Shrub Control and Water Yield on Texas Rangelands: CURRENT STATE OF KNOWLEDGE

Transcription

1 Shrub Control and Water Yield on Texas Rangelands: CURRENT STATE OF KNOWLEDGE 2005 Texas Agricultural Experiment Station Research Report 05-1

2 AUTHORS Bradford P. Wilcox, Professor, Department of Rangeland Ecology and Management and Texas Agricultural Experiment Station, Texas A&M University, College Station, TX William A. Dugas, Resident Director of Research, Texas Agricultural Experiment Station, Blackland Research Center, Temple, TX M. Keith Owens, Professor, Texas Agricultural Experiment Station, Texas A&M University Agricultural Research and Extension Center, Uvalde, TX Darrell N. Ueckert, Regents Fellow and Professor, Texas Agricultural Experiment Station, Texas A&M University Agricultural Research and Extension Center, San Angelo, TX Charles R. Hart, Associate Professor and Extension Range Specialist, Texas Cooperative Extension, Texas A&M University Extension Center, Fort Stockton, TX ACKNOWLEDGMENTS The authors express appreciation to Larry Butler (U.S. Department of Agriculture-Natural Resources Conservation Service), Peter Ffolliott (University of Arizona), Wendy Gordon (Texas Parks and Wildlife Department), Rebecca Griffith (U.S. Army Corps of Engineers), Lee MacDonald (Colorado State University), Bill Mullican (Texas Water Development Board), Johnny Oswald (Texas State Soil and Water Conservation Board), Tom Thurow (University of Wyoming), Karl Wood (New Mexico State University), and numerous colleagues in the Texas Agricultural Experiment Station and Texas Cooperative Extension for their thoughtful reviews of this paper. The encouragement and support of Ron Lacewell and Bob Whitson for the preparation and publication of this report and the editing skills of Ann Shurgin and the design skills of Steven Keating (Agricultural Communications) are also greatly appreciated. Special appreciation is extended to all the scientists listed in the references section, whose research findings made this synthesis possible.

3 Texas Agricultural Experiment Station Research Report 05-1 TABLE OF CONTENTS [ Executive Summary 2 ] [ Introduction 3 ] [ A Conceptual Framework: Four Generalizations 4 ] [ Vegetation Types and Geographic Zones 5 ] - Saltcedar in Riparian Zones 5 - Ashe Juniper on Rangelands with Subsurface Flow 8 - Honey Mesquite on Uplands or Outwash Plains 11 with Little Subsurface Flow [ Ongoing Activities 13 ] [ Needed Research 15 ] [ Conclusions 17 ] [ References 18 ] This report was funded in part by the Joe Skeen Institute for Rangeland Restoration, a joint effort of the Texas Agricultural Experiment Station, New Mexico State University, and Montana State University. The Joe Skeen Institute is supported by the U.S. Department of Agriculture-Cooperative State Research, Education, and Extension Service, and by the U.S.D.A. Forest Service. p. 1

4 EXECUTIVE SUMMARY The disparity between the supply and demand for water in Texas has focused attention on brush control as a method for increasing the water supply for Texans. This review (1) summarizes how shrubs affect the water cycle at the tree, hillslope, small catchment, and landscape scales and (2) examines whether broadscale reductions in woody plant cover might increase the water supply. The greatest information need is at the landscape scale, but research at this scale is costly and difficult. Hence, most field investigations have focused on the tree or hillslope scales. However, extrapolating these data to larger scales is problematic because hydrological processes may change as the scale increases. A review of the scientific literature has allowed the authors to make the following broad generalizations: 1. The relationship between shrub removal and increased water yields becomes stronger as annual rainfall increases. 2. The linkage between shrub removal and increased water yields is weaker in upland areas and stronger in areas adjacent to stream channels, where shrubs may be using shallow groundwater that is hydrologically connected to stream flow. 3. In upland areas, the linkage between removal of woody plants and increased water yield is stronger where water can move rapidly through the soil or parent material to recharge springs or shallow aquifers. 4. In areas where there is little subsurface water movement and woody plants are not accessing groundwater, shrub control is unlikely to significantly affect either groundwater recharge or stream flow. The data suggest that water can be salvaged by controlling dense stands of saltcedar in riparian zones, dense stands of Ashe juniper in areas with rapid subsurface flow (e.g., where springs are present), and mesquite on soils that develop deep cracks when dry. However, in each case there is considerable uncertainty with respect to the amount of water that can be salvaged and the fate of the salvaged water. Hydrologic models have predicted substantial increases in water yields following shrub control, but these results must be used with caution. There is a threefold variation in the predicted water yield increases from different models, and in some cases the predicted increases are at least three times larger than the measured values. More research is needed to fully understand the extent to which shrub control can increase water yields. Key gaps include community-level evapotranspiration studies, isotope studies to determine the sources of water used by shrubs, groundwater surface water connections, nested watershed studies, and landscape scale studies on saltcedar in various habitat types. The hydrological condition of Texas watersheds and the partitioning of water within the hydrological cycle are determined by complex interactions between soils, vegetation, and climate. Brush encroachment is only one change that has affected the health of Texas watersheds droughts, excessive grazing, water impoundments, urbanization, and water transfers also have altered the hydrological cycle. It would be valuable to shift the debate away from a focus on brush control and water yields to a broader assessment of best management practices for improving watershed health and sustainability. p. 2

