that corresponds to existing state, regional, or national vegetation classifications. The classification includes natural vegetation as well as vegetation types resulting from major, recent, and ongoing anthropogenic disturbances (e.g., old fields, crop field, and developed land). In the present project, this map was updated as information was gathered during the course of prairie evaluations and invasive species surveys. The new map corresponds to 2011 training area boundaries and includes training units 102 and 103, which were not part of the earlier vegetation map.

Objective 2. Locate tracts of native prairie and assess their current quality. Prairies are critical reservoirs of native biological diversity, so resource managers and researchers at FRMR need

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 1 accurate information about the location and condition of individual prairie tracts. In 2002/2003, KBS conducted field surveys to delimit boundaries of individual native prairie tracts and to determine the natural community type of each tract following Lauver et al. (1999). The condition of each tract was estimated using Floristic Quality Assessment (FQA) protocols, tools for which were developed in 2002 by staff of the KBS and R. L. McGregor Herbarium, both at the University of Kansas. After completion of the FQA evaluation, prairies were ranked or graded using standard methodology employed by the Kansas Natural Heritage Inventory (KSNHI). In the present project, the condition of each native prairie tract was re-evaluated and an updated rank was assigned. The condition of each tract was compared between the 2002/2003 and 2011/2012 survey periods.

Objective 3. Determine locations and severity of infestations of two weed species of greatest concern to installation resource managers. Several aggressive, non-native plant species present a danger to the military’s training mission and to the ecological integrity of natural communities on FRMR. In 2002/2003 KBS used field surveys and standardized field methods to determine the locations and severity of infestations of populations of four weed species: musk- thistle (Carduus nutans L.), Johnson grass (Sorghum halepense (L.) Pers.), sericea lespedeza (Lespedeza cuneata), and black locust (Robinia pseudoacacia). In the present project, surveys were conducted for the latter two species. These data were compared with the data collected in 2002/2003 and with data collected by Ft. Riley staff in 2007/2008 and Dynamac Corporation in 1999.

Objective 4. Document locations of protected and rare animal and plant species. In 2002/2003, KBS used Natural Heritage Inventory methodology to document information about any rare species of plant or animal encountered during fieldwork. In the present project, no directed surveys were conducted for rare species, but data were collected ancillary to vegetation surveys.

Objective 5. Compare coverage of woody vegetation across the installation between 2002 and the current period. The encroachment of woody vegetation into grassland areas is a major management concern on FRMR. For the vegetation map produced by KBS from 2002/2003 surveys, analysts used on-screen digitizing and photo-interpretation techniques to delineate polygons of woodland and forest cover. This technique is subjective, especially for transitional land cover types such the woodland class, and the detail of digitizing varies by image analyst. In the present project, an object-oriented classification approach, which utilizes spectral characteristics in addition to spatial context or patterns to derive land cover classes, was used to increase the accuracy of discriminating between grassland, woodland, and forest land cover classes. A second goal of this objective was to develop an objective repeatable mapping approach that allows ongoing monitoring of the vegetation at the installation.

The location of FRMR and associated landscape and cultural features referenced in this report are shown in Figure 1.1. Training area numbers are shown in Figure 1.2.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 2

FIGURE 1.1 Location of the Ft. Riley Military Reservation (FRMR), and landscape and cultural features referenced in this report.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 3

FIGURE 1.2 Training areas boundaries on FRMR in 2011/2012.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 4

Chapter 2: Current Vegetation Conditions 2.1. Introduction

The vegetation of FRMR has been mapped and described in three earlier studies (Agri-Service Associates, Inc. 1985, USACE 1985, Freeman and Delisle 2004). The 1985 studies used aerial photographs and field surveys to identify the dominant vegetation types on the installation (cropland, grassland, woodland, farmstead, water, and miscellaneous), to assess vegetation conditions qualitatively, and to map field boundaries of each cover type in each of nearly 100 training areas. They provided baseline information about the locations and conditions of vegetation types on FRMR, but their usefulness was limited by their simple classification system and non- quantitative methods. Vegetation of the installation was later mapped and described (Freeman and Delisle 2004), addressing the inadequacies of the earlier work. The vegetation classification of Lauver et al. (1999) was used to classify natural and near-natural vegetation, and a new classification system was developed for cultural vegetation types (vegetation significantly modified by human activities, such as grasslands dominated by naturalized species) and cultivated land cover (e.g., agricultural land or tree plantations) (see Table 2.1, modified from Freeman and Delisle 2004 ).

As part of Objective 1 of this study, we utilized the vegetation classification and map of Freeman and Delisle (2004) as a framework to update information about the current condition of vegetation on the installation. Primarily, the work involved adjusting and refining the earlier vegetation map when field observations showed that current vegetation conditions differed from those described in Freeman and Delisle (2004).

2.2. Methods

Field surveys were conducted from May–September in 2011 and in July in 2012 by one or two 2- person crews. General survey protocols described in Freeman and Delisle (2004) were followed, although weed surveys protocols were modified in part (see Chapter 4), and a new technique was attempted to improve the accuracy and consistency of mapping communities with woody and shrubby vegetation (see Appendix A). In each training area, field crews 1) compared the 2004 vegetation map with current conditions and recorded changes, 2) conducted prairie assessments as necessary (see Chapter 3), 3) recorded locations of weed species of concern (see Chapter 4), and 4) recorded the locations of rare species tracked by the Kansas Natural Heritage Inventory (see Chapter 5).

In each training area, each prairie requiring floristic quality assessment usually was assigned to one member of a field crew, who surveyed that entire prairie. Some large prairies were surveyed by two field crew members, who were responsible for different parts of the prairie. Observations were combined upon completion of the surveys. Crew members discussed general survey routes so those on foot could focus on prairie assessments while crew members in the ATV, which could cover more ground, could focus on weed mapping. Regardless if on foot or in the ATV, crew members recorded the locations of all weed points or polygons that they encountered. In addition, differences and suspected differences between vegetation types mapped in 2004 and during the current survey

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 5 were recorded on field maps and verified. The typical 640-acre training area usually required 5–8 person-hours to complete. All training areas except 101 and the MPRC were surveyed in 2011; remaining areas were surveyed in July 2012.

A general comparison of 2002/2003 and 2011/2012 vegetation patterns was carried out within the limits of the data, but a host of problems prevented quantitative field-by-field and cover class-by- cover class comparisons (see discussion).

As part of Objective 5 of this study, the Kansas Applied Remote Sensing Program utilized an object-based image analysis approach to map encroaching woody vegetation. The approach incorporates spectral and contextual data, and groups pixels into image segments. Image segments were classified into four categories: non-vegetation, grassland, woody encroachment, forest/woodland. A multi-level segmentation approach was used where vegetated and non- vegetated segments were classified first, and then vegetated segments were further segmented into finer vegetation categories. Multiple rule-sets were developed, tested, and compared to determine optimal variables and algorithms for separating vegetation classes. A detailed summary of the procedure and results are presented in Appendix A.

2.3. Results and Discussion

Figure 2.1 shows the vegetation map updated from field surveys carried out in 2011/2012. Revisions were minor and included adjustments due to changes in training unit boundaries, slight modifications in polygon boundaries within training units, and changes in vegetation types due to redeterminations of polygons or the discovery of coding errors in the 2002/2003 map. A rigorous comparison of the 2002/2003 and 2011/2012 vegetation maps was deemed unnecessary because the changes were so minor.

Undoubtedly, the most dramatic change in vegetation on FRMR was the increase in distribution and abundance of sericea lespedeza, which is discussed in detail in Chapter 4. As was noted during earlier surveys (Freeman and Delisle 2004), this aggressive, non-native species was a serious problem in many Semi-natural/Altered vegetation communities. The concern then was that, if unchecked, sericea lespedeza eventually could jeopardize the ecological conditions of Natural/Near-natural herbaceous communities on and off the installation. We have no detailed information about the status of the species on lands off of but immediately adjacent to the installation. However, our perception is that the overall severity of the infestation of sericea lespedeza on FRMR has increased since 2003 in spite of determined efforts to control its spread, and its increased occurrence on high quality prairies on the installation may signal a trajectory that ultimately could lead to declines in the ecological condition and economic value of these prairies.

2.4. Conclusions

The most recent version of the vegetation map of FRMR (Freeman and Delisle 2004) was updated based on field surveys in 2011/2012. Revisions were minor and included adjustments due to changes in training unit boundaries, polygon boundaries within training units, and vegetation types resulting from polygon redeterminations or earlier coding errors. The most significant vegetation change observed from 2002/2003 to the present is the increase in coverage of sericea lespedeza, which has increased in occurrence on high quality prairies on the installation.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 6

TABLE 2.1 Classification of Natural/Semi-natural and Cultural vegetation types known to occur on FRMR. Only common names are used in the classification. Detailed information about each type is provided in Freeman and Delisle (2004) in the following appendices: B and C for Natural/Near-natural types; D for Semi-natural/Altered types; and E for Planted/Cultivated types. I. Natural/Semi-natural Vegetation a. Natural/Near-natural Vegetation i. Forest Communities 1. Ash-Elm-Hackberry Floodplain Forest 2. Cottonwood-Sycamore Floodplain Forest 3. Cottonwood-Black willow Floodplain Forest ii. Woodland Communities 4. Mixed oak Ravine Woodland iii. Herbaceous Communities 5. Flint Hills Tallgrass Prairie 6. Sand Prairie iv. Sparse Vegetation Communities 7. Limestone Butte Sparse Vegetation 8. Riverine Sand Flats-Bars Sparse Vegetation b. Semi-natural/Altered i. Ruderal Vegetation 9. Cropland-Abandoned 10. Brome Field 11. Ruderal-Mixed ii. Invasive Vegetation 12. Sericea lespedeza Herbaceous Vegetation 13. Smooth brome/Japanese brome Herbaceous Vegetation iii. Modified/Managed Vegetation 14. Overgrazed Tallgrass Prairie 15. Woodland-Brushy II. Planted/Cultivated Vegetation 16. Fire Break 17. Food Plot 18. Cultivated Field 19. Tree Plantation 20. Hedgerow/Windbreak 21. Lawn

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 7

FIGURE 2.1 Vegetation of FRMR based on surveys conducted in 2011/2012.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 8

Chapter 3: Prairie Assessments

3.1. Introduction

Prairies are critical reservoirs of biological diversity in Kansas and the Great Plains, providing habitat for a majority of the state’s native organisms. As land use and land management practices change, many prairie tracts are threatened due to fragmentation, isolation, and degradation. Issues of particular concern to the maintenance of biodiversity of prairies are 1) exotic species, especially plants, that invade prairies and out-compete native species, 2) encroachment of woody vegetation, and 3) military training activities that result in direct or indirect damage. Before these factors can be mitigated effectively, accurate baseline data about the location, quantity, and quality of prairie resources are needed.

Freeman and Delisle (2004) located and estimated the ecological conditions of 116 Flint Hills Tallgrass Prairies on FRMR during the 2002/2003 field seasons. None of the prairies met the 1000- acre minimum size standard used for this community type in ecoregional planning (see discussion below). However, assessment criteria that considered landscape context, size, and condition found that 34% of the prairies ranked as A-grade or B-grade, indicating they are least impacted by humans. The remaining 66% were C-grade or D-grade, most of which were small, isolated, and moderately to severely impacted by past or ongoing human activities. The largest prairies, which also generally graded the highest, were concentrated in the south, east, and northwest parts of the installation. Prairies generally were most abundant in those parts of the installation with the greatest topographic relief. Areas with comparatively lower relief, especially in the central part of the FRMR, generally had a much higher incidence of past cultivation.

In fulfillment of Objective 2 of this study, KBS staff revisited and reassessed the conditions of prairies on FRMR during the 2011/2012 field seasons. Generally speaking, we followed the same assessment protocols described in Freeman and Delisle (2004). Though described in detail in Freeman and Delisle (2004), we summarize those protocols below to aid the reader.

Ecoregional conservation planning is a tool that uses principles of conservation biology and ecology to identify priority areas for conservation (Groves et al. 2002). Although normally employed at the ecoregional scale (106 acres), this approach also has utility at the local scale (<101–104 acres).

The conservation planning framework used in ecoregional conservation planning has seven primary steps (Groves et al. 2002). First, conservation targets are identified – the species and communities that are most significant in the area of interest. Second, information about these targets is gathered, and data gaps are identified and filled though field surveys, rapid ecological assessment, or other approaches. Third, conservation goals are established. The quality and quantity of target species and community occurrences needed to protect biodiversity in the area of interest are determined. Fourth, existing conservation areas are identified. Fifth, the viability of conservation targets is estimated. Size, condition, and landscape context are the primary attributes considered. Sixth, a portfolio of conservation sites is assembled. Site selection criteria are developed and employed. Seventh, priority conservation areas are identified. An explanation of each step follows.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 9

1. Identify conservation targets – Conservation targets may include ecosystems and communities, imperiled, endangered, endemic, or keystone species, and abiotic factors that help maintain the structure and function of ecosystems and natural communities.

2. Collect information and identify gaps – Information about conservation targets may be obtained from a variety of sources, including existing data sources or expert workshops. Often, new data are acquired from rapid ecological assessments, site-specific surveys, or by remote-sensing methods.

3. Establish conservation goals – Normally, conservation goals are established by considering the representation and quality of the conservation targets within the planning area. This phase of the planning process involves asking two questions: How much or how many of each target should be conserved, and how should the targets be distributed across the planning region?

4. Assess existing conservation areas – An important early step in conservation planning is to determine which biological targets already have adequate protection within existing conservation areas. As employed in most planning efforts, existing management areas are identified, conservation targets within them are enumerated, and the level of protection afforded each target is assessed.

5. Evaluate viability of conservation targets – Three key factors are evaluated in this process: landscape context, size, and condition (Figure 3.1). Estimates of each factor for each conservation target are entered into a series of evaluation matrices to determine which occurrences have the highest viability. Normally, landscape context and size are weighted more heavily than is condition. The rationale is that landscape context and size cannot increase, or can do so only slightly with time, whereas condition is a more variable attribute and can be increased fairly quickly with appropriate management inputs. Also, the assessed condition of a prairie remnant may vary with season, observer, and management or environmental conditions. After landscape context and size have been evaluated (Table 3.1), results from that matrix are entered into a landscape context/size × condition matrix (Table 3.2). The results from that matrix then can be analyzed spatially in GIS to identify sites of highest conservation priority.

Landscape Context/ Landscape ContextSize Landscape Context Size Matrix × GIS Analysis (acres) × Size Matrix Condition Matrix Condition FIGURE 3.1 Summary of evaluation process for estimating viability of conservation targets.

A. Landscape Context – Landscape context refers to the general condition of the landscape in which a site occurs, considering such issues as disturbance regimes, fragmentation, topography, and biological diversity. Landscape context is ranked A–D. Generally speaking, A-grade landscapes have little if any impact from land conversion and are dominated by natural communities. Natural processes, and species interactions and migrations can occur across all natural communities and experience no complete barriers. Surrounding vegetation is >80% natural. B-grade landscapes have experienced some land conversion, but natural communities remain well-connected. Natural processes, and species interactions and migrations can occur across many natural communities and experience few barriers. Surrounding vegetation is 50–80% natural. C-grade landscapes are

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 10 fragmented by cultural land, including cropland or developed areas. Barriers severely affect many natural processes, species interactions, and migrations, and many species are unable to maintain viable populations. Surrounding vegetation is 20–50% natural. At the low end of the spectrum, D- grade landscapes are surrounded almost entirely by cultural land. Natural processes and species migrations are severely compromised and cannot occur at natural scales. Only a subset of the historic biological diversity is viable within natural communities.