5 INTRODUCTION The increasing Texas population and associated municipal and industrial growth are placing greater demands on the state s water supply. The issue of available supply is particularly acute during times of drought, as was learned during the drought of the late 1990s to Texas will have to either provide more water or reduce the amount being used. One potential mechanism for increasing water supply is to reduce the amount of water consumed by shrubs and trees on Texas rangelands. During the past century, the density and coverage of shrubs have increased dramatically on Texas rangelands, and many former grasslands or savannas have now converted to shrublands or closed-canopy woodlands. This report examines how changes in shrub density and coverage affect the water cycle, and the extent to which broadscale reductions in woody plant cover might increase water yields in Texas. The increasing Texas population and associated municipal and industrial growth are placing greater demands on the state s water supply. Texas will have to either provide more water or reduce the amount being used, especially during drought conditions. A change in rangeland vegetation cover can have a large effect on the water cycle. The most direct and least desirable effect occurs when herbaceous vegetation cover is lost due to excessive grazing or drought. In this case, more of the precipitation becomes surface runoff, particularly during intense rainfall events. Although this increase in surface runoff might increase water yields, this is never desirable because of the accelerated erosion and potentially severe degradation of water quality, stream channels, and aquatic ecosystems. A more desirable way of increasing water yields (stream flow and groundwater recharge) is to manage vegetation to decrease total evapotranspiration. Total evapotranspiration is composed of interception of water by vegetation and mulch and subsequent evaporation from these surfaces, transpiration by plants, and evaporation from the soil surface. A reduction in total evapotranspiration generally increases the amount of water that percolates below the root zone into groundwater, which may become stream flow. Given the undesirability of increasing surface runoff, this paper explores the potential to reduce total evapotranspiration and thereby increase water supplies. The focus is on Texas rangelands, including riparian zones. One reason for the controversy about the effects of shrubs on the water budget is that the scale at which information is needed is often different from the scale at which data are collected. Society is often most interested in data collection at the landscape scale because this is the scale at which water supplies are evaluated and measured. Unfortunately, this scale has the least information available on how vegetation may affect stream flow and recharge, and it is difficult to conduct experiments at the landscape scale. Most data on shrub water use have been collected at the tree or small-plot scales, and attempts have been made to extrapolate these data to larger scales. Such extrapolations introduce uncertainty and require a sound understanding of how hydrologic processes may change with increasing spatial scale. One way to raise the level of confidence in extrapolation is to compare estimates made at multiple scales. This review summarizes the current state of scientific knowledge at the following spatial scales: (1) individual tree or small plot, (2) hillslope or stand, (3) small catchment, and (4) landscape. It is important to recognize that the hillslope scale can encompass many trees or shrubs, and this increase in scale integrates key hillslope processes such as overland flow, depression storage, and sediment deposition. Small catchments incorporate the additional processes of channel flow and groundwater recharge. The landscape scale encompasses watersheds of several square miles. p. 3

6 One potential mechanism for increasing water supply is to reduce the amount of water consumed and intercepted by shrubs and trees on Texas rangelands by controlling these plants. A CONCEPTUAL FRAMEWORK: FOUR GENERALIZATIONS A review of the scientific literature allows four broad generalizations about vegetation control and water yields. The first generalization is that the relationship between the removal of upland shrubs or trees and water yields becomes stronger as annual rainfall increases. Conversely, the lower the rainfall, the lower the potential for increasing water yields by managing vegetation. A common rule of thumb is that annual rainfall must be at least about 18 in. for vegetation removal to increase water yields (Hibbert 1983). This rule of thumb was derived from research on chaparral rangelands in Arizona and California. However, there may be no lower precipitation limit for riparian areas with shallow groundwater levels, because there is a more direct and immediate linkage between a reduction in transpiration and a potential increase in stream flow. Humid landscapes typically offer more opportunity than semiarid landscapes for manipulating vegetation to increase water yields, all other factors being equal. There is extensive literature showing that stream flow increases following a reduction in woody plant cover, but most of this research has been conducted in areas receiving, on average, 30 in. or more of rain a year. Relatively few studies have shown that stream flow can be increased by reducing the cover of woody plants in semiarid landscapes. This is because there is sufficient energy to evaporate most of the precipitation that falls in semiarid areas, so most of the rainfall is lost to evapotranspiration regardless of the vegetation present, unless the saved water is quickly transferred to a stream or is stored in a shallow aquifer that is not available for plant transpiration. The second generalization is that the linkage between shrub removal and increased water yields is generally stronger in riparian areas than in upland areas because trees in riparian areas may be directly accessing the shallow groundwater that is often hydrologically connected to streams, rivers, or lakes. However, efforts to reduce the amount of woody vegetation in riparian zones may be undesirable because of the potential adverse effects on stream-bank stability, wildlife habitat, aesthetics, and biodiversity. An important exception is areas where saltcedar (Tamarix spp.) an aggressive alien invader is present in high densities along streams and around shorelines. The third generalization is that for upland areas, the linkage between removal of woody plants and increased water yields is stronger in areas where water can move rapidly through the soil or parent material (e.g., in areas where springs currently exist or historically have existed). Rapid downward flow can occur in areas with deep sandy soils or fractured limestone. These areas often support intermittent or perennial springs, seeps, and streams. In these areas, subsurface water that would otherwise become recharge or stream flow could be available for deep-rooted shrubs to transpire. The fourth generalization is that in areas where there is little subsurface water movement and woody plants are not accessing groundwater, reducing woody plant cover is less likely to increase either groundwater recharge or stream flow unless direct runoff is increased. In these areas, most of the precipitation infiltrating the soil remains in the upper portion of the soil profile and is available for evaporation or use by herbaceous vegetation. p. 4

7 VEGETATION TYPES AND GEOGRAPHIC ZONES Three species exemplify the generalizations listed above: saltcedar in riparian zones, Ashe juniper (Juniperus asheii) in areas with subsurface flow, and honey mesquite (Prosopis glandulosa var.glandulosa) in upland areas with little subsurface flow. These examples can be extended to other species and other geographic zones, but in Texas these other species and zones are of much less importance. The linkage between shrub removal and increased water yields is stronger in areas adjacent to stream channels, where dense stands of saltcedar may be using shallow groundwater that is hydrologically connected to stream flow. Saltcedar in Riparian Zones Many of the semiarid riparian areas in Texas have become dominated by dense infestations of saltcedar. Saltcedar was introduced in the past as an ornamental plant and to stabilize stream banks. It often forms dense canopies that largely exclude native vegetation, and these stands are not suitable habitat for most native wildlife. Saltcedar densities are often much greater than those of native phreatophytic (having deep roots that reach the water table) trees, and saltcedar has colonized areas along streams and lakes where native phreatophytic trees were not extensive. Although the effect of saltcedar along reservoirs is a major concern for many Texas municipalities, little is known about its effect on the water budget in these locations. In this review, the results of saltcedar water-use studies in riparian areas along streams at the tree, stand, and landscape scales are presented. Most of the relevant research has been conducted in Arizona, California, Nevada, and New Mexico, rather than in Texas. A comprehensive review of water use by saltcedar, at least at the tree and stand scales, can be found in Glenn and Nagler (2005). Tree Scale Water-use measurements at the leaf and tree levels are critical for understanding the physiology of saltcedar and how environmental conditions affect such key processes as transpiration. Tree-level measurements are typically made with sap flow gauges or lysimeters. Sap flow gauges measure the flux of water past a point in the stem, and these data often are scaled to the entire tree, based on leaf or sapwood area. Lysimeters require the tree to be in a large pot that is placed in the field. The roots are confined to the pot, and evapotranspiration can be calculated from the changes in weight or soil moisture. Saltcedar may be a large consumer of water. An early report stated that a saltcedar tree could use 200 gal. of water per day. Although this number has been widely cited, the authors were unable to find the original citation to verify how the estimate was derived. Extrapolating from short periods of p. 5