TABLE 3.1 Generalized evaluation matrix for landscape context rating × size grade. The grade for a given site is determined by estimating the landscape context grade, the size grade, and noting the grade in the cell in which the column and row of those grades, respectively, intersect.

Landscape Context Grade A B C D A A A B B B B B B C Size Grade C B C C C D C C D D

TABLE 3.2 Generalized evaluation matrix for landscape context/size rating × condition grade. The grade for a given site is determined by estimating the landscape context/size grade (from Table 3.1), the condition grade, and noting the grade in the cell in which the column and row of those grades, respectively, intersect.

Landscape Context/Size Grade A B C D A A A B C Condition B A B B C Grade C B C C D D C D D D

B. Size – Determining the size of a natural community may appear straight-forward, but several issues complicate this process: patch size and minimum distance separating two occurrences.

Patch size denotes the size and landscape position of a natural community (Lauver et al. 1999). Four patch types are recognized: matrix, large-patch, small-patch, and linear. Matrix communities occur on the dominant landforms in an ecoregion and form extensive and often contiguous cover, usually >1,000 acres. Large-patch communities generally occur on subdominant landform features and form large but interrupted cover, usually 20–1,000 acres. Small-patch communities occur on specialized landforms and microhabitats, and generally are <20 acres. Linear communities are long, narrow communities usually associated with riverine features.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 11

Size standards have been established for many natural communities to distinguish viable from non- viable occurrences and, for viable occurrences, to rank them (A–D, with A being the best and D being the worst). During ecoregional planning, each community occurrence must meet the minimum size set for its type to be considered for conservation purposes. For example, for Flint Hills Tallgrass Prairie, a matrix community type, occurrences <1,000 acres usually are not considered viable (able to support ecosystem functions necessary to maintain high levels of native biodiversity for more than 100 years). Unless there are mitigating factors, such as high restoration potential or other, nearby occurrences to which smaller occurrences might be connected, substandard occurrences usually are excluded from planning.

A second factor complicating the size issue is how far apart two occurrences of the same community type can be before they are considered distinct occurrences. Several evaluation guidelines are available to assist in making this determination for terrestrial natural communities. Basically, two tracts are treated as distinct if they are separated by: 1) a substantial barrier to natural processes and/or to native species, such as a busy highway, developed area, or large body of water; 2) cultural vegetation that limits connection of patches; large areas of FRMR formerly were cultivated but have undergone more than 40 years of succession and, while usually classified as ruderal community types, these areas often are dominated by native species within a patchwork of natural/ruderal types, so a substantial amount of species migration is possible; 3) a different community type coverage >0.5 mile wide if the communities frequently do not occur in a mosaic, or 1–2 miles wide if frequently in a mosaic; 4) a tract subjected to management that is significantly different from that employed on them; or 5) a major break or change in ecological land unit.

C. Condition – Condition refers to impact that human disturbance has had on a site. Condition can be estimated by any of several available methods. Most Natural Heritage programs use subjective field assessments, which are based on estimates of native species richness, abundance of exotic species, and ecological processes. As with landscape context, condition may be ranked from A–D, with A being the best (least affected by human disturbance) and D being the worst (severely affected by human disturbance).

Floristic Quality Assessment (FQA) is a standardized tool used to estimate the floristic quality of a natural area based on the vascular plants growing there (Freeman and Morse 2002). By extension, it can be used to assess the overall ecological quality of a site. Ecologists, botanists, environmental professionals, and land managers use FQA to establish baseline assessments, to conduct long-term monitoring, and to assess restoration progress in a variety of ecological settings (Herman et al. 1997, Taft et al. 1997). Developed in the 1970s (Wilhelm 1977, Swink and Wilhelm 1979), the method has been refined from its original form (Wilhelm and Ladd 1988, Taft et al. 1997, Rooney and Rogers 2002) and now is in use or development in numerous states and provinces in the U.S. and Canada (Taft et al. 1997).

The method was developed to avoid subjective measures of natural community quality, such as “high” or “low”. Some elements of FQA still are subjective, but the method has clear advantages over other evaluation tools, including repeatability and ease of application. Ideally, FQA should be used with other content-based and context-based measures (sensu Rooney and Rogers 2002) to estimate the integrity of native plant communities (Taft et al. 1997).

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 12

The FQA method is based on calculating an average coefficient of conservatism (C) and a floristic quality index (FQI) for a site. It may be used to compare several sites supporting the same community type (e.g., several Flint Hills Tallgrass Prairies) but should not be used to compare different community types (Rooney and Rogers 2002). A coefficient of conservatism is an integer from 0–10 that is assigned to each native plant species in a given geographic region – often a state or province. Naturally occurring hybrids and infraspecific taxa usually are not assigned coefficients.

Coefficients of conservatism express two basic ecological tenets: plants differ in their tolerance of the type, frequency, and amplitude of anthropogenic disturbance, and plants vary in their fidelity to remnant natural plant communities (Taft et al. 1997). As employed in FQA, these two principles exhibit an inverse relationship: the lower a species’ tolerance of human-mediated disturbance, the higher its likelihood of occurring only in a natural plant community. Low coefficient values (0–3) denote taxa often found in highly disturbed habitats and without a strong affinity for natural communities. High coefficient values (7–10) denote species that tolerate only limited disturbance and usually are found in natural communities. With these principles as a guide, the C value applied to each species represents a relative rank based on observed behavior and patterns of occurrence in Kansas natural communities. Non-native species are not assigned coefficients because they were not part of the pre-settlement landscape. They do have an effect on FQA, however, and they may be incorporated in the assessment process.

The FQA process begins with a thorough inventory of vascular plants at a site of interest. The checklist then is used to calculate a floristic quality index (FQI) for the site. Two approaches have been proposed for calculating the FQI. In its original form (Wilhelm 1977, Taft et al. 1997), a mean C value (mean C) is calculated first. The mean C value for a site is the arithmetic mean of the coefficients of all native vascular plants occurring on the entire site (mean C = ΣC/N), without regard to dominance or frequency. Non-native taxa are excluded from the calculation of mean C. The FQI is the mean C multiplied by the square root of the total number of taxa (√N) inventoried on the site (FQI = mean C × √N). Separate calculations may be made using N = all taxa (native and non-native) and N = native taxa only (see analysis and discussion in Taft et al. 1997). The basic formula for FQI combines the conservatism of the taxa with a measure of the taxon richness of the site. By multiplying by √N instead of N, the formula reduces the effect of the size of the site (larger sites tend to have a larger total number of species). If the sampling method involves transects or quadrats a mean C and FQI can be calculated for each sample.

Rooney and Rogers (2002) have shown that a modified FQI, which is simply the mean C value for the site (mean C = ΣC/N), has greater power in reflecting the degree of habitat degradation. They argue that because the original FQI formula combines two independent measurements, species richness and the C values of species in the survey, identical FQI scores can be obtained for two natural communities that differ markedly in their quality. For example, a high FQI score could be due to either a large number of common species present at the site, each with low C values, or relatively few rare species at the site, each with high C values. Their approach is computationally simpler than the original FQI, and it is not strongly affected by sample size or species richness.

6. Assemble portfolio of conservation areas – Following assessment, sites are assembled into a portfolio that best meets the conservation goals established for the targets. The portfolio helps identify where those goals can be met, and where restoration activities may be necessary to meet conservation goals.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 13

7. Identify priority conservation areas – The final step of the conservation planning process involves identifying conservation priorities based on issues that may influence long-term strategies, including existing levels of protection, conservation value, feasibility, and other factors.

3.2. Methods

Flint Hills Tallgrass Prairie tracts identified by Freeman and Delisle (2004) were the conservation targets for this phase of the study. As in our earlier study, we did not assume that any particular number or distribution of Flint Hills Tallgrass Prairie occurrences would be best in meeting management or conservation goals on the installation, so explicit goals were not set. All occurrences of Flint Hills Tallgrass Prairie were considered to have equal levels of protection on FRMR despite the fact that some are buffered more from potentially detrimental influences than are others. None of the prairies on the installation is protected per se, and training activities and on-the-ground management decisions quickly can change conditions of individual sites.

Evaluation of long-term viability of prairies was the primary emphasis of our study. Landscape condition on FRMR varies from B (50–80% natural vegetation; mostly in the west and south) to C (20–50% natural vegetation; mostly in the central part) but was considered fairly uniform across the installation. A landscape context grade of C was assumed for all prairies in our analysis, so we did not consider it in the landscape context grade × size grade matrix. Size grades, unweighted by landscape context grades, were used directly in the landscape condition/size grade × condition grade matrix to prevent grade compression and to provide better spread and ranking of sites.

Prairie size was determined from the digital vegetation coverage. Following strict assessment guidelines, only occurrences of Flint Hills Tallgrass Prairie (a matrix community type) >1,000 acres should be included in ecoregional assessments (Freeman and Delisle 2004). Smaller occurrences are assumed to have low, long-term viability or not to be viable. Few tracts of Flint Hills Tallgrass Prairie on FRMR meet the minimum size requirements, but because resource managers must work with what is on the installation, the 1,000-acre cut-off was not used to filter occurrences. All prairies, regardless of size, were included in our analysis, with size used to sort occurrences in the evaluation matrix. To facilitate comparisons with earlier study (Freeman and Delisle 2004), the following size classes were used to assign grades: D = 0–200 acres; C = 201–400 acres; B = 401– 600 acres; and A = >601 acres.

Again, to facilitate comparison of data, we used the same critera as described in Freeman and Delisle (2004) for determining whether prairies were distinct or not. Application of those criteria meant that most training areas were treated functionally as their own management areas. Most are surrounded by perimeter roads or trails that slow, but do not prevent, dispersal of plant propagules. Some roads are robust fire guards and greatly reduce the chance of fire spreading from one unit to another. For these and other reasons, we considered prairies in different training areas to be separate even though two tracts might be separated by the width of a gravel road. Within a training area, any two prairies sharing part of a boundary (point or line) were combined as a single occurrence. Furthermore, any two prairies separated by 0.5 mi or less of any herbaceous community, natural or altered, were combined for purposes of evaluation. Prairies separated by more than 0.25 miles by a non-herbaceous community type (e.g., forest or woodland) were considered distinct. Some small, isolated prairies (mostly <10 acres) were excluded from our analysis.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 14

For Floristic Quality Assessment, a list of all native and naturalized species observed in each prairie was compiled using a master field checklist designed specifically for FQA. Presence data for each prairie were entered into a customized Excel application that calculated and summarized FQA metrics, including species richness (all taxa and native taxa only), percent of non-native taxa, mean conservatism (all taxa and native taxa only), floristic quality index (all taxa and native taxa only), and number of state-rare taxa (S1 and S2).

Floristic data were gathered for all prairies from May–September 2011 except Training Areas 101 and the MPRC; the latter units were surveyed in July 2012. Floristic quality assessments were conducted at 120 sites in total. Condition grades were assigned based on native species only and using the same index classes as in Freeman and Delisle (2004): D = 15.40–23.80; C = 23.81–32.20; B = 32.21–40.60; and A = >40.61.

After final assessment through the evaluation matrices, each prairie or prairie complex was assigned a final grade (A–D, with A being the best and D being the worst) that summarized all evaluation factors: landscape condition, size, and condition. Finally, site grades were added to the GIS coverage as attributes so the data could be summarized spatially.

3.3. Results and Discussion

3.3.1. Prairie Assessment Results for 2011/2012

Field survey data are not included in this report but are available upon request from the authors. Electronic copies of all files containing FQA species lists and metrics, which are based on field survey data, have been submitted to the FRMR Department of Public Works. Floristic quality assessment metrics derived from field surveys are summarized in Table 3.3. Grades assigned to each prairie estimating landscape context, size, landscape context × size, condition, and landscape context/size × condition are summarized in Table 3.4.

Histograms of size data and floristic quality indices for all 120 Flint Hills Tallgrass Prairies assessed in this study are shown in Figures 3.2 and 3.3, respectively. Size vs. floristic quality index data are summarized in Figure 3.4. Prairie locations and grades are shown in Figure 3.6.

Prairie size ranged from 12–2,172 acres and mean size was 239.08 acres (Figure 3.2). However, the distribution of sizes is positively skewed (skewness = 4.232); nearly 70% of all prairies assessed were smaller than the mean.

Floristic quality index values ranged from 16.13–45.62 and were roughly normally distributed (Figure 3.3). The scatterplot of size vs. floristic quality index for all Flint Hills Tallgrass Prairies (Figure 3.4) shows the relationship between these two variables. Floristic quality index increased with size but approached an asymptote when the index reached the low 40s. Size and the index showed a moderately positive relationship (r = 0.503, Spearman’s rho test for nonparametric correlation p = 0.01).

Vertical lines on Figure 3.4 indicate breaks between size classes (D = 0–200 acres, C = 201–400 acres, B = 401–600 acres, and A = >601 acres). Horizontal lines approximate breaks between index classes (D = 15.40–23.80, C = 23.81–32.20, B = 32.21–40.60, and A = >40.61). Grades for sites shown in Figure 3.4 were determined by applying the criteria from Table 3.2. Floristic quality index VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 15 grades for all 120 prairies were as follows: D = 5, C = 30, B = 64, and A = 21. Overall grades for all 120 prairies were as follows: D = 29, C = 44, B = 34, and A = 13.

In addition to floristic quality index, we examined the relationship between prairie size and percent of non-native species. Non-native species richness often is an index of disturbance, and it might be assumed that the percent of non-native species will increase as prairie size increases, up to a point. However, the data do not support that hypothesis. Figure 3.5 shows virtually no relationship between prairie size and percent of non-native species (Pearson correlation r = -0.056; p = 0.543).

Figure 3.6 shows the locations and grades of prairies documented during this study. On FRMR, as in many other parts of the state, level or nearly level ground is more likely to be cultivated or developed, while prairies occur most frequently in areas with comparatively greater relief. This is evident from Figure 3.6. Extensive tracts of native Flint Hills Tallgrass Prairie occur on uplands and on upper slopes of valleys along the south, east, and northeast sides of the installation (Training Blocks C, F, I, M, and training units south of Vinton School Road). In these areas, tributaries to the Republican River, Kansas River, and Wildcat Creek have eroded short, steep-sided valleys. The bottoms and lower slopes of these valleys usually are dominated by Ash-Elm-Hackberry Floodplain Forest or Mixed oak Ravine Woodland, but fire and haying have controlled the spread of woody vegetation onto the upper slopes and uplands in most places. The second area of concentration of prairies on FRMR is in the northwest part (Training Blocks A, D, G, H, J, K, N, and the MPRC). Again, most large tracts are associated with the upper reaches of tributaries to the Republican River, namely Rush Creek, Farnum Creek, Madison Creek, Dry Creek, and several other unnamed tributaries.

The divide between watersheds draining westward to the Republican River and eastward to Wildcat Creek or southeastward to the Kansas River is approximated by Old Highway 77, which extends from near the northwest corner of Custer Hill north to the town of Riley. Along this divide is a 2– 4-mile wide band of fairly level ground, along which much of the land formerly was cultivated, and where native prairies are relatively uncommon. Not coincidentally, some of the worst infestations of Lespedeza cuneata on the installation are found here (see Chapter 4).

In our analysis, we closely followed all criteria for determining when two prairies should be treated as one or two occurrences. One exception was Training Area 65, where we conducted FQA on five separate prairies in the TA rather than a single assessment for all five tracts as criteria would dictate. This was done to obtain FQA data from several sites with similar management and topographic features but which differed in size. The prairies in TA65 were suitable for this purpose. However, had we followed our criteria, as was done in adjacent training areas, all five prairies in TA65 would have been part of a single, large, B-grade complex.