8 observations to longer time periods, or from trees to stands, can lead to very high water-use estimates (e.g., as much as 111 in. per year [van Hylckama 1974; Gay and Fritschen 1979]). Studies at the leaf- and tree-level scales have shown that saltcedar and native riparian woody phreatophytes use similar amounts of water. Nagler et al. (2003) showed that saltcedar, cottonwood (Populus spp.), and willow (Salix spp.) trees that were ft tall all used gal. of water per day. They concluded that saltcedar is not the water spender that has been anecdotally reported. This conclusion is supported by an isotopic study showing that saltcedar was at least as efficient in its water use as native woody plants along the Colorado River (Busch and Smith 1995). In southern Nevada, stands of four different phreatophytes had similar transpiration rates on a leaf-area basis when there was moderate to high soil water availability (Sala et al. 1996). Collectively, most research at the tree scale indicates that saltcedar trees do not inherently use more water than other phreatophytic trees, assuming leaf area per tree and environmental factors are equal. Saltcedars act like other phreatophytic plants: When moisture is available they use it at high rates, but when moisture is limited they decrease transpiration. Stand Scale A variety of techniques have been used to estimate water use by saltcedar at the stand level. These include sap flow measurements, groundwater monitoring, large lysimeters, and micrometeorological techniques. Estimated values for stand-level water use have ranged from 17 to 67 in. per year (Gay and Fritschen 1979; Weeks et al. 1987; Devitt et al. 1998). Differences in techniques and stand densities have resulted in such large discrepancies in water-use estimates that there is no consensus on saltcedar water use at the stand level. Micrometeorological techniques, such as the energy balance (Dugas et al. 1998) or the eddy covariance (Dahm et al. 2002) technique, measure latent heat fluxes (evapotranspiration) from a plant community. Both methods require uniform, upwind conditions (fetch) from the sensors, as well as large areas with and without the plant species of interest. Many riparian communities often lack adequate fetch, making it difficult to apply these methods to compare water use of saltcedar versus that of other riparian plants. Dahm et al. (2002) used the eddy covariance method to estimate season-long evapotranspiration from several riparian communities along the middle Rio Grande in New Mexico. Saltcedar stands in floodplains had higher evapotranspiration rates (43 48 in. per year) than saltcedar stands in areas that did not flood (30 in. per year). Evapotranspiration from a mixed stand of cottonwood was 48 in. per year, as compared to 38 in. per year for a mature cottonwood stand. Another study using micrometeorological methods reported that saltcedar stands used 30 in. of water in a dry year and 58 in. in a wet year (Devitt et al. 1998). Increasing salinity generally makes it more difficult for saltcedar to uptake and transpire water; thus, the salt content of shallow groundwater also affects the amount of water transpired (Busch and Smith 1995; Glenn et al. 1998; Smith et al. 1998). p. 6 Dense stands of saltcedar along the shorelines of reservoirs are a major concern for many Texas municipalities, but little is known about saltcedar s effect on the water budget in these locations.

9 Other work at the stand scale indicates that water use by saltcedar is comparable to that for native phreatophytes (Glenn and Nagler 2005). The main variable determining stand-level water use in many communities is the total amount of leaf area. As indicated by Sala et al. (1996) and Anderson (1982), saltcedar will only transpire more water than native phreatophytes when it has a higher stand density or leaf-area index, which often occurs when saltcedar dominates riparian zones. Studies generally agree that saltcedar growing in areas with shallow water tables transpires more water than plants growing in areas with deep water tables (van Hylckama 1970; Carman and Brotherson 1982; Horton et al. 2001). As a result, saltcedar growing in close proximity to rivers or streams would transpire more water than plants growing further from the streambed or water source (Devitt et al. 1997; Hays 2003). As saltcedar stands mature and develop a dense monoculture, water use increases with increasing stand size and density (Davenport et al. 1982; Sala et al. 1996; Devitt et al. 1997). Landscape Scale Only a few studies have attempted to measure water use by saltcedar at the landscape level, because one must measure both water inputs (precipitation, surface water and groundwater inflows) and outflows (change in soil water storage, surface and subsurface outflows) for each stream or river segment (Goodrich et al. 2000). In the first study, saltcedar was removed from the bottomlands of Arizona s Safford Valley to determine potential water savings (Gatewood et al. 1950). The results indicated that 9,300 acres of phreatophytic vegetation, primarily saltcedar, used approximately 28,000 acre-ft of water per year. Welder (1988) documented the hydrologic effect of a large-scale saltcedar control program that was initiated in 1967 along the 82-mile Acme-Artesia reach of the Pecos River in New Mexico. A total of 21,500 acres of saltcedar was killed by bulldozing or root plowing, but saltcedar was not controlled within 50-ft-wide buffer zones along both sides of the river. Although a companion study (Weeks et al. 1987) found that water use by saltcedar at the stand scale was about 12 in. per year greater than the replacement vegetation, Welder reported no detectable change in stream flow as a result of clearing the saltcedar. Weeks et al. speculated that any potential increase in base flow may have been offset by groundwater pumping, or by unusually dry and wet years following saltcedar clearing. One possibility is that evapotranspiration from saltcedar remained high following the mechanical treatment because of the rapid regrowth of saltcedar. The available literature indicates it is probable but not conclusive that where saltcedar stands are dense, extensive, and close to a waterway or shallow groundwater source, removing saltcedar and replacing it with native (particularly non-woody) plants may result in less overall water use. Water salvage may be an important factor in locations such as the Pecos River in Texas, where a gallery phreatophyte forest along the river was not present before saltcedar invasion and where grasses would dominate if saltcedar were killed, or in lake basins with receding water levels invaded by dense stands of saltcedar. The natural grasslands and scattered woody plants along the river with less leaf area, shallower roots, and a shorter growing season would probably use less water than saltcedar trees. The authors are not aware of any data in the peer-reviewed literature on the amount of water used by herbaceous vegetation that has replaced saltcedar stands in riparian zones. p. 7

10 Ashe Juniper on Rangelands with Subsurface Flow Springs on Texas rangelands are commonly associated with limestone or karst geology, and these are a good indicator of subsurface flow. Two important features of these sites facilitate the presence of springs: (1) shallow soils that cannot store much water and (2) fractured parent material that allows rapid, deep infiltration of rainfall. Ashe juniper rangelands in central Texas commonly exhibit these attributes. Tree Scale Interception. Evergreen shrubs such as juniper have a large interception capacity because they retain their leaves year round and have high leaf area per tree (Hicks and Dugas 1998). Owens and Lyons (2003) estimated that about 47% of annual precipitation is intercepted by an Ashe juniper tree and its litter layer. The proportion of interception varies dramatically with storm size, as interception may be close to 100% for storms with less than 0.5 in. of rainfall and only 20% for large storms. Large storms generate most of the stream flow on Texas rangelands and provide most of the aquifer recharge. Transpiration. Transpiration from an Ashe juniper community is expected to be greater than from an herbaceous community because Ashe juniper transpires water throughout the year, typically has a much greater community leaf area, and can access water at greater depths. A mature Ashe juniper can transpire as much as 33 gal. of water per day, or about in. per year (Owens and Ansley 1997). These data show that dense stands of juniper can intercept and transpire large quantities of water and thereby have a major impact on the water budget in semiarid regions. If juniper is removed, other vegetation will increase in density and coverage, and these species also will intercept and transpire water. Without specific information on the amount and type of replacement vegetation, however, it is difficult to predict how much water could be saved by removing juniper. Hillslope Scale Evapotranspiration. At the hillslope scale, evapotranspiration is usually estimated using micrometeorological methods. Dugas et al. (1998) measured evapotranspiration from an Ashe juniper community using the energy balance method. They took measurements from two paired areas of about 40 acres each over a 5-year period. All Ashe juniper trees were removed from one of the areas after 2 years. Average evapotranspiration from both areas for the 5-year period was about 25 in. per year. The difference in evapotranspiration (untreated minus treated) was about 1.5 in. per year in the first 2 years following treatment. The treatment effect disappeared in the third year (i.e., evapotranspiration was similar in the treated and untreated areas). Small Catchment Scale Small catchments with springs. Many springs in Texas have dried up in the past 150 years. This may be due to increased groundwater pumping (Brune 2002), an increase in woody plant cover, or other factors. There are many anecdotal accounts of springs drying up after the encroachment of woody plants and of spring flow returning after the woody plant cover was removed or A high probability of water yield increases may be associated with control of dense stands of Ashe juniper where it grows on shallow soils that cannot store much water; where fractured parent material allows rapid, deep infiltration of rainfall; and where springs are present. The karst rangelands of the Edwards Plateau are the source areas for many Texas rivers. p. 8