A-grade and B-grade prairies are concentrated in the south, east, and northwest parts of the installation, where prairie is most plentiful. Sites in Training Blocks H and K generally exhibit less relief than do prairies with similar grades in other parts of FRMR. C-grade and D-grade prairies are concentrated along the divide through the central part of the installation. They also occur sporadically in other parts of the installation.

3.3.2. Comparison of 2011/2012 and 2002/2003 Prairie Assessments

During the 2011/2012 field seasons, 120 prairies were surveyed on FRMR; 116 were surveyed in VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 16

2002/2003. Several factors affected the number and sizes of prairies assessed during the two survey periods, including changes in training area boundaries and the addition of new survey areas. In 2011/2012, Training Areas 47S (an addition to TA 47), 102, and 103 were new units added to our survey. Also, access to the entire MPRC (not all of which could be surveyed in 2002/2003) and application of prairie delimitation criteria caused us to treat all prairie in the MPRC as one large occurrence. Areas of the MPRC surveyed in 2002/2003 were treated as five separate prairies. Adjustments in training area boundaries also affected the size of some prairies. However, size changes due to within-unit changes in land use, adjustments to polygon boundaries, and corrections of earlier mapping errors were minor. Excluding prairies unique to one survey period, the correlation between sizes for the two survey periods is significant (Spearman’s rho = 0.921; p = 0.01).

The correlation between floristic quality indices obtained in 2011/2012 and 2002/2003 (Figure 3.7) is positive and significant (r2 = 0.541; p = 0.01). While 41 prairies (34.2%) received the same condition grade in 2011/2012 as in 2002/2003, significantly more (54 prairies; 49.1%) received higher grades in 2011/2012 as compared to 2002/2003 (Figure 3.8). Only 15 prairies (12.5%) received lower grades in 2011/2012 as compared to 2002/2003.

Field protocols and analysis procedures are not believed to have contributed to the observed increase in floristic quality index grades among so many prairies in 2011/2012. Field assessment protocols were basically identical for both survey periods. The membership of the field crews conducting FQA differed between the two survey periods, but all crew members had extensive floristic experience in the Flint Hills, and field identifications routinely were checked by the junior author, who had the most botanical field experience on FRMR. Consequently, we do not believe that observer bias was a significant factor. Also, as described earlier, data analysis procedures (coefficients of conservatism, FQI calculations, grade criteria and assignment methods) were identical, so assessment methodology was not a factor.

Environmental conditions during the two survey periods could account for some of the differences, but isolating climate influences would be difficult because 1) assessments during each survey period had to be carried out over a 2-year period, 2) there was considerable within-year and between-year variation in conditions in both survey periods, and 3) surveys were conducted when training areas were accessible, so timing of assessment could not be controlled across the population of prairies assessed.

The observed increases may reflect actual improvement in prairie quality between the two survey periods. Prairie condition can change quickly depending on management regimes. Widespread and periodic use of prescribed burns, control of the timing and frequency of haying, efforts to control woody vegetation encroachment, and measures directed at minimizing the impacts from military training on prairies may be having a beneficial effect on many prairies on the installation. Potentially countering this however, is the apparent spread and increase in abundance of sericea lespedeza across the installation (see Chapter 4). Compared to earlier estimates (Freeman and Delisle 2004), this aggressive weed has become much more prevalent on native prairies across FRMR, and it is not clear what impact this spread, and efforts to control it, ultimately will have on vegetation conditions.

Comparing prairies surveyed both in 2011/2012 and 2002/2003, 74 received the same overall grade (Figure 3.9). However, 31 prairies (28.7%) received a higher grade in 2011/2012 as compared to 2002/2003. Only three prairies (2.8%) received a lower grade in 2011/2012 compared to VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 17

2002/2003. The increase in overall grades is a function of the large number of prairies receiving higher condition grades – not changes in landscape context or prairie sizes. As a group, prairies with higher overall grades were represented roughly proportional to the number of prairies in the size classes on FRMR and were more-or-less uniformly distributed across the installation.

As mentioned above, Training Areas 102 and 103, and a new addition to Training Area 47 that we have named 47S in this report, were assessed using FQA for the first time in 2011. Training Area 47S is a gently southeast-sloping, 12-acre tract located along the west-central part of the installation. It is designated for youth/handicapped hunting. The area was in the process of being cleared of woody vegetation, which was particularly dense at the east and west ends, when it was surveyed in 2011. Consequently it was rather severely disturbed. TA 47S supports small patches of native prairie vegetation but had among the lowest floristic quality indices of any of the areas surveyed in 2011/2012.

Training Areas 101 and 102, 109 acres and 458 acres respectively, are located in the southwest corner of the Impact Area. The southern edge of the floodplain of Threemile Creek separates TA 102 (to the south) from TA 103. The training areas exhibit about 40 meters of relief, with floodplains and deep draws and ravines dominated by forest communities, and slopes and level uplands dominated by tallgrass prairie. Flint Hills Tallgrass Prairie in TA 102 was graded as A (FQI = 40.77), but because of its small size received an overall grade of C. TA 103 also was graded as A (FQI = 43.69) and received an overall grade of A. The high quality of prairies in both training areas probably is due to the former inclusion of these training areas within the boundary of the Impact Area, which restricted training in them, their relief, and frequent fires that limited woody encroachment on slopes and uplands.

3.4. Conclusions

One hundred twenty Flint Hills Tallgrass Prairies on FRMR were assessed for floristic quality in 2011/2012 following protocols used in a similar study in 2002/2003, when 116 prairies were assessed. The two survey periods had 108 prairie sites in common. Training Areas 47S, 102, and 103 were assessed using FQA for the first time in 2011. TA 47S received an overall grade of D, but TAs 102 and 103 each received an overall grade of A.

Prairie condition grades, based on floristic quality indices, for prairies assessed in 2011/2012 were: D = 5, C = 30, B = 64, and A = 21. Overall grades for all 120 prairies were: D = 29, C = 44, B = 34, and A = 13. None of the prairies meets the 1000-acre minimum size standard used for this community type in ecoregional planning.

Using assessment criteria that considered landscape context, size, and condition, we found that 78.7% of the prairies were A-grade or B-grade in 2011/2012, a significant increase from 33.6% in 2002/2003. These prairies are least impacted by humans compared to the remaining 21.3%, which are C-grade or D-grade. As a consequence of receiving higher condition grades, 47 prairies (43.5%) received higher overall grades in 2011/2012 as compared to 2002/2003.

When individual prairie condition grades from 2011/2012 and 2002/2003 were compared, in 2011/2012, 52 prairies (48.1%) received higher grades, 41 prairies (34.2%) received the same grade, and 15 prairies (12.5%) received lower grades. When individual overall prairie grades in 2011/2012 and 2002/2003 were compared, in 2011/2012, 31 prairies (28.7%) received a higher grade, 74 VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 18 received the same grade, and three (2.8%) received a lower grade. Grade changes in the prairies compared were a function of condition changes, not changes in landscape context or prairie sizes. Prairies with higher overall grades are more-or-less uniformly distributed across FRMR.

The reason for the increase in number of prairies with higher condition grades is not clear, but it may reflect actual improvement in floristic quality across the installation resulting from management strategies. Observer bias, field protocols, and data assessment procedures are not believed to have significantly affected the results. The influence of climate on the results is potentially complicated and has not been explored.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 19

TABLE 3.3 Floristic quality assessment data for all sites evaluated on FRMR in 2011/2012. Column codes are: Site = training area+field number; Ra = species richness, all taxa; Rn = species richness, native taxa only; %N = percent of all taxa at site that are non-native; Ca = mean conservatism, all taxa; FQIa = floristic quality index, all taxa; Cn = mean conservatism, native taxa only; FQIn = floristic quality index, native taxa only; and Size = area of site (acres). See text for definitions of metrics. Assessments in sites marked with an asterisk cannot be compared against 2002/2003 surveys due to significant changes in training area boundaries that affected how fields were partitioned.

Site Ra Rn %N Ca Cn FQIa FQIn Size TA3-1 102 93 8.82 3.67 4.02 37.03 38.78 220 TA4-1 133 119 10.53 3.44 3.84 39.63 41.89 91 TA5-1 137 125 8.76 3.66 4.02 42.89 44.90 98 TA6-1 111 97 12.61 2.82 3.23 29.71 31.78 145 TA7-1 132 117 11.36 3.23 3.64 37.08 39.38 254 TA8-1 148 138 6.76 3.33 3.57 40.52 41.97 111 TA9-1 80 73 8.75 3.75 4.11 33.54 35.11 36 TA10-1 120 112 6.67 3.32 3.55 36.33 37.61 120 TA11-1 118 110 6.78 3.53 3.79 38.39 39.76 48 TA11-2 89 82 7.87 3.65 3.96 34.45 35.89 171 TA12-1 163 145 11.04 3.01 3.39 38.46 40.78 457 TA13-1 145 136 6.21 3.46 3.69 41.69 43.05 395 TA14-1 157 148 5.73 3.34 3.54 41.82 43.07 603 TA15-1 140 115 17.86 2.56 3.12 30.34 33.48 93 TA16-1 129 104 19.38 2.15 2.66 24.39 27.16 118 TA17-1 139 124 10.79 3.49 3.91 41.14 43.55 412 TA20-1 160 140 12.5 2.99 3.42 37.87 40.48 249 TA21-1 131 121 7.63 3.44 3.73 39.40 41.00 125 TA22-1 162 148 8.64 3.43 3.75 43.60 45.62 292 TA23-1 120 110 8.33 3.48 3.79 38.07 39.76 321 TA24-1 107 95 11.21 2.98 3.36 30.84 32.73 488 TA25-1 106 98 7.55 3.28 3.55 33.8 35.15 348 TA26-1 75 72 4.00 3.79 3.94 32.79 33.47 54 TA26-2 85 79 7.06 3.53 3.80 32.54 33.75 51 TA27-1 154 140 9.09 3.32 3.65 41.18 43.19 367 TA29-1 159 143 10.06 3.32 3.69 41.87 44.15 503 TA30-1 133 123 7.52 3.61 3.90 41.62 43.28 523 TA31-1 164 149 9.15 3.13 3.45 40.14 42.11 276 TA32-1 120 112 6.67 3.72 3.98 40.71 42.14 465 TA33-1 114 109 4.39 3.77 3.94 40.27 41.19 110 TA34-1 74 61 17.57 2.50 3.03 21.51 23.69 29 TA35-1 132 115 12.88 2.83 3.24 32.47 34.78 212 TA35-2 94 88 6.38 3.49 3.73 33.83 34.96 28 TA35-3 131 117 10.69 3.34 3.74 38.27 40.49 91 TA36-1 117 102 12.82 2.38 2.73 25.70 27.53 366 TA37-1 136 122 10.29 3.18 3.55 37.13 39.20 337 TA38-1 103 92 10.68 2.45 2.74 24.83 26.27 108 TA39-1 121 111 8.26 2.99 3.26 32.91 34.36 267 TA40-1* 94 83 117.00 2.26 2.55 21.87 23.27 45 TA41-1 87 79 9.20 2.60 2.86 24.23 25.43 150 TA42-1* 80 70 12.50 2.53 2.89 22.58 24.14 118 TA43-1 81 73 9.88 2.83 3.14 25.44 26.80 139

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 20

Site Ra Rn %N Ca Cn FQIa FQIn Size TA44-1 117 104 11.11 2.99 3.37 32.36 34.32 76 TA45-1 94 84 10.64 2.69 3.01 26.09 27.60 116 TA46-1 87 80 8.05 2.72 2.96 25.41 26.5 292 TA47-1 113 98 13.27 3.07 3.54 32.64 35.05 64 TA47-2* 98 87 3.26 3.26 3.67 32.22 34.20 111 TA47S 72 63 12.5 1.78 2.03 15.08 16.13 12 TA48-1 134 116 13.43 3.09 3.57 35.76 38.44 463 TA49-1 95 81 14.74 2.38 2.79 23.19 25.11 45 TA50-1 146 132 9.59 3.16 3.5 38.24 40.21 286 TA51-1 140 124 11.43 2.86 3.23 33.81 35.92 249 TA52-1 101 91 9.90 2.76 3.07 27.76 29.25 65 TA53-1 114 99 13.16 3.21 3.70 34.28 36.78 108 TA53-2 102 88 13.73 2.77 3.22 28.02 30.17 32 TA54-1 134 122 8.96 2.99 3.29 34.64 36.30 370 TA55-1 98 88 10.20 2.68 2.99 26.57 28.04 206 TA56-1 113 98 13.27 2.68 3.09 28.50 30.61 83 TA57-1 131 121 7.63 3.02 3.26 34.51 35.91 329 TA58-1 137 126 8.03 3.10 3.37 36.31 37.86 711 TA59-1 120 111 7.50 3.30 3.57 36.15 37.59 505 TA60-1 102 90 11.76 3.58 4.06 36.14 38.47 197 TA60-2 96 85 11.46 3.36 3.80 32.97 35.03 250 TA61-1 111 105 5.41 3.39 3.58 35.69 36.69 789 TA63-1 121 111 8.26 3.48 3.79 38.27 39.96 611 TA64-1 140 123 12.14 3.09 3.52 36.60 39.04 349 TA65-1 80 69 13.75 2.59 3.00 23.14 24.92 79 TA65-2 87 78 10.34 2.93 3.27 27.34 28.87 56 TA65-3 112 101 9.82 2.84 3.15 30.05 31.64 155 TA65-4 99 84 15.15 3.39 4.00 33.77 36.66 29 TA65-5 109 99 9.17 3.13 3.44 32.66 34.27 81 TA66-1 109 99 9.17 3.41 3.76 35.63 37.39 209 TA66-2 115 106 7.83 3.45 3.75 37.02 38.56 123 TA67-1 129 116 10.08 3.05 3.39 34.60 36.49 900 TA68-1 126 107 15.08 2.67 3.15 30.02 32.58 214 TA70-1 133 118 11.28 3.10 3.49 35.72 37.93 341 TA71-1 93 85 8.60 3.33 3.65 32.15 33.62 377 TA71-2 99 90 9.09 3.04 3.34 30.25 31.73 105 TA72-1 114 101 11.4 3.19 3.60 34.09 36.22 528 TA73-1 102 92 9.80 2.73 3.02 27.53 28.98 167 TA73-2 72 62 13.89 2.78 3.23 23.57 25.40 56 TA74-1 105 95 9.52 2.74 3.03 28.11 29.55 212 TA75-1 203 181 10.84 2.91 3.26 41.41 43.85 526 TA76-1 84 74 11.90 2.75 3.12 25.20 26.85 19 TA76-2 94 85 9.57 3.23 3.58 31.36 32.97 44 TA77-1 130 117 10.00 2.98 3.31 33.94 35.78 527 TA78-1 86 79 8.14 2.85 3.10 26.42 27.56 165 TA79-1 140 127 9.29 2.91 3.21 34.48 36.20 392 TA80-1 119 108 9.24 3.24 3.56 35.29 37.05 104 TA81-1 106 98 7.55 2.92 3.15 30.01 31.21 288 TA82-1 118 105 11.02 2.94 3.30 31.94 33.86 141 TA83-1 70 66 5.71 2.50 2.65 20.92 21.54 31 TA83-2* 78 74 5.13 3.15 3.32 27.85 28.60 147 TA84-1* 59 56 5.08 3.1 3.27 23.82 24.45 104 TA85-1 101 86 14.85 2.56 3.01 25.77 27.93 159