11 reduced. The Upper Colorado River Authority has reported many such events as a result of large-scale brush control in and around the North Concho River watershed. The only study quantifying an increase in spring flow following juniper removal was by Wright (1996). This study showed that the partial removal of Ashe juniper in an 8-acre catchment in the Seco Creek watershed caused spring flow to increase from 3.1 gal. per min during the 2-year pre-treatment period to 3.8 gal. per min following the treatment, even though there was less precipitation in the post-treatment period. This increase in flow is equivalent to about 1.5 in. per year. Increased spring flow following removal of woody plants can be very significant for livestock, local stream ecology, and wildlife, but it may not necessarily increase downstream water yields. Small catchments without springs. A few studies have examined the effect of juniper removal on small catchments where no springs were present. Richardson et al. (1979) compared runoff from two 10-acre catchments over an 11-year period. Juniper was removed from one of the catchments in the fifth year by root plowing, and this decreased surface runoff by about 20%, or 1 in. per year, due to the increase in surface roughness and resultant increase in infiltration and surfacewater storage. In another paired-catchment study, removing juniper cover by hand cutting had little influence on surface runoff from 14- and 9-acre catchments in the Seco Creek watershed (Dugas et al. 1998). Surface runoff was only about 5% of total precipitation and occurred only during high-intensity storms. Many springs in Texas have dried up in the past 150 years due to increased groundwater pumping, an increase in woody plant cover, or other factors. One study in the Seco Creek watershed documented an increase in spring flow following partial removal of Ashe juniper. Researchers in Uvalde County evaluated the influence of juniper removal on runoff from small catchments in the Edwards Plateau region, where the average annual rainfall is about 28 in. (Wilcox et al. 2005). They monitored stream flow over a 13-year period from nine catchments ranging in size from 10 to 15 acres. Following a 3-year observation period, juniper was partially or entirely removed from six of the catchments. This removal had no significant effect on runoff during two observation periods when the annual rainfall was well above average and slightly above average, respectively. Evapotranspiration and recharge were not measured. In 1988, researchers initiated runoff measurements on seven 10-acre, juniper-dominated watersheds on Edwards County rangelands (Thurow and Hester 1997). Average annual rainfall in this area is about 23 in. All of the woody vegetation was cut down and removed from some of the sites to determine the degree to which brush cover influenced water yield. The amount of runoff was similar (0.2 in. per year) from watersheds with 100% grass cover, 70% grass/30% shrub cover, and 40% grass/60% shrub cover. Thurow and Hester made the assumption that recharge in shrub-free areas would be substantially higher (3.7 in. per year) because of reduced interception by juniper. This estimate should be interpreted with caution because it was not a direct measurement but rather the result of a water balance calculation. These studies reveal that in the absence of springs, removing Ashe juniper may not increase surface runoff. Shrub removal may decrease water yield if mechanical control increases surface roughness and surface storage. p. 9

12 Landscape Scale The greatest interest in using juniper control to increase water yields is at the landscape and river basin scales, but experiments at these scales are the most difficult and expensive. For this reason, indirect methods, such as analyzing trends in stream flow and hydrologic modeling, are used to evaluate the effect of vegetation removal on water yields. Analysis of stream flow. Stream flow data from the early 1900s are available for many of the major rivers in Texas. These long-term data provide insights into the variability of stream flow, trends over time, and the relationship of stream flow to climate. The careful analysis of historic stream flow data can be used to assess the sensitivity of stream flow to landscape scale vegetation changes. For example, if woody plant cover on the Edwards Plateau increased dramatically during the last century and if stream flow concomitantly decreased, as some modeling exercises predict, the stream flow record should reflect those changes, barring any confounding changes in precipitation or land use. To date, only a few such analyses have been conducted for the Edwards Plateau. The Lower Colorado River Authority (LCRA) found no changes in stream flow on the Pedernales River from 1939 to 2000 (LCRA 2000). The analysis of stream flow records from most Texas rangelands shows that flood flows from large storms account for a very high percentage of the total annual discharge. For these events, the presence or absence of shrubs may have very little effect on the amount of runoff. Hydrologic modeling studies. A series of modeling studies have investigated the influence of shrub cover on stream flow and recharge in the Edwards Plateau region. Each of these studies has predicted an increase in water yield following broadscale shrub control. Using the Soil and Water Assessment Tool (SWAT) model, Bednarz et al. (2001) predicted that removing about 200,000 acres of shrubs from the 800,000-acre Pedernales River catchment would increase the average annual flow by about 35%. This is about 5 in. per year from the treated area. It is interesting that the examination of annual flow of the Pedernales River (LCRA 2000) did not show a decrease in annual water yield over the past 60 years as woody plant cover and density were increasing, although it is possible that 60 years ago the brush cover was already dense enough to have reduced annual water yields. Wu et al. (2001), working in the Cusenberry Draw in the western Edwards Plateau, used the Simulation of Production and Utilization of Rangeland (SPUR) model to estimate that in areas with no juniper, runoff would account for up to 15% of the precipitation, or about 3 in. in an average year. They predicted that water yields would drop to almost zero if the amount of woody plant cover increased to 20%. More recent modeling work on the Edwards Plateau predicts a smaller increase in stream flow resulting from shrub control. Afinowicz et al. (2005), also using the SWAT model, predicted a 1.5 in. per year increase in stream flow following brush control for a smaller stream, the north fork of the Upper Guadalupe River. To summarize, all modeling exercises on the Edwards Plateau have predicted that reducing the amount of shrub cover will increase stream flow and/or groundwater recharge, but the magnitude of the predicted increases differs by a factor of three. Predicted increases from earlier modeling exercises are about three times greater than the increases measured in the field. Results from more recent modeling work are in line with field observations. The results of any modeling study must be interpreted with caution, as their validity depends on the accuracy of model inputs (e.g., representative precipitation, soil properties, and pre- and post-treatment vegetation characteristics) and the model assumptions (e.g., effect of vegetation changes on surface runoff). p. 10