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 21

Site Ra Rn %N Ca Cn FQIa FQIn Size TA85-2 131 116 11.45 3.09 3.49 35.39 37.60 213 TA86-1 122 108 11.48 2.89 3.26 31.87 33.87 147 TA87-1 123 107 13.01 2.79 3.21 30.93 33.16 17 TA87-2 97 82 15.46 2.61 3.09 25.69 27.94 159 TA87-3* 53 49 7.55 3.77 4.08 27.47 28.57 19 TA88-1 117 103 11.97 2.85 3.23 30.79 32.81 246 TA89-1 120 108 10.00 3.00 3.33 32.86 34.64 197 TA90-1 129 113 12.40 3.22 3.68 36.63 39.13 137 TA90-2 114 105 7.89 3.32 3.61 35.50 36.99 282 TA91-1 162 147 9.26 3.20 3.53 40.78 42.81 645 TA92-1 147 132 10.20 3.14 3.5 38.11 40.21 202 TA93-1 82 73 10.98 2.38 2.67 21.53 22.82 134 TA94-1 95 89 6.32 2.88 3.08 28.11 29.04 67 TA95-1 110 103 6.36 2.97 3.17 31.18 32.22 160 TA96-1 116 102 12.07 2.81 3.20 30.27 32.28 81 TA96-2 145 130 10.34 3.03 3.38 36.54 38.59 197 TA97-1 136 119 12.50 3.24 3.71 37.82 40.43 273 TA97-2 112 105 6.25 3.36 3.58 35.53 36.69 72 TA98-1 161 139 13.66 2.96 3.43 37.59 40.46 149 TA99-1 136 119 12.50 2.96 3.38 34.47 36.85 179 TA100-1* 103 90 12.62 3.17 3.63 32.22 34.47 124 TA101-1* 122 104 14.75 2.97 3.48 32.77 35.50 213 TA102-1* 121 117 3.31 3.64 3.77 40.09 40.77 117 TA103-1* 135 131 2.96 3.70 3.82 43.03 43.69 458 MPRC-1* 177 153 13.56 2.82 3.26 37.51 40.34 2172

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 22

TABLE 3.4 Grades for landscape context, size, and condition of Flint Hills Tallgrass Prairies evaluated on FRMR in 2011/2012. Column codes are: Site = training area+field number; LCgrade = landscape context score (A–D); S = size (acres); Sgrade = size grade (A–D); LCgrade × Sgrade = landscape context grade × size grade (from Table 3.1); FQIn = floristic quality index, native taxa only; FQIngrade = floristic quality index grade; and LC/Sgrade × FQIngrade = landscape context grade/size grade × floristic quality index grade (from Table 3.2). See text for explanation of grades.

LCgrade × LC/Sgrade × Site LCgrade S Sgrade FQIn FQIngrade Sgrade FQIngrade TA3-1 C 230 C C 38.78 B B TA4-1 C 91 D D 41.89 A C TA5-1 C 98 D D 44.90 A C TA6-1 C 145 D D 31.78 C D TA7-1 C 254 C C 39.38 B B TA8-1 C 111 D D 41.97 A C TA9-1 C 36 D D 35.11 B C TA10-1 C 120 D D 37.61 B C TA11-1 C 48 D D 39.76 B C TA11-2 C 171 D D 35.89 B C TA12-1 C 457 B B 40.78 A A TA13-1 C 395 C C 43.05 A B TA14-1 C 603 B B 43.07 A A TA15-1 C 93 D D 33.48 B C TA16-1 C 118 D D 27.16 C D TA17-1 C 412 B B 43.55 A A TA20-1 C 249 C C 40.48 A B TA21-1 C 125 D D 41.00 A C TA22-1 C 290 C C 45.62 A B TA23-1 C 321 C C 39.76 B B TA24-1 C 488 B B 32.73 B B TA25-1 C 348 C C 35.15 B B TA26-1 C 54 D D 33.47 B C TA26-2 C 51 D D 33.75 B C TA27-1 C 367 C C 43.19 A B TA29-1 C 503 B B 44.15 A A TA30-1 C 523 B B 43.28 A A TA31-1 C 276 C C 42.11 A B TA32-1 C 465 B B 42.14 A A TA33-1 C 110 D D 41.19 A C TA34-1 C 29 D D 23.69 D D TA35-1 C 212 C C 34.78 B B TA35-2 C 28 D D 34.96 B C TA35-3 C 91 D D 40.49 A C TA36-1 C 366 C C 27.53 C C TA37-1 C 337 C C 39.20 B B TA38-1 C 108 D D 26.27 C D TA39-1 C 267 C C 34.36 B B TA40-1* C 45 D D 23.27 D D TA41-1 C 150 D D 25.43 C D TA42-1* C 118 D D 24.14 C D TA43-1 C 139 D D 26.80 C D TA44-1 C 76 D D 34.32 B C

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 23

LCgrade × LC/Sgrade × Site LCgrade S Sgrade FQIn FQIngrade Sgrade FQIngrade TA45-1 C 116 D D 27.60 C D TA46-1 C 292 C C 26.50 C C TA47-1 C 64 D D 35.05 B C TA47-2* C 111 D D 34.20 B C TA47S* C 12 D D 16.13 D D TA48-1 C 463 B B 38.44 B B TA49-1 C 45 D D 25.11 C D TA50-1 C 286 C C 40.21 B B TA51-1 C 249 C C 35.92 B B TA52-1 C 65 D D 29.25 C D TA53-1 C 108 D D 36.78 B C TA53-2 C 32 D D 30.17 C D TA54-1 C 370 C C 36.30 B B TA55-1 C 206 C C 28.04 C C TA56-1 C 83 D D 30.61 C D TA57-1 C 329 C C 35.91 B B TA58-1 C 711 A A 37.86 B A TA59-1 C 505 B B 37.59 B B TA60-1 C 197 D D 38.47 B C TA60-2 C 254 C C 35.03 B B TA61-1 C 789 A A 36.69 B A TA63-1 C 614 B B 39.96 B B TA64-1 C 351 C C 39.04 B B TA65-1 C 79 D D 24.92 C D TA65-2 C 56 D D 28.87 C D TA65-3 C 155 D D 31.64 C D TA65-4 C 31 D D 36.66 B C TA65-5 C 83 D D 34.27 B C TA66-1 C 209 C C 37.39 B B TA66-2 C 123 D D 38.56 B C TA67-1 C 900 A A 36.49 B A TA68-1 C 214 C C 32.58 B B TA70-1 C 342 C C 37.93 B B TA71-1 C 377 C C 33.62 B B TA71-2 C 105 D D 31.73 C D TA72-1 C 526 B B 36.22 B B TA73-1 C 167 D D 28.98 C D TA73-2 C 56 D D 25.40 C D TA74-1 C 210 C C 29.55 C C TA75-1 C 526 B B 43.85 A A TA76-1 C 20 D D 26.85 C D TA76-2 C 44 D D 32.97 B C TA77-1 C 527 B B 35.78 B B TA78-1 C 165 D D 27.56 C D TA79-1 C 392 C C 36.20 B B TA80-1 C 105 D D 37.05 B C TA81-1 C 288 C C 31.21 C C TA82-1 C 143 D D 33.86 B C TA83-1 C 31 D D 21.54 D D TA83-2* C 147 D D 28.60 C D TA84-1* C 104 D D 24.45 C D TA85-1 C 159 D D 27.93 C D TA85-2 C 214 C C 37.60 B B

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 24

LCgrade × LC/Sgrade × Site LCgrade S Sgrade FQIn FQIngrade Sgrade FQIngrade TA86-1 C 148 D D 33.87 B C TA87-1 C 17 D D 33.16 B C TA87-2 C 159 C C 27.94 C C TA87-3* C 19 D D 28.57 C D TA88-1 C 246 C C 32.81 B B TA89-1 C 196 D D 34.64 B C TA90-1 C 137 D D 39.13 B C TA90-2 C 282 C C 36.99 B B TA91-1 C 645 A A 42.81 A A TA92-1 C 202 C C 40.21 B B TA93-1 C 134 D D 22.82 D D TA94-1 C 67 D D 29.04 C D TA95-1* C 149 D D 32.22 B C TA96-1 C 81 D D 32.28 B C TA96-2 C 197 D D 38.59 B C TA97-1 C 273 C C 40.43 B B TA97-2 C 72 D D 36.69 B C TA98-1 C 149 D D 40.46 B C TA99-1 C 179 D D 36.85 B C TA100-1* C 124 D D 34.47 B C TA101-1* C 214 C D 35.50 B C TA102-1* C 109 D D 40.77 A C TA103-1* C 458 B B 43.69 A A MPRC-1* C 2172 A A 40.34 B A

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 25

FIGURE 3.2 Histogram of size for 120 Flint Hills Tallgrass Prairies assessed on FRMR in 2011/2012.

FIGURE 3.3 Histogram of floristic quality index values for 120 Flint Hills Tallgrass Prairies assessed on FRMR in 2011/2012.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 26

FIGURE 3.4 Scatterplot of prairie size vs. floristic quality index for 119 Flint Hills Tallgrass Prairies assessed on FRMR in 2011/2012. The MPRC, which was assessed as a single 2,172-acre prairie, had an index of 40.34.

FIGURE 3.5 Scatterplot of prairie size vs. percent non-native species for 120 Flint Hills Tallgrass Prairies assessed on FRMR. The dashed line is linear fit line.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 27

FIGURE 3.6 Locations and grades of Flint Hills Tallgrass Prairies on FRMR in 2011/2012.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 28

FIGURE 3.7 Scatterplot of floristic quality indices for prairies assessed both in 2002/2003 and 2011/2012.

FIGURE 3.8 Condition grade changes among prairies assessed in 2011/2012 as compared to 2002/2003.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 29

FIGURE 3.9 Overall grade changes among prairies in 2011/2012 as compared to 2002/2003.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 30

Chapter 4: Weed Surveys

4.1. Introduction

Invasive plants, especially those that are non-native, are among the greatest threats to natural ecosystems worldwide. Problems associated with them have increased dramatically with expanding human populations, world travel, and international trade. An estimated 5,000 non-native plant species (also called exotic, alien, or introduced) occur in the U.S. today (Morse et al. 1995). Non- indigenous plants occupy an additional 4,600 acres of wildlife habitat per day in the U.S. (Babbitt 1998), and invasive weeds on croplands are estimated to cost the U.S. $26.4 billion annually (Pimentel et al. 2000). Combined annual losses and damages plus control costs from aquatic weeds, crop weeds, weeds in pastures, and weeds in lawns, gardens, and golf courses are close to $34 billion (Pimentel et al. 2000).

More than 500 of the approximately 2,200 (23%) of species and named hybrids of vascular plants documented in Kansas have been introduced since the arrival of Euro-Americans. Exotic plants are of particular concern because many natural controls formerly regulating their populations are absent in the new, non-native environment. Among their many adverse impacts on natural communities (Randall 1995, 1996), non-native plant species often out-compete native species, reducing biodiversity and modifying habitat structure.

Five noxious weed species requiring control have been reported on FRMR: musk thistle (Carduus nutans), field bindweed (Convolvulus arvensis), sericea lespedeza (Lespedeza cuneata), kudzu (Pueraria lobata), and Johnsongrass (Sorghum halepense) (US Army 2010). All except kudzu occur widely across the installation. Objective 3 of this project was to conduct surveys for sericea lespedeza and black locust (Robinia pseudoacacia), an aggressive tree species that, while not listed as noxious in Kansas, can invade prairies, oak savannas, and upland forest. Information about these two species is summarized below.

Lespedeza cuneata (Fabaceae; sericea lespedeza). Native to eastern Asia, sericea lespedeza is a perennial legume with slightly woody stems that can grow to 2 m tall. It was introduced into many parts of the U.S. for erosion control and as food and cover for wildlife, the reason it apparently was planted on FRMR in the mid- to late 1980s. Compared to native grassland species, sericea lespedeza is unpalatable to livestock because of the high concentration of tannins in its tissues. Seeds are dispersed in the fall, may be spread by birds, and can remain viable for over 20 years. It is found extensively along roadsides but also can invade other sites, including thickets, fields, meadows, prairies, and woodlands. It is very drought hardy. Burning, grazing, and fertilization can provide some control on rangeland. Late spring burns on non-rangeland have achieved some success. Sericea lespedeza can become highly invasive, forming dense populations that diminish native biodiversity or impede efforts at ecosystem restoration; it is particularly problematic in rangeland in the southern Flint Hills of Kansas. The species is a serious threat to prairie and woodland communities.

Robinia pseudoacacia (Fabaceae; black locust). Black locust’s original range was in the southeastern U.S. on the lower slopes of the Appalachian Mountains, with some outliers further north on slopes and forest margins in Illinois, Indiana, and Missouri. It is a rapid-growing,

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 31 deciduous tree in the legume family that can grow to 30 m tall. Most natural reproduction is by root suckering and stump sprouting. Black locust becomes a management problem when it aggressively invades dry prairies and savannas, and shades native species. It is found in a variety of disturbed sites such as pastures, degraded woods, old fields, roadsides and rights-of-way. Mowing and burning largely are ineffective control measures because of the plant’s vigorous vegetative propagation. Management has concentrated more on chemical control. Annual haying may prevent first year seedlings from spreading into prairies.

4.2. Methods

The locations and severity of infestations of sericea lespedeza and black locust were documented from May–September, 2011, and in July 2012 in conjunction with prairie assessments. Surveys were conducted on foot and with the use of an ATV. Prairies were surveyed for weeds as floristic quality assessments were being done; non-prairie areas were surveyed simultaneously. Each training unit was surveyed so as to cover as much area and as many different land use types as possible. No attempt was made to use transects to locate and quantify infestations. An attempt was made to locate all occurrences that had been recorded in 2002/2003 surveys (Freeman and Delisle 2004); the status of these occurrences (not found, expansion/contraction, changes in canopy cover) was recorded on field forms and maps. Depending on the terrain, a unit could be surveyed in 2-6 hours. Not all parts of every training unit could be surveyed due to terrain or vegetation conditions. If an area with a previous weed occurrence could not be reached, the old occurrence was retained in the new dataset. This could have led to overestimations in areas that had been treated in the intervening years. Fortunately, this happened only rarely.

Locations were recorded with hand-held Garmin GPS units with an average accuracy greater than 10 m. Multiple GPS readings were recorded and exported into a GIS to delineate the boundary of polygon occurrences. Population boundaries also were drawn on aerial photographs and entered into a GIS using head’s-up digitizing. Occurrences were determined to be either points or polygons based on their size.

Mapping procedures followed the recommendations in Carpenter et al. (2002) and Anonymous (2002), with minor modifications. The minimum mapping unit for each occurrence was <1m2 (individuals mapped), and the minimum distance between adjacent occurrences was 100 m (i.e., any two plants closer than 100 m were mapped as part of the same occurrence). The minimum mapping distance was increased from that used in our 2002/2003 study to avoid mapping dozens of small occurrences with no foreseeable management benefits. Each point or polygon was assigned the following attributes: date observed, observer(s), centroid position, canopy cover (for polygons, using 10 cover classes), and area (calculated in ArcGIS). Stands of sericea lespedeza and black locust that had been killed, usually by spraying, were excluded from our estimates unless there was evidence of live plants emerging from among the dead ones. This situation was not uncommon. Young black locust trees, most presumably suckers, often were seen in and among dead trees, and green sericea lespedeza ramets often were observed emerging from among dead ramets that appeared to have been killed by spraying in the previous growing season.

Training areas 3–103 were surveyed in 2011; the MPRC was surveyed in 2012. Neither the Impact Area nor most of the developed and residential parts of the installation were surveyed.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 32

4.3. Results and Discussion

4.3.1. Weed Survey Results for 2011/2012

Black locust was recorded in 39 training areas, with an estimated 142 acres infested (Table 4.1, Figure 4.1). Many stands contained dead trees, presumably killed by spraying.

Sericea lespedeza was recorded in 101 training areas, including the MPRC, with an estimated 21,604 acres infested (Table 4.1, Figure 4.2). Occurrences were documented in all training areas except those comprised almost entirely of forest and woodland vegetation. Of 21,604 acres with sericea lespedeza, 5,369 acres (24.9%) occur in prairies distributed as follows: 770 acres in A-grade prairies, 1,929 acres in B-grade prairies, 1,226 acres on C-grade prairies, and 1,444 acres on D-grade prairies. Surveys confirmed the most severe infestations are in the central, north-central, and east-central parts of the installation. Presently, the species is less of a problem south of Vinton School Road. Also, training blocks in the northwest part of the installation (J, K, and N) are less severely infested.

TABLE 4.1 Number of training areas (including the MPRC) in which sericea lespedeza and black locust were documented, and total area infested (in acres), in 2002/2003 and 2011/2012. Area figures exclude populations recorded as points. Species # Areas 2003 Area (acres) in #Areas in 2012 Area (acres) in 2003 2012 Black locust 31 150 39 142 Sericea lespedeza 94 12,927 101 21,604

4.3.2. Comparison of 2011/2012 and 2002/2003 Weed Survey Results

Nearly all of the occurrences of black locust documented in 2002/2003 were relocated in 2011/2012. Though the number of training areas in which populations were found increased from 31 to 39, that increase was due in part to changes in training area boundaries. The estimated number of infested acres decreased by 5.3% from 2002/2003 to 2011/2002. Many populations showed evidence of damage from herbicide application, but many sprayed colonies also showed signs of suckering even where mature trees had been killed. However, black locust appears to continue to be a localized pest on the installation.

A comparison of acres infested by sericea lespedeza by cover class is presented in Table 4.2. The most dramatic change was in the 1-10% cover class, where the number of infested acres increased 2.1 times from 2002/2003 to 2011/2012. Sericea lespedeza also showed an overall increase in acres across the sum of the other cover classes from 2002/2003 to 2011/2012 (from 1134.1 acres; 8.8% of total for year up to 3171.7 acres; 14.7% of total for year, respectively), though infested acres showed both increases and decreases among individual cover classes when the two survey periods were compared.

As stated above, 24.9% of the sericea lespedeza observed in 2011/2012 (polygons only, points excluded from calculations) occurred in prairies, a slight percentage increase from 23% in 2002/2003 (Figure 4.3). Sericea lespedeza occupied 19% of native prairie acres in 2011/2012, an

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 33

FIGURE 4.1 Populations of black locust documented on FRMR in 2011/2012.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 34

FIGURE 4.2 Populations of sericea lespedeza documented on FRMR in 2002/2003 and 2011/2012.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 35 increase from 11% in 2002/2003 (from 131 to 770 acres in A-grade prairies; 901 to 1,929 acres in B- grade prairies, 1,022 to 1,226 acres in C-grade prairies, and 867 to 1,444 acres in D-grade prairies.). In A- and B-grade prairies only, sericea lespedeza increased from 6% to 13% of prairie acres infested.

Minimizing the spread of sericea lespedeza into native prairies is identified as a management goal on FRMR (US Army 2010). Despite control efforts, sericea lespedeza was more abundant on native prairies in 2011/2012 than in 2002/2003. Many point occurrences have expanded into polygon occurrences, and many polygons have increased in size. Although the current level of infestation in prairies usually is not severe, this trend is disturbing. Tracked vehicles likely are the primary vector for introduction of sericea lespedeza into native prairies, but other methods of dispersal, such as by birds, also may be occurring. To help slow the spread throughout the installation, aggressive control measures including aerial spraying are used on formerly cultivated ground and other non-native grasslands that are heavily impacted and serve as seed reservoirs. Spot and patch or direct, ground- spraying by truck, ATV, or backpack equipment is used in native prairies to minimize impacts to non-target plants.

TABLE 4.2 Number of infested acres and percent of cover in 10 cover classes for sericea lespedeza on FRMR in 2002/2003 and 2011/2012. Area figures exclude populations recorded as points. % Canopy Cover Area (acres) in % of total cover Area (acres) in % of total cover 2003 in 2003 2012 in 2012 1-10 8758.33 67.76 18432.37 85.32 11-20 2163.91 16.75 2022.91 9.36 21-30 870.52 6.73 476.55 2.21 31-40 374.11 2.89 184.64 0.85 41-50 310.47 2.4 144.84 0.67 51-60 80.35 0.62 70.97 0.33 61-70 31.54 0.24 202.65 0.94 71-80 205.41 1.59 54.53 0.25 81-90 68.56 0.53 10.46 0.05 91-100 63.62 0.49 4.11 0.02

4.3.3. Comparison of 2007 and 2008 Survey Data from FRMR with 2011/2012 Survey Data from KBS

We attempted to compare sericea lespedeza survey data collected for this study in 2011/2012 with that collected by FRMR staff in 2007 and 2008. The FRMR surveys were conducted using two methodologies. Linear scouting was used primarily in areas dominated by open land. Using an ATV, survey transects were driven on a north-south axis with ca 200 m between tracks over the entire training unit. In areas characterized by relatively high levels of woody growth and/or stream channels, this method was not feasible. Instead, surveys were conducted following the contours of the landscape or vegetation outline. With both survey techniques, sericea lespedeza was recorded with a GPS unit and the level of infestation was categorized as heavy, moderate, or light. A direct comparison of data gather by FRMR and us could not be carried out because of differences

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 36

FIGURE 4.3 Locations of prairies and infestations of sericea lespedeza on FRMR in 2011/2012.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 37 in survey and data collection methods. We estimated congruence of the two data sets by examining the overlap of all the 2007 and 2008 data points with our 2011/2012 sericea lespedeza polygons. Spatial analysis was conducted in GIS to determine the number of points falling within each polygon as a percentage of the total number of points (within and outside the polygons) in each training area. This gave us an objective estimate of congruence between the two datasets. A host of factors could affect the congruence of survey data, including differences in survey methods or effort, observer error, timing of surveys, and actual changes in coverage due to control measures or biological phenomena (recruitment or natural mortality). Among these variables, we hypothesized that control efforts would account for most of the incongruence measured in any training area, so we attempted to identify training areas where no spraying occurred from 2007– 2010, allowing us to exclude spraying as a variable. Nine training areas were identified that, as far as the data available to us indicated, had not been sprayed for sericea lespedeza from 2007–2010. Unfortunately, of those nine training areas, weed surveys were conducted only in two areas (TA52, TA55) in both 2007/2008 and 2011/2012. We estimated the congruence between the two data sets in these two training areas at 95% and 59%, respectively. However, this was far too small a sample to draw any meaningful inferences, and we concluded that there was no way rigorously to test the congruence of survey efforts without the confounding effects of spraying.

Infestations of sericea lespedeza also were estimated during the 1999 growing season using different methods from those described above. Degree of infestation was estimated in five classes: heavy, moderate, light, not infested, and not scouted. The polygons produced for that study suggested that very liberal criteria were used to determine the sizes of areas infested. Despite any real methodological differences between the 1999 and 2010/2011 surveys, and the fact that there have been continued efforts to control some infestations through herbicide use, the generalized maps from the two studies show many similarities in both location and severity of infestations across the installation. The 1999 surveys suggested that the most severe infestations were in training areas in the central and eastern part of the installation.

4.3.4. Assessment of Potential Impact of Spraying on Sericea Lespedeza

An issue of particular interest is the impact that spraying may be having on sericea lespedeza on FRMR. Examining this issue proved to be extremely challenging due to differences in survey methods or effort, observer bias, timing of surveys, changes in coverage due to control measures or biological phenomena, and incomplete data. To complicate matters, we did not have access to information about where spraying occurred from 2002–2006, and we were unable to take into account the potential effect of prescribed burns conducted later in the growing season as part of the overall strategy to control sericea lespedeza. In spite of these variables, we attempted to examine this issue within the limits of the data available to us, hypothesizing that spraying would decrease the estimated coverage of sericea lespedeza in training areas on FRMR.

The estimated area of coverage of sericea lespedeza, regardless of cover class, was obtained for each training area for the 2002/2003 and 2011/2012 survey periods. Point data were excluded from the estimates because they had no measurable area. Spraying information from 2007–2010 was obtained from FRMR. Whether or not any spraying (spot, broadcast, aerial) was documented in a training unit in 2007, 2008, 2009, or 2010 was recorded, and the total number of years of spraying in each training area (ranging from 0–4) was recorded. Data were analyzed as a case of repeated measures, with the 2002/2003 acres representing pre-spray conditions and 2011/2012 acres representing post-

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 38 spray conditions.

Data were available for 96 training areas. No spraying occurred in any of the four years in nine training areas (3, 4, 11, 17, 18, 22, 52, 55, and MPRC); at least one year of spraying occurred in all other 87 training areas (Figure 4.4). Due to the non-normal distribution of the data, a non- parametric sign test was performed to determine if there was a difference in 1) the distribution of samples in training areas in which no spraying occurred from 2007–2010, and 2) the distribution of samples in training areas in which at least one year of spraying occurred from 2007–2010. In each case, the distribution of samples pertains to the number of training areas in which the estimated acres of sericea lespedeza decreased from 2002/2003 to 2011/2012 compared to the number of training areas in which the estimated number of acres of sericea lespedeza increased from 2002/2003 to 2011/2012.

We found no statistical difference in the distribution of samples in non-sprayed units (p = 1.0), meaning that increases and decreases in estimated coverage appeared to be equal. Even though we might predict an increase in estimated coverage in training areas in which there has been no recent control, several factors might explain this result. First, the magnitude of change (increase or decrease in coverage) in the six areas south of Vinton School Road was very small. These training areas historically have had low levels of infestation and few large weed occurrences (polygons). We did document more point occurrences in some of these training areas, but as noted above, point occurrences were not included in coverage estimates for training areas. Not surprisingly, among those training areas with no spraying from 2007–2010, the two with the largest increase (TA 52 with 100x increase; TA 55 with 51x increase) each contain limited areas of native prairie, are located in the central part of the installation where infestation levels generally are extremely high, and experience frequent disturbance from tracked vehicles.

There was a statistical difference in the distribution of samples in sprayed units (p <0.001), but not in the direction that would be predicted. Positive ranks (69) were far more frequent than negative ranks (18), indicating that estimated coverage increased in training areas in which at least one year of spraying had occurred from 2007–2010. This is seen in Figure 4.5, in which most of the points (training areas) are positioned above the reference line, indicating that the estimated coverage of sericea lespedeza increased in most training areas in 2011/2012 as compared to 2002/2003.

Our findings beg the obvious question of why this species appears to be spreading on FRMR in spite of ongoing efforts to control it and evidence that herbicide can provide effective control (Ohlenbusch and Bidwell 2007, Emry 2008). We can only speculate about the possible reasons, but recent studies suggest issues potentially at play.

Emry (2008) and Emry et al. (2011) found that control of sericea lespedeza, but not necessarily eradication, depends on both accurate and complete identification of patches followed by application of herbicides. Using a model of spatial spread of sericea lespedeza, Emry et al. (2011) found that detection of populations, which is a function of the number and size of patches, plays an important role in control. Small, isolated patches or individuals are less likely to be detected than are larger, clustered patches. The effectiveness of control can be nullified if patches go undetected, allowing plants to persist, spread locally, and produce seeds that can expand the area of infestation. Emry et al. (2011) made two management recommendations based on their study: 1) detection and treatment of all target plants is critical, and 2) approaches should be adopted that allow detection success to be estimated and, in cases where detectability is low, survey techniques need to be VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 39 improved to increase detection (such as double-observer surveys) or control approaches that are less affected by observer bias should be used (for example, broadcast spraying vs. spot spraying). Also, including abundance in predictions of spread may prevent managers from underestimating future spread and the resources required effectively to manage populations (Emry 2008, Emry et al. 2011). At present, we lack detailed demographic information for sericea lespedeza and the role of seed banks on recruitment. Both issues are potentially important in understanding the population dynamics of this species.

Studies by Emry (2008) and Emry et al. (2011), and our survey results suggest that managers at FRMR should consider several management strategies going forward. First, adoption of survey protocols that yield consistent and accurate estimates of sericea lespedeza populations should be adopted. A study to estimate overall detection probabilities and observer biases, such as used by Alexander et al. (2009), could help determine the efficacy of different survey approaches. Second, a spraying regimen (type, location, and frequency of application) should be adopted that takes into account detection issues as well as biodiversity concerns. If detection probabilities are high and observer biases are low for recently used survey protocols, spot-spraying should provide effective control on tallgrass prairies, though spraying may have to be done annually due to the high probability of new patches becoming established via seed immigration from surrounding sources. At the same time, spot-spraying should cause less collateral damage to native plants compared to broadcast or aerial spraying. If detection probabilities are determined to be low, new protocols that yield better detection need to be implemented because the alternative, use of control approaches less affected by observer bias (such as broadcast or aerial spraying), would result in higher collateral damage to native biodiversity. Given the distribution and abundance of sericea lespedeza in many other parts of the installation, broadcast and/or aerial spraying may be the only control measures that will keep sericea lespedeza in check. Regardless of the effectiveness of detection and control measures, management of sericea lespedeza on FRMR undoubtedly will be a long-term and resource-intensive proposition.

4.4. Species of Potential Concern

Freeman and Delisle (2004) reported the first documented occurrence of Rubus bifrons Vest (European or Himalayan blackberry, often called R. armeniacus Focke or R. discolor Weihe & Nees in the North American botanical literature) on FRMR at a single site in Training Area 65. This Old World species is cultivated in the western U.S. for its fruits, but it has become widely naturalized in the western and eastern U.S., and in several Canadian provinces. Though it provides food and shelter for a variety of wildlife, it can spread aggressively and become a pest, especially in disturbed habitats (Tirmenstein 1989). At the time of its discovery on FRMR, it was suggested that this species should be controlled to prevent possible spread on the installation.

During surveys in 2011, small populations of Himalayan blackberry were discovered in four more training areas in the northwest part of FRMR. Specifically, the species was documented in prairies in Training Areas 65, 67, 70, 75, and 76, all in the drainages of Timber, Dry, and Madison creeks. Occurrences generally included 1–15 plants, and in no location was the species abundant. Plants still were present in the area where it first was discovered (39.24704°N, -96.93996°W) as well as along the east-west gravel road between Training Areas 65 and 76 (39.24893°N, -96.940309°W), just north of the original site. Occurrences in other training areas comprised few, widely scattered plants in open prairie or formerly cultivated areas.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 40

Himalayan blackberry can be controlled via chemical and/or mechanical methods, and it is recommended that plants be eliminated when they are discovered to prevent further spread on the installation. Even though Himalayan blackberry currently is not a pest, control of this non-native species now may prevent future environmental impacts.

4.5. Conclusions

Field surveys were carried out to identify the locations and severity of infestations of black locust (Robinia pseudoacacia) and sericea lespedeza (Lespedeza cuneata) on FRMR. Populations of black locust are small, isolated, occupy approximately 142 acres, and are not a serious threat to biodiversity.