13 Honey Mesquite on Uplands or Outwash Plains with Little Subsurface Flow Tree Scale Mesquite has been reported to intercept much less ambient rainfall than juniper only between 15% and 30% (Navar and Bryan 1990; Desai 1992; Martinez-Meza and Whitford 1996). This difference is attributed to the following properties: Mesquite is deciduous; it has flat, waxy leaves that are much less effective in holding water than scale-like juniper leaves; and it has much less leaf area per tree. The amount of interception by mesquite canopies is similar to that for herbaceous vegetation. Water use by mesquite depends on a number of factors, including water availability and tree density. Mesquite may take advantage of both deep and shallow water. On mesquite sites in northern Texas, where deep water was not available, Ansley and co-workers found that mesquite relies principally on shallow soil water (Ansley et al. 1990; Ansley, Jacoby, et al. 1992; Ansley, Price, et al. 1992). Other work in the same area indicates that on an individual tree basis, mesquite trees are capable of using from 8 to 54 gal. of water per day (Ansley et al. 1991; Ansley et al. 1994; Ansley et al. 1998). When moisture is limited, mesquite trees may decrease transpiration rates by 35% 75% (Dugas et al. 1992). Tree density also affects water use. Ansley et al. (1998) found that the rate of water use per tree was much higher in an open savanna than for trees of the same height in dense stands. When these numbers are scaled up to a stand basis, the amount of water used by mesquite is roughly equivalent in open savannas and in dense mesquite woodlands. The amount of transpiration by mesquite also depends on whether the plants have access to groundwater. When soil water contents in a northern Texas study were low due to lack of rainfall, mesquite trees in a riparian zone where the water table was 4.9 ft deep had nearly double the transpiration rate of trees on upland or lowland sites with no access to groundwater (Cuomo et al. 1992). However, when soil water contents were higher, transpiration rates were similar at all three sites. Mesquite trees are water limited in most Texas landscapes, and they use more water when it becomes available from rainfall or shallow groundwater. Control of mesquite on upland sites where there is very little subsurface water flow is not likely to increase water yield or reduce total evapotranspiration. Mesquite is a phreatophytic species in some desert regions, depending largely on deep roots for obtaining water from belowground aquifers. Nilsen et al. (1981) and Nilsen et al. (1983) documented the use of groundwater at a depth of ft by western honey mesquite (P. glandulosa var. torreyana) in the Sonoran Desert of southern California. Phillips (1963) found live mesquite roots at a depth of about 175 ft in a gravel bed about 100 yd long and ft thick in an open pit mine about 20 miles southeast of Tucson, Arizona. The depth of groundwater at this site was 240 ft. p. 11

14 Hillslope Scale In general, larger-scale studies have demonstrated that mesquite removal has a small effect on either surface runoff or groundwater recharge. At a Rolling Plains site where the average annual precipitation was 25 in., mesquite control had minimal, if any, effect on soil water (Carlson et al. 1990). The flush of herbaceous growth following mesquite removal means that community-level evapotranspiration will not change. This is why there was no change in soil moisture following mesquite removal (Heitschmidt and Dowhower 1991; Dugas and Mayeux 1991). In southern Texas, where the average annual precipitation was 28 in., there was little difference in soil moisture storage or evapotranspiration between adjacent mesquite- and grass-dominated communities (Weltz and Blackburn 1995). A study in the Blackland Prairie of Texas, where annual precipitation is about 34 in., documented an increase in deep soil moisture following mesquite removal (Richardson et al. 1979). The measured increase was about 3.1 in. in the upper 5 ft of soil. The fact that soil water storage increased in the Blackland Prairie but not in the other two mesquite rangelands is probably explained by the formation of vertical soil cracks during dry periods at the Blackland Prairie site. These cracks allow water to move deeper into the soil profile during rainfall events, where it would not be available to herbaceous plants. Small Catchment Scale Dugas and Mayeux (1991) used energy balance techniques to measure evapotranspiration rates on mesquite rangelands in the Rolling Plains of Texas. They concluded that under circumstances of low grazing pressure and low runoff potential, honey mesquite removal would provide little if any additional water for off-site uses in the short term (161). Evapotranspiration was somewhat greater from the treated (no mesquite) site under dry conditions, but under wet conditions there was no significant difference in evapotranspiration between the mesquitecovered and mesquite-free sites. The small difference between the two sites was attributed to the vigorous growth of herbaceous vegetation following mesquite control on the treated site, a phenomenon that has been documented by other researchers (Dahl et al. 1978; Heitschmidt et al. 1986; Heitschmidt and Dowhower 1991). Landscape Scale Analysis of stream flow. An analysis of historic stream flow in the North Concho River by the Upper Colorado River Authority (UCRA 1998) showed a marked decline in stream flow since about UCRA personnel hypothesized that this decrease was a direct result of a large increase in brush density prior to By inference, removing this brush particularly in areas adjacent to streams where there is shallow groundwater would restore flows to the levels observed prior to The authors of this report speculate that the change in stream flows was at least partly due to changes in precipitation intensity and/or a reduction in excessive grazing. A reduction in storm intensity and a reduction in grazing would decrease Mesquite growing in riparian zones where the water table is shallow can have transpiration rates nearly double those of trees growing on upland sites with no access to groundwater. These large mesquite trees are about 100 feet from the North Concho River. p. 12