Sericea lespedeza was recorded in 101 training areas, including the MPRC, with an estimated 21,604 acres infested. The estimated infested acres increased 2.1 times since the completion of comparable surveys in 2002/2003. Sericea lespedeza occupied 19% of native prairie acres in 2011/2012, an increase from 11% in 2002/2003. This increase does not take into account point data – occurrences of sericea lespedeza that were smaller than the minimum size for mapping as polygons. These point occurrences increased dramatically on native tallgrass prairies. Surveys indicate that areas identified as being the most severely infested in 2002/2003, specifically in the central, north-central, and east- central parts of the installation, generally are as severely infested if not more severely infested now as compared to 2002/2003.

Attempts to compare data from other surveys, specifically those during the years from 2007–2010, with surveys for this study, and to assess the potential impact of spraying on sericea lespedeza were limited because of differences in survey methods or efforts, potential observer error, timing of surveys, control efforts, biological phenomena, and incomplete data that could not be adequately controlled among the various data sets. Our data do suggest that sericea lespedeza is increasing in distribution and abundance on FRMR in spite of efforts to control it. While the most severely impacted parts of the installation are formerly cultivated areas, intrusion into native tallgrass prairies is increasing.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 41

FIGURE 4.4 Number of years for which some type of spraying (spot, broadcast, aerial) has been recorded in training areas on FRMR.

FIGURE 4.5 Scatterplot of estimated acres of sericea lespedeza in training areas in 2011/2012 and 2002/2003. Each square represents a different training area. The dotted line is a reference line where r2 = 1.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 42

Chapter 5: Rare Species

5.1. Introduction

Previous surveys on FRMR have documented numerous rare and protected animal species (US Army 2010). The Kansas Natural Heritage Inventory tracks site-specific information on a subset of these species based largely on the number of known occurrences extant in the state (Table 5.1). Additional rare species observed within the last 25 years that are resident for at least part of the year include vertebrates (Regal Fritillary), fishes (Common Shiner, Johnny Darter), reptiles (Texas Horned Lizard, Western Hognose Snake), birds (Bell’s Vireo, Black-billed Cuckoo, Common Poorwill, Dickcissel, Field Sparrow, Grasshopper Sparrow, Greater Prairie-chicken, Kentucky Warbler, Loggerhead Shrike, Northern Flicker, Prothonotary Warbler, Red-headed Woodpecker, Rusty Blackbird, Upland Sandpiper, Wood Thrush), and mammals (Southern Bog Lemming). These species fall into one or more of the following categories: Army Species At Risk, Kansas Species of Greatest Conservation Need, U.S. Fish and Wildlife Service Birds of Conservation Concern.

TABLE 5.1 Species tracked by the Kansas Natural Heritage Inventory documented on FRMR within the last 25 years. Birds occurring only as migrants or transients are excluded. Data from the Fort Riley 2010 Integrated Natural Resources Management Plan and the Kansas Natural Heritage Inventory database.

Scientific Name Common Name G Rank S Rank Federal State SGCN Status Status Haliaeetus leucocephalus Bald Eagle G5 S2B, S4N Tier 2 Accipiter striatus Sharp-shinned Hawk G5 S1B, S4N N/A Ammodramus henslowii Henslow’s Sparrow G4 S3B C Tier 1 Charadrius melodus Piping Plover G3 S1B, S2N T T Tier 1 Sternula antillarum Least Tern G4 S1B E E Tier 1 Asio flammeus Short-eared Owl G5 S2B, S3N C Tier 1 Seiurus aurocapilla Ovenbird G5 S1B N/A Hybognathus placitus Plains Minnow G4 S2S3 T Tier 1 Notropis topeka Topeka Shiner G3 S2 E T Tier 1 Phoxinus erythrogaster Southern Redbelly Dace G5 S2S3 C Tier 1 Cycleptus elongatus Blue Sucker G3G4 S3 C Tier 1 Gryllotalpa major Prairie Mole Cricket G3 S3 C Tier 1 G Rank: Global conservation status rank: G1 = imperiled; G5 = secure. S Rank: State conservation status rank: S1 = imperiled; S5 = secure. B = Breeding season; N = non-breeding season See http://www.natureserve.org/explorer/ranking.htm for complete definitions of G and S Ranks Federal Status: E = Endangered; T=Threatened State Status: E = Endangered; T=Threatened; C=Species in Need of Conservation SGCN: Kansas Species of Greatest Conservation Need

Objective 4 of this project was to document locations of protected and rare animal and plant species, if any were encountered.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 43

5.2. Methods

Field crews recorded observations of rare species concurrent with vegetation surveys conducted during the 2011/2012 field seasons. No species-specific surveys using specialized survey techniques needed to detect most of these species were employed.

5.3. Results and Discussion

No state-rare plant species were encountered during this study. Two plant species, Chenopodium pallescens Standl. (pale goosefoot) and Sporobolus heterolepis (A. Gray) A. Gray (prairie dropseed) were removed from the state’s list of rare plants based on data collected at FRMR in 2002/2003. Populations of both species were observed on FRMR during vegetation surveys in 2011/2012, including many large, healthy populations of prairie dropseed.

No previously undocumented rare animals were observed. Henslow’s Sparrow was the only animal tracked by the KSNHI documented during this study. The Short-eared Owl is present only in winter; our field work was conducted from May through September. As our objective was to document rare species if they were encountered during vegetation surveys, we did not expect to encounter the Prairie Mole Cricket or the several fish and bird species known to occur on FRMR, and none were observed.

Henslow’s Sparrow was detected at 22 locations on 17 different training areas (Fig. 5.1) with observation dates ranging from May 24 through August 5, 2011. Observations were distributed throughout the installation except south of Vinton School Road where none were detected. Several observations were in areas not defined as Henslow’s Sparrow habitat in the FRMR Integrated Natural Resources Management Plan (US Army 2010). Birds were found most frequently in formerly cultivated fields (16 observations) with a smaller number detected in native prairie (6 observations). Weather conditions were not always suitable for detecting singing birds (high winds, rain, high temperatures). In addition, not all members of the field crew were equally skilled at identifying the Henslow’s Sparrow, whose vocalizations can be difficult to detect. Thus, we do not consider this a comprehensive survey for this species.

Henslow’s Sparrow is a Kansas Species of Greatest Conservation Need categorized as a Tier 1 species in the Kansas Comprehensive Wildlife Conservation Plan (Wasson et al. 2005) and an Army- designated Species at Risk (US Army 2010). Populations have declined sharply throughout its range in the central and eastern United States, and loss of habitat is believed to be an important factor in this decline. It is considered an area-sensitive species that needs large tracts of grassland habitat (> 30 ha) in which to breed. It nests in dense grasslands with standing dead vegetation and a well- developed layer of litter, and thus generally is found in tracts that have not been burned or hayed for 2-5 years. Management actions identified in the Fort Riley INRMP include maintaining and improving nesting habitat by managing for 3-year old grasslands within the Henslow’s Sparrow habitat area, controlling sericea lespedeza, and reducing encroachment of woody vegetation.

5.4. Conclusions

FRMR is known to support populations of several rare animal species that due to the lack of directed surveys were not encountered during this study. Henslow’s Sparrow was the only state-rare animal species tracked by KSNHI documented during this study. A Rare Species Management Plan VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 44 is in effect for this species. No state-rare plant species were documented. Two plant species formerly on the state list of rare species, Chenopodium pallescens and Sporobolus heterolepis, are no longer tracked by KSNHI.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 45

FIGURE 5.1 Locations of Henslow’s Sparrow sightings on FRMR in 2011.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 46

Literature Cited

Agri-Service Associates, Inc. 1985. Fort Riley land use inventory and condition survey project report. Agri-Service Associates, Inc., Manhattan, KS. 2 pp + appendices and maps.

Alexander, H.M., N.A. Slade, W.D. Kettle, G.L. Pittman, and A.W. Reed. 2009. Detection, survival rates, and dynamics of a cryptic plant, Asclepias meadii: applications of mark-recapture models to long-term monitoring studies. J. Ecol. 97: 267–276.

Anonymous. 2002. Weed mapping standards. Unpublished report from the North American Weed Management Association. (http://www.nawma.org/).

Babbitt, B. 1998. Statement by Secretary of the Interior Bruce Babbitt on invasive alien species, “Science in Wildlife Weed Management” Symposium, Denver, CO, April 8, 1998. U.S. Dept. of the Interior News Release. (http://www.doi.gov/news/archives/speeches&articles/weedbab.htm).

Carpenter, A. T., T. A. Murray, and J. Buxbaum. 2002. Inventorying and mapping invasive plants. Nat. Areas J. 22: 163-165.

Emry, D.J. 2008. Population ecology and management of the invasive plant, Lespedeza cuneata. PhD dissertation. University of Kansas.

Emry, D.J., H.M. Alexander, and M.K. Tourtellot. 2011. Modelling the local spread of invasive plants: importance of including spatial distribution and detectability in management plans. J. Applied Ecol. 48: 1391–1400.

Freeman, C. C. and J. M. Delisle. 2004. Vegetation of the Fort Riley Military Reservation, Kansas. Open-file Report No. 119. Kansas Biological Survey. Lawrence, KS. 110 pp.

Freeman, C. C. and C. A. Morse. 2002. Kansas floristic quality assessment: coefficients of conservatism. Unpublished report of the R. L. McGregor Herbarium and Kansas Biological Survey, University of Kansas. Lawrence, KS. 15 pp. + appendix.

Groves, C. R., D. B. Jansen, L. L. Valutis, K. H. Redford, M. L. Shaffer, J. M. Scott, J. V. Baumgartner, J. V. Higgins, M. W. Beck, and M. G. Anderson. 2002. Planning for biodiversity: putting conservation science into practice. BioScience 52: 499–512.

Herman, K. D., L. A. Masters, M. R. Penskar, A. A. Reznicek, G. S. Wilhelm, and W. W. Brodowicz. 1997. Floristic quality assessment: development and application in the state of Michigan (USA). Natural Areas J. 17: 265-279.

Lauver, C. L., K. Kindscher, D. Faber-Langendoen, and R. Schneider. 1999. A classification of the

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 47

natural vegetation of Kansas. Southwestern Nat. 44: 421-443.

Morse, L. E., J. T. Kartesz, and L. S. Kutner. 1995. Native vascular plants. Pp. 205–209. In: Our living resources: a report to the nation on the distribution, abundance, and health of U.S. plants, animals, and ecosystems. U.S. Dept. of Interior, National Biological Service. Washington, DC. 530 pp.

Ohlenbusch, P.D. and T. Bidwell. 2001. Sericea lespedeza: history, characteristics, and identification. MF-2408. Kansas State University. Agricultural Experiment Station and Cooperative Extension Service.

Pimentel, D., L. Lach, R. Zuniga, and D. Morrison. 2000. Environmental and economic costs of nonindigenous species in the United States. BioScience 50: 53–65.

Randall, J. M. 1995. Assessment of the invasive weed problem on preserves across the United States. Endangered Species Update 12: 4–6.

Randall, J. M. 1996. Weed control for the preservation of biological diversity. Weed Technology 10: 370–381.

Rooney, T. P. and D. A. Rogers. 2002. The modified floristic quality index. Natural Areas J. 22: 340 –344.

Swink, F. and G. S. Wilhelm. 1979. Plants of the Chicago region. 3rd edition. Morton Arboretum, Lisle, IL. 922 pp.

Taft, J. B., G. S. Wilhelm, D. M. Ladd, and L. A. Masters. 1997. Floristic quality assessment for vegetation in Illinois: a method for assessing vegetation integrity. Erigenia 15: 3–23.

Tirmenstein, D. 1989. Rubus discolor. In: Fire Effects Information System, [Online]. U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, Fire Sciences Laboratory (Producer). http://www.fs.fed.us/database/feis/ [2012, November 29].

US Army, Fort Riley, Kansas – Environmental Division, Directorate of Public Works. 2010. Integrated Natural Resources Management Plan, Fort Riley, Kansas. 120 pp + Appendix A- M.

USACE. 1985. Land utilization vegetation and condition survey, Ft. Riley, Kansas. U.S. Army Corps of Engineers, Kansas City District.

Wasson, T., L. Yasui, K. Brunson, S. Amend, and V. Ebert. October 2005. A Future for Kansas Wildlife, Kansas’ Comprehensive Wildlife Conservation Strategy. Dynamic Solutions, Inc. in cooperation with Kansas Department of Wildlife and Parks. 170 pp.

Wilhelm, G. S. 1977. Ecological assessment of open land areas in Kane County, Illinois. Kane County Urban Development Division. Geneva, Illinois.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 48

Wilhelm, G. S. and D. Ladd. 1988. Natural area assessment in the Chicago region. Pp 371–375 In: Trans. 53rd North American Wildlife and Natural Resources Conference.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 49

This page intentionally left blank.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 50

Appendix A. Mapping Encroachment of Woody Vegetation.

TESTING A NEW METHOD FOR MAPPING ENCROACHMENT OF WOODY VEGETATION ON THE FORT RILEY MILITARY RESERVATION USING AN OBJECT-BASED IMAGE ANALYSIS APPROACH

Dana L. Peterson

Kansas Applied Remote Sensing Program, Kansas Biological Survey University of Kansas, 2101 Constant Avenue, Lawrence, KS 66047-3759 [email protected]

Submitted December 31, 2012

Citation: Peterson, Dana L. 2012. Testing a new method for mapping encroachment of woody vegetation on the Fort Riley Military Reservation using an object-based image analysis approach. Pp. 50–77. In Delisle, J. M., C. C. Freeman, and D. L. Peterson. 2012. Vegetation Survey and Mapping of the Fort Riley Military Reservation, Kansas. Open-file Report No. 174. Kansas Biological Survey. Lawrence, KS. 78 pp.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 51

A.1. Introduction and Background

The encroachment of woody vegetation into grassland areas is a major management concern on Fort Riley Military Reservation (FRMR). Woody or shrubby vegetation is located along the edges of woodlands, in isolated patches along smaller, intermittent drainages and ravines, and scattered throughout grasslands (US Army 2010). Encroachment into grasslands is an impediment to military training exercises and reduces habitat quality for native species. The vegetation of FRMR has been mapped and described in three earlier studies (Agri-Service Associates, Inc. 1985, USACE 1985, Freeman and Delisle 2004). For the map produced by KBS in 2004, analysts used on-screen digitizing and photo-interpretation techniques to delineate polygons of woodland and forest cover. This technique is subjective, especially for transitional land cover types such as woodlands, and the detail of digitizing varies by image analyst. Delineating woody encroachment was not a specific objective of earlier mapping efforts. Formerly cultivated areas were usually designated as such regardless of the extent of the invasion of woody growth. The general mapping of tree and woodland stands and areas of woody encroachment were captured, but delineations were somewhat coarse. Overlaying the map on 2003 imagery also reveals errors of commission and omission for both the forest and woody encroachment classes (Figure A.1).

FIGURE A.5 A subset of FRMR showing coarse delineations of woody and forest classes from 2002/2003. The areas indicated show errors of omission (e.g., woody encroachment delineated as grassland) and errors of commission (e.g., grassland delineated as woody encroachment).

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 52

A.2. Methods

A.2.1. Mapping Approach

For Objective 5 of this study we tested a new approach to map the encroachment of woody vegetation on the installation and used this approach to compare woody cover in two time periods. This mapping endeavor addresses two key data needs identified by land managers at FRMR. First, it provides an update to the 2002-2003 mapping effort, and second, it provides an opportunity to develop a repeatable, objective mapping methodology for future and historical mapping efforts. An object-based image analysis (OBIA) mapping approach was selected due to the ability to incorporate context, texture, and spectral information, and previous research has shown an OBIA can offer improvements in mapping land use and land cover over the traditional per-pixel approach (Myint et al. 2011, Platt and Rapoza 2008, Thomas et al. 2003, Wang et al. 2004, Whiteside et al. 2011, Yaun and Bauer 2006). The task of mapping woody encroachment was divided among three image analysts. A series of rule-sets was developed and tested to classify two subsets of the study area into the following classes: 1) non-vegetation (includes bare soil, mowed or harvested fields, rock outcrops, tree shadows, water, roads, and urban features); 2) grassland (native and former cropland); 3) woody encroachment; and 4) forest/trees. To show how woody encroachment has changed over time, we apply one of the rule-sets to 2006 aerial imagery over a subset of FRMR. Aerial imagery from 2006 was selected because that is the earliest year 4-band imagery is available over the study area, and our preliminary results indicate that the normalized difference vegetation index (NDVI) is critical for separating vegetation types. It should be noted that we do not have LiDAR data for 2006 for the FRMR and therefore, used a rule- set that did not utilize LiDAR data. Furthermore, no field data were available for 2006 to perform an accuracy assessment of the mapped area. Due to the iterative nature of rule-set development and time-intensive processing of rule-sets in OBIA along with the high spatial resolution of the data used in the classification, rule-sets were developed and tested on multiple spatial subsets. Testing the rule-sets on spatial subsets allows the analyst quickly to modify and test changes as part of optimizing a rule-set. Another objective for testing rule-sets on other areas is to minimize over-fitting a rule-set to a particular area not representative of the entire study area. Subsets were selected based on the two distinct landscapes observed within the FRMR (Figure A.2). The South subset represents a landscape with more topographic relief and vegetation composed of native prairie, woodland, forest, urban, and woody encroachment. The Central subset represents less topographic relief and is a mix of native prairie, forest, urban, woody encroachment, and former cropland in various stages of succession to grassland. A potential risk of using subsets is a failure to capture adequate landscape heterogeneity that represents the larger study area and, as a result, the rule-set will not generalize well to the entire extent. The subsets selected included areas where field data were collected in 2012 and also included large extents of all the vegetation classes being mapped. Using field data collected in 2012, accuracy levels were calculated for these subsets. Once mapping accuracy levels were calculated and compared, the rule-set with the best overall accuracy was applied to the entire scene or study area. A separate accuracy assessment was completed for the entire study area.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 53

FIGURE A.2 Subsets in FRMR used to develop and test rule-sets using an object-based image analysis (OBIA) classification approach.

A.2.2. Data Preprocessing

A combination of software was used for data preprocessing , accuracy assessment, and change detection. Software used include ESRI ArcGIS 10.1, Erdas Imagine 10, Quick Terrain Modeler (QTM), and eCognition 8. eCognition 8 was used for image classification. All data were projected to the universal coordinate system, UTM, Zone 14, NAD83. Because 4-band county mosaics were unavailable for 2010, quarter quads of 4-band (red, green, blue, near infrared (NIR)) aerial imagery flown in 2010 was obtained from the Data Access and Support Center (DASC) at the Kansas Geological Survey at the University of Kansas. The aerial imagery has

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 54 a spatial resolution of 1-meter. The quarter-quads were mosaicked to create one 4-band image. The layer was subset to the study area. The aerial imagery was flown on July 23, 2010, and July 27, 2010. For 2006, county mosaics in MrSID format were downloaded from the DASC website. The MrSID Generation 3 file format uses lossless compression so no data loss occurred. The MrSID files were converted to ERDAS Imagine (.img) file format and layer-stacked to create a 4-band image. The data were subsetted to the study area. The aerial imagery was flown between June 18, 2006, and July 6, 2006. LiDAR data from 2010 were obtained from the GIS Division at FRMR and processed using QTM software. A digital terrain model (DTM) and a digital surface model (DSM) were created from the LiDAR data using the last and first returns, respectively. A normalized digital surface model (nDSM) was then created by subtracting the DTM (representing surface elevation) from the DSM (representing the top of the surface canopy) to obtain the height of landscape features such as buildings and tree canopy height. Trees and buildings have the highest values, woody encroachment areas have moderate values, and grassland and bare ground have the lowest values (Figure A.3b; see Figure A.3a for landscape context).

FIGURE A.3 a. Natural-color composite of the 2010 aerial imagery for a subset of the FRMR to use as visual reference for other layers calculated; b. Example of the median filter of the nDSM derived from 2010 LiDAR data.

The normalized difference vegetation index (NDVI) was calculated for each pixel as part of rule-sets developed within eCognition software. NDVI is used as a surrogate for photosynthetically active vegetation or live biomass. The standard calculation for NDVI was used ((NIR-Red/NIR+Red)). Trees have relatively high NDVI, woody encroachment areas have moderate to high NDVI,

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 55 grassland areas have lower NDVI, and unvegetaed areas have NDVI near zero or negative (Figure A.4a). Multiple image texture layers were derived and compared, namely Sobel’s edge detection and gray level co-occurrence matrix (GLCM) texture measures. The assumption is that texture combined with spectral data will aid in discriminating among vegetation types. The Sobel operator was used to produce an edge detection layer using the NIR band. Prior to calculating Sobel, the NIR band was smoothed using a 3-by-1 kernel filter in an effort to reduce Sobel’s sensitivity to spectral noise, especially with high-resolution, 1-meter data. The Sobel operator emphasizes areas with high spatial frequency, which corresponds to edges. Trees and woody vegetation have higher edge values than does more homogeneous vegetation, such as grassland (Figure 4b). Sobel’s edge detection was calculated as a processing step in the rule-sets developed.

FIGURE A.4 a. Example of the normalized difference vegetation index (NDVI); b. Example of the Sobel operator applied the near-infrared (NIR) band from 2010 aerial imagery.

In addition to Sobel’s edge detection, the following GLCM texture measures were calculated for image objects: angular second moment, contrast, correlation, dissimilarity, entropy, homogeneity, and mean object NDVI in all directions (0°, 45°, 90°, and 135°) (Haralick et al. 1973). Research has shown that while GLCM texture measures are computational intense and add significant processing time, they improve classification accuracy levels, making processing more cost-effective (Wijaya and Gloaguen 2007, Kim et al. 2009, Lu and Weng 2005, Haung et al. 2009). GLCM tabulates the frequency of different combinations of gray levels for two pixels in a relative position for the entire image.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 56

The GLCM texture measures are calculated using the following equations (Trimble 2012):

GLCM Angular Second Moment:

GLCM Contrast:

GLCM Correlation:

GLCM Dissimilarity:

GLCM Entropy:

GLCM Homogeneity:

GLCM Mean

GLCM Standard Deviation

Parameters: • i is the row number • j is the column number • Pi, j is the normalized value in the cell i, j • N is the number of rows or columns • μi; j is the GLCM mean • σi; j is the GLCM standard deviation

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 57

While research has shown that GLCM texture measures improve mapping efforts, one analyst focused efforts on identifying which texture measures would be most suitable to separate vegetation types in FRMR. The analysis was performed using the two most separable vegetation types, forest/trees and grasslands. The South subset was used for this comparative analysis. In eCognition software, GLCM texture measures are calculated as temporary files, so each time the rule-set is opened, the GLCM texture measures are recalculated, making it a time-intensive task. To test the texture measures, photointerpretation was used to select image objects for the tree and grassland classes. These image objects and associated texture measures were stored as membership functions for each texture measure. These two classes were selected because they likely would be more separable than the transitional woody encroachment class. If the analysis were fruitful, the texture measures would be tested on the more difficult to map class: woody encroachment. An analysis was performed to evaluate the overlap in the sample distributions of the two vegetation classes. This would determine the utility of the GLCM texture measures. The results indicated that sample distributions of GLCM texture measures for forest/trees and grasslands substantially overlap, with values ranging from 31–76%. GLCM texture measures are a tabulation of the frequency of different combinations of gray levels occurring at the pixel level (Trimble 2012). We found the GLCM texture measures did not differentiate between grassland and forests, and there was relatively high overlap with the texture measures between the tree and grassland vegetation classes. Based on our analysis, Sobel of NIR was identified to be the most useful texture measure for the FRMR. A.2.3. Rule-Set Development

Mapping was divided among analysts so different approaches could be explored and compared. Specifically, we wanted to test the OBIA classification technique with and without the nDSM to allow us to determine whether vegetation classes can be mapped when concurrent LiDAR data are unavailable, and also to determine the utility of LiDAR and derived variables in the classification of vegetation classes. These findings will aid land managers in determining whether ongoing LiDAR collections are required to monitor future changes in woody encroachment at FRMR. It was anticipated that NDVI could adequately separate forest and grassland vegetation types, but it was uncertain if NDVI could map a transitional vegetation type such as woody encroachment. The nDSM derived from the 2010 LiDAR was expected to provide critical information for differentiating between forest (including individual trees) and woody encroachment, and between grassland and woody encroachment. We were aware that variation in woody encroachment density could complicate the mapping effort. Some of the rule-sets developed are described below. There are commonalities between the rule- sets that are worth describing and are general best practices for rule-set develop for the FRMR that were explored early in the project. One commonality was the use of a multi-resolution and multi- level segmentation approach. The segmentation process groups pixels into image objects based on weightings of layers defined by the user. Parameters used to define the shape and size of image objects includes scale, shape, and compactness. For this study area, we found that working from coarser to finer segmentation levels (i.e., coarser to finer image objects) was more efficient and produced image objects that were more useful for the rules in the classification. For each rule-set, we mapped classes that were most distinct first and classes that were less distinct thereafter. We found it best to separate image objects into non-vegetated and vegetated classes first and then re-segment the vegetated class into finer objects for additional rules for classification to map grassland, woody encroachment, and trees. Membership functions for vegetated and non- VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 58 vegetated classes had little overlap, with values ranging from 0–4%. Segmentation parameters of scale, shape, and compactness were based on trial and error. This process is time-consuming but is necessary to maintain an appropriate level of mapping detail. Rule-set 1: NDVI and the median filter of the nDSM (nDSM_med) were layers calculated as part of the rule-set. A median filter was applied to the nDSM to reduce noise and false readings observed in the nDSM. Once these additional layers were calculated, the image was segmented using the 4- bands of aerial imagery and NDVI. Using photo interpretation, vegetated and non-vegetated image objects were selected and stored as samples to create membership functions for vegetated and non- vegetated classes separately. Membership functions for each class were used in a nearest neighbor classifier where each image object was identified as vegetated or non-vegetated. The vegetated class was extracted and the nDSM_med was used in a series of multi-level segmentations to separate canopy height into trees (>2m) and woody encroachment (0.2m–2.0m). A multi-level segmentation approach allows the user to identify threshold levels of layers or variables to separate objects into classes. Following the first multi-level segmentation that identified the nDSM_med greater than 2m as trees, several rules were utilized to refine the tree class. One function identified objects as trees that were less than 200 pixels in area and 75% of a relative border to existing objects classified as trees. The second rule uses a grow function that allowed image objects classified as trees that were less than 15,000 ft2 to grow into other image objects. The second multi-level segmentation identified woody encroachment in the vegetated class if the nDSM_med was greater ≥ 0.2m and <2.0m. Next, small image objects of woody encroachment that were enclosed by trees were recoded to trees. The remaining image objects of woody encroachment identified by the nDSM_med were used as samples in a nearest neighbor classifier to map woody encroachment. Any remaining image objects in the vegetated class were recoded to the grassland class. The merge region function was used to group image objects for each class into single image objects. Then the merged objects were export into a geotiff file format. Rule-Set 2: NDVI and Sobel of the NIR band were calculated as part of the rule-set. Once these layers were calculated, the image was segmented using the 4-bands of aerial imagery and NDVI. Using photo interpretation, vegetated and non-vegetated image objects were selected and stored as samples to create membership functions for vegetated and non-vegetated classes separately. The membership functions for each class were used in a nearest neighbor classifier where each image object was assigned as vegetated or non-vegetated. Any objects with NDVI<0 were assumed to be non-vegetated and assigned to the non-vegetated class, and objects with NDVI>0 were assigned to the vegetated class. The vegetated class was extracted and re-segmented using finer scale shape and compactness parameters. Using photointerpretation, image objects for the grassland and tree classes were collected and stored as membership functions. These samples were used in a nearest neighbor classifier to assign image objects to the grassland or tree class. Once the grassland and tree classes were generated, the image was analyzed to determine threshold values of mean NDVI and mean Sobel of image objects that would best separate grassland from woody encroachment and trees from woody encroachment. These threshold values were used to map woody encroachment in the grassland and tree classes. Next, small clumps of grassland and woody encroachment that were enclosed by trees were recoded to the tree class. The merge region function was used to group image objects for each class into single image objects. Then the merged objects were export into a geotiff file format. Rule-Set 3: The same methods were used as in Rule-Set 2 with the addition of the nDSM. . The goal was to determine whether adding the nDSM at the end of the rule-set would improve the VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 59 mapping of the vegetation types. Using the same values as in Rule-Set 1, multi-level segmentation was used to reassign image objects based on height information in the nDSM. For example, image objects with nDSM<0.2m were recoded to grassland, image objects 0.2m–2.0m were recoded to woody encroachment, and image objects ≥2.0m were recoded to trees. Next, small clumps of grassland and woody encroachment that were enclosed by trees were recoded to the tree class. The merge region function was used to group image objects for each class into single image objects. Then the merged objects were export into a geotiff file format. Rule-Set 4: Rule-Set 2 was used as a starting point to map vegetation types in 2006 for the South subset. Initially the rule-set was run using membership functions developed using the 2010 data, but the results were poor. This was largely due to differences in data collection, with the 2006 aerial images acquired a month earlier in the growing season. Therefore, new membership functions were built using image objects based on the 2006 aerial imagery. A.2.4. Accuracy Assessment

Accuracy assessment was performed using field data collected in February and August of 2012. Assessment was performed on the combined subsets mapped for each of the three rule-sets created and then for the final rule-set used to map the entire study area. GPS readings and vegetation types were recorded at various locations on the installation and used as validation or reference sites (Figure A.5). Data were collected at 213 sites. For woody encroachment, canopy height and density were visually estimated and recorded. Field points were overlaid on aerial imagery and inspected for positional accuracy. Several points were moved several meters in order to be in the correct location, such as a patch of woody encroachment or inside a forest stand. Due to potential positional error in GPS readings, a three-by-three window was generated around each GPS point. Within that point, a focal majority of the vegetation class mapped was calculated and compared to the vegetation class recorded in the field. An error matrix, errors of commission and omission, and an overall accuracy level are reported. Errors of commission identify the degree to which a vegetation class was overestimated, such as the percent of reference sites that were mapped as forest when they should have been mapped as encroachment or grassland. Errors of omission identify the degree to which a vegetation class was underestimated, such as the percent of reference sites that were mapped as encroachment or grassland when they should have been mapped as forest. The overall accuracy is calculated by summing the correctly mapped reference sites and dividing by the total number of reference sites. A.3. Results

A.3.1. Subset Mapping

Figures A.5–A.8 show the raw 2010 NAIP aerial imagery as a false color composite and the associated mapping results for each of the rule-sets developed for the Central and Sout subsets. There are differences between the maps for both subsets and in the magnitude of differences varies between the two subsets. As mentioned previously, the South subset contains grasslands dominated by native prairie, woody encroachment, and forests and trees. Overall, the trends in woody encroachment are similar among the maps (Figures A.6–A.8). The most noticeable difference among the maps for the South subset is the more solid, contiguous areas of woody encroachment and forest mapped using Rule-Set 1. There is less total area mapped as woody encroachment using

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 60

Rule-Set 2 and 3 than using Rule-Set 1. A visual comparison between the aerial imagery and the maps indicates many of the woody encroachment areas were mapped.

FIGURE A.5 False-color composite (NIR, Red and Green bands) of the South subset in FRMR. Vegetated areas are shown in gradations of red and non-vegetated areas are shown in gradations of blue. Numbers indicate training areas.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 61

FIGURE A.6 Rule-Set 1 mapping results for the South subset of the FRMR.

FIGURE A.7 Rule-Set 2 mapping results for the South subset of the FRMR.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 62

FIGURE A.8 Rule-Set 3 mapping results for the South subset of the FRMR.

The Central subset had more problems with the classification of vegetation types (Figures A.9– A.11). This area is dominated by grassland that was formerly cropland, woody encroachment, and to a lesser extent, forest and trees. Knowledge of the area mapped indicates woody encroachment was substantially overestimated using all three rule-sets. A second observed issue is with forests mapped using Rule-Set 2 and 3, specifically in Training Areas 49, 52, 55, and 56. These areas contain high NDVI values and, in some cases, high Sobel edge detection values. This area is representative of many of the former cropland areas at FRMR. NDVI and Sobel widely varied across former cropland areas, making it much more difficult to separate woody encroachment from grasslands. The areas with high NDVI likely reflect areas that are dominated by weedy forbs. Unfortunately, the use of the nDSM in Rule-Set 1 and 3 did not better separate grassland and woody encroachment in these areas.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 63

FIGURE A.9 False-color composite (NIR, Red and Green bands) of the Central subset in FRMR. Vegetated areas are shown in gradations of red and non-vegetated areas are shown in gradations of blue. Numbers indicate training areas.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 64

FIGURE A.10 Rule-Set 1 mapping results for the Central subset of the FRMR.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 65

FIGURE A.11 Rule-Set 2 mapping results for the Central subset of the FRMR.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 66

FIGURE A.12 Rule-Set 3 mapping results for the Central subset of the FRMR.

A comparison of accuracy levels reveals that Rule-Set 1 mapped the vegetation types more accurately than the other two rule-sets (Table A.1). All rule-sets had relatively high omission errors for woody encroachment, meaning encroachment was underestimated in some areas. Error matrices are presented in Tables A.2–A.4. Columns in the error matrix show vegetation classes identified in the field (reference sites); rows show the corresponding mapped vegetation class for reference sites. The major diagonal, shown in bold, are reference sites that were mapped correctly. For Rule-Set 1, woody encroachment was confused more with grassland than with trees (Table A.2) while for Rule-Set 2 and Rule-Set 3, woody encroachment was confused slightly more with trees than with grassland (Tables A.3 and A.4).

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 67

TABLE A.1 Accuracy assessment results for Rule-Set 1–3 for the two subset areas combined. Errors of omission and commission are reported along with overall accuracy levels. Rule-set 1 had the highest overall accuracy level.

Woody Measure Grassland Trees/Forest Encroachment

Sample Size 38 14 85

Omission Error 39.5% 28.6% 45.9%

Rule-Set 1 Commission Error 58.9% 23.1% 28.1%

Overall Accuracy 57.3%

Omission Error 50.0% 0% 56.5%

Rule-Set 2 Commission Error 53.7% 69.6% 24.5%

Overall Accuracy 51.5%

Omission Error 50.0% 0% 54.1%

Rule-Set 3 Commission Error 50.0% 70.83% 22.0%

Overall Accuracy 52.9%

TABLE A.2 Error matrix for Rule-Set 1. Using this rule-set, woody encroachment was confused more with grassland than with trees (shaded cells).

Reference Non- Row Classes Mapped Vegetated Grassland Trees Woody Total

Non-Vegetated 1 1 0 3 5

Grassland 0 23 0 33 56 Trees 0 0 10 3 13

Mapped Woody 0 14 4 46 64 Column Total 1 38 14 85 138

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 68

TABLE A.3 Error matrix for Rule-Set 2. Using this rule-set, woody encroachment was confused more with trees than with grasslands (shaded cells).

Reference Non- Row Classes Mapped Vegetated Grassland Trees Woody Total

Non-Vegetated 1 1 0 0 2

Grassland 0 19 0 22 41 Trees 0 6 14 26 46

Mapped Woody 0 12 0 37 49 Column Total 1 38 14 85 138

TABLE A.4 Error matrix for Rule-Set 3. Using this rule-set, woody encroachment was confused more with grassland than with trees (shaded cells).

Reference Non- Row Classes Mapped Vegetated Grassland Trees Woody Total

Non-Vegetated 1 1 0 0 2

Grassland 0 19 0 19 38 Trees 0 7 14 27 48

Mapped Woody 0 11 0 39 50 Column Total 1 38 14 85 138

Rule-Set 2 was applied to 2006 4-band aerial imagery in an effort to detect changes in the extent and location of woody encroachment in the South subset (Figure A.13). The South subset was selected because woody encroachment was mapped better in that area than in the Central subset. The most noticeable difference in the 2006 map is the extent mapped as non-vegetated and woody encroachment. For many areas, grasslands in early June were either not very photosynthetically active or possibly had been mowed the previous fall. As a result, these areas had low NDVI values, similar to non-vegetated image objects, and thus were mapped as non-vegetated. In comparison to the 2010 maps, fewer areas were mapped as woody encroachment. Again, this could have been due to earlier date of imagery in 2006. Without field data to use as reference sites in an accuracy assessment, it is difficult to discern how accurately woody encroachment was mapped, whether it was under- or over-estimated, and whether an earlier date of imagery (e.g., early June) maps woody encroachment more accurately than the later date from 2010.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 69

FIGURE A.13 Rule-Set 4 mapping results for the South subset of the FRMR.

A.3.2. Entire FRMR Mapped Using Rule-Set 1

Even with the noted issue of overestimating woody encroachment in former cropland areas, we applied Rule-Set 1 to produce a map of the entire installation (Figure A.14). As we anticipated, woody encroachment mapped more accurately in native prairie than in former cropland. The overall accuracy level for the FRMR using Rule-Set 1 is 62.8%, 5.5% higher than what was mapped in the subsets.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 70

FIGURE A.14 Rule-Set 1 mapping results for the entire FRMR.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 71

TABLE A.5 Accuracy assessment results for Rule-Set 1 for the entire FRMR. Errors of omission and commission are reported along with overall accuracy levels.

Woody Measure Grassland Trees/Forest Encroachment

Sample Size 76 27 110

Entire Omission Error 29.0% 40.7% 40.9% FRMR Commission 43.2% 20.0% 30.9% Error

Overall Accuracy 62.8%

TABLE A.6 Error matrix for Rule-Set 1 applied to the entire FRMR. Using this rule-set, woody encroachment was more confused with grassland than with trees (shaded cells).

Reference Data Non- Row Classes Mapped Vegetated Grassland Trees Woody Total

Non-Vegetated 2 3 0 1 6

Grassland 0 54 1 40 95 Trees 0 0 16 4 20

Mapped Woody 0 19 10 65 94 Column Total 2 76 27 110 215

A.4. Discussion

A.4.1. Change Detection of Woody Encroachment

Figures A.15 and A.16 show the woody encroachment mapped using 2006 and 2010 data, respectively. All other classes mapped were combined to create the “non-woody encroachment” class. Figure A.17 shows the change in woody encroachment between the two time periods. The change map shows areas where woody encroachment did not change, where non-woody areas changed to woody, and where woody areas changed to non-woody. This map illustrates the potential to map change over time. It is difficult to assess the accuracy of this map without an accuracy assessment of the 2006 map. Furthermore, it is questionable whether the 2010 map, with moderate to low accuracy levels, is suitable for change detection. Preferably both maps would have higher accuracy levels in order to be used for assessment and management, especially with this fine scale of mapping.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 72

FIGURE A.15 Woody encroachment mapped in the South subset for 2006.

FIGURE A.16 Woody encroachment in the South subset for 2010.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 73

FIGURE A.17 The change in woody encroachment from 2006 to 2010 in the South subset of FRMR.

A.4.2. Limitations & Future Directions

Most LiDAR are acquired during leaf-off conditions. While forest canopies were dense enough to provide data returns during leaf-off, the returns do not represent dense leaf-on canopies and, as a result, produce shorter canopy heights than would be expected and, in some areas, underestimate tree and forest spatial extents. Additional rules were utilized in rule-sets to expand the forest extent to be more representative. Woody encroachment canopies during leaf-on and leaf-off are less dense than forest and, as a result, there were many woody encroachment areas where no returns were acquired. Furthermore, for woody encroachment areas where returns were available, the nDSM typically underestimated canopy height, making it difficult to identify an appropriate canopy threshold height for classification. As a result, the height thresholds used in the classification tend to be lower than what is found in the landscape. Woody encroachment in the southern portion of FRMR was more separable than in the central portions and in other areas that were formerly cropland. Figure A.18 shows a small area visited during February 2012 and shows most reference sites were accurately mapped as forest, woody encroachment, and grassland. The results suggest that woody encroachment can be mapped and monitored in areas of native prairie using the methods described herein. However, as we observed, rule-sets developed for one year may not produce optimal results in another year without modifying the rule-set, such as collecting new data samples for the image classification. This is largely due to the variation in temporal period of the aerial imagery, which directly reflects differences in the phenologic stages of the vegetation types being mapped.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 74

FIGURE A.18 a. False color composite of 2010 aerial imagery for area where data were collected in February 2012. This area is located in the southern portion of the FRMR, where native prairie is the dominant grassland and woody encroachment is visually distinct; b. The map shows how the vegetation types are mapped relatively well in this area.

Areas in the FRMR that were formerly cropland vary widely in spectral and texture characteristics. These areas are of varying grassland quality with many areas containing large patches of weedy, annual forbs. These areas have relatively high NDVI and Sobel edge detection values that are similar to areas of woody encroachment. Field data were collected in former cropland and included field sites for grassland and woody encroachment. Visual inspection of these field sites showed little or no distinction between grassland/herbaceous vegetation and woody encroachment (Figure A.19). Due to the inability to distinguish visually between woody encroachment and grassland (former cropland), it was too difficult to sample image objects of woody encroachment in these areas to be used to calculate membership functions for a nearest neighbor classification. Because interpreting woody encroachment was possible in the South subset where native prairies exist, membership functions were built using samples from that area and applied to the Central subset containing former cropland. The results were poor and thus, one of many rule-sets not presented in any detail in this report. Because of the inability to build a sample of woody encroachment with any confidence, multi-level segmentation of NDVI, Sobel edge detection, and the nDSM were used to map woody encroachment.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 75

FIGURE A.19 a. False color composite of 2010 aerial imagery for area where data were collected in August 2012. This area is located in the central portion of the FRMR, where grasslands were former cropland and woody encroachment is difficult to distinguish, especially in the north-eastern portion of the image; b. Woody encroachment is overestimated in this area.

More work is needed to understand the complexity of mapping woody encroachment in former cropland areas. Future mapping efforts could focus on collecting field sites across the FRMR, with more focus in former cropland areas, to be used for training the classification, versus using photo interpretation of image objects. eCognition provides a multitude of image object parameters that, due to time constraints, were left unexplored. eCognition is unique from other image processing software packages. It is a powerful software package that allows users an almost infinite number of combinations to develop rule-sets. While this is an advantage of the software, it also can be a disadvantage in that there is no “standard” classification approach to adopt, and developing tailored rule-sets is time-consuming. The rule-sets used to produce land use or land cover maps are specifically tailored to the application and study area, making almost every rule in the rule-set a trial and error process. While time intensive, eCognition offers a classification approach, and many applications have shown it to exceed the mapping ability of traditional per-pixel classification approaches.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 76

Literature Cited

Agri-Service Associates, Inc. 1985. Fort Riley land use inventory and condition survey project report. Agri-Service Associates, Inc., Manhattan, KS. 2 pp + appendices and maps.

Conner R. W. and C. A. Harlow. 1980. A Theoretical Comparison of Texture Algorithms. IEEE Transactions on Pattern Analysis and Machine Intelligence. PAMI-2(3):205-222.

Freeman, C. C. and J. M. Delisle. 2004. Vegetation of the Fort Riley Military Reservation, Kansas. Open-file Report No. 119. Kansas Biological Survey. Lawrence, KS. 110 pp.

Haralick R. M, K. Shanmugan, and I. Dinstein. 1973. Textural Features for Image Classification. IEEE Transactions on Systems, Man and Cybernetics. SMC-3(6):610–621.

Huang, X., L. P. Zhang, and L. Wang. 2009. Evaluation of morphological texture features for mangrove forest mapping and species discrimination using multispectral IKONOS imagery. IEEE Transactions on Geoscience Remote Sensing Letters 6:393–397. Kim, M., M. Madden, and T. Warner. 2009. Forest type mapping using object-specific texture measures from multispectral IKONOS imagery: segmentation quality and image classification issues. Photogrammetric Engineering and Remote Sensing 75: 819–829. Lu, D. and Q. Weng. 2005. Urban classification using full spectral information of Landsat ETM+ imagery in Marion County, Indiana. Photogrammetric Engineering and Remote Sensing 71:1275–1284. Myint, S. W., P. Gober, A. Brazel, S. Grossman-Clarke, and Q. Weng. 2011. Per-pixel vs. object- based classification of urban land cover extraction using high spatial resolution imagery. Remote Sensing of Environment 115(5): 1145-1161. Platt, R.V. and L. Rapoza. 2008. An evaluation of an object-oriented paradigm for land use/land cover classification. The Professional Geogr. 60: 87-100. Quick Terrain Modeler: LiDAR Exploitation Software by Applied Imagery, http://www.appliedimagery.com/index.php.

Thomas, N., C. Hendrix, and R. G. Congalton. 2003. A comparison of urban mapping methods using high-resolution digital imagery. Photogrammetric Engineering & Remote Sensing 69(9):963–972. Trimble. 2012. eCognition® Developer 8.7.2. Reference Book. 1, D-80339. München Germany: Trimble Germany GmbH, Trappentreustr., pp. 313-328.

Wang, L., W. Sousa, and P. Gong. 2004. Integration of object-based and pixel-based classification for mangrove mapping with IKONOS imagery. International J. of Remote Sensing 25(24):5655-5668.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 77

Whiteside, T. G., G. S. Boggs, and S.W. Maier. 2011. Comparing object-based and pixel-based classifications for mapping savannas. International Journal of Applied Earth Observation and Geoinformation 13(6):884-893. Wijaya, A. and R. Gloaguen. 2007. Comparison of multisource data support vector Machine classification for mapping of forest cover, In: Proceedings from the Geoscience and Remote Sensing Symposium, IGARSS 2007. IEEE International . 1275-1278.

Yaun, F. and M. E. Bauer. 2006. Mapping impervious surface area using high resolution imagery: A comparison of object-based and per pixel classification, In: ASPRS Annual Conference Proceedings, May 1-5, 2006, Reno, Nevada. Unpaginated CDROM, 8pp. US Army, Fort Riley, Kansas – Environmental Division, Directorate of Public Works. 2010. Integrated Natural Resources Management Plan, Fort Riley, Kansas. 120 pp + Appendix A- M.

USACE. 1985. Land utilization vegetation and condition survey, Ft. Riley, Kansas. U.S. Army Corps of Engineers, Kansas City District.

VEGETATION SURVEY AND MAPPING OF THE FT. RILEY MILITARY RESERVATION PAGE 78