15 surface runoff and increase infiltration, and most of the increased infiltration would likely be lost in evapotranspiration. There is no basis for an unambiguous conclusion about the cause of the observed decrease in annual water yields. Modeling studies. Modeling studies on several major watersheds have predicted increases in stream flow following mesquite removal. For example, in the mesquite-dominated Lake Arrowhead watershed in north central Texas, where annual rainfall averages 28 in., mesquite removal was predicted to increase annual water yields by about 5 in. per year (TAES 2002). The predicted increases in water yields from modeling studies are much higher than the values observed in field studies and are inconsistent with the conceptual framework laid out in this report. These inconsistencies suggest that further study is needed to refine the models, and that additional field research is needed to document the effects of mesquite removal at larger scales. In the North Concho River watershed assessment and feasibility study (UCRA 1998), the Soil and Water Assessment Tool (SWAT) model also was used to predict changes in stream flow as a result of shrub control on the watershed. Key assumptions in these modeling scenarios were that alluvial aquifers were severely depleted as a direct result of deep-rooted mesquite and that normal flows would not be reestablished until the alluvial aquifer was sufficiently recharged. ONGOING ACTIVITIES Riparian Zones A large-scale ecosystem restoration program on the Pecos River in western Texas began in The primary treatment is the aerial application of herbicides to kill saltcedar. Spraying began in 1999, and 9,664 acres along 271 miles of the Pecos River had been treated by 2004, resulting in about 85% 90% saltcedar mortality. Monitoring efforts were initiated at the outset of the project to evaluate the effects of saltcedar control on river water salinity, stream flow, and water salvage. To date, river water salinity has not been affected, and irrigation delivery to the river from Red Bluff Reservoir was suspended in due to drought conditions (Hart et al. 2005). The stand-level water use prior to saltcedar control had been calculated from diurnal fluctuations in the water table, and this was estimated at 114 in. per year in the riparian zone, 35 in. per year at the edge of the riparian zone, and 1.6 in. per year in the uplands (Hays 2003). Preliminary results are encouraging. However, no fixed values have been derived for the amount of water salvaged, and the fate of the salvaged water has not been ascertained. Future research will continue to determine the volume and fate of salvaged water, estimate saltcedar water use with sap flow measurements, and investigate surface water and shallow groundwater interactions. Saltcedar has invaded the shoreline of the upper Colorado River, which gets its red color from runoff laden with silt and clay, largely due to natural geological erosion of soils in the Rolling Plains ecological region. p. 13

16 Ashe Juniper in Karst Areas with Subsurface Flow A number of ongoing studies in the Edwards Plateau region are examining the influence of shrubs on water yield. The USDA-Natural Resources Conservation Service and the U.S. Geological Survey are leading several watershed-scale studies in this region. The two main study areas are the Honey Creek watershed and Government Canyon. Both studies include comparisons of stream flow from watersheds with different degrees of shrub removal. Collaborators from Texas A&M University, the University of Texas Austin, and the University of Texas San Antonio are also conducting detailed investigations of evapotranspiration, interception, surface runoff, and groundwater recharge. Large-scale rainfall simulation studies are being conducted at the Honey Creek area and at Camp Bullis. Additional work, focusing on evapotranspiration, is ongoing at the Texas State University Freeman Ranch. Measurements of precipitation and runoff are being made on two 100-acre watersheds by the Texas Agricultural Experiment Station in Coryell County. Measurements were made for about 2 years before woody plants were removed from one watershed in February Preliminary results suggest it will be difficult to identify small changes in runoff following woody plant removal due to the variability between the two watersheds and the uncertainties in measuring high flows, as nearly all of the runoff is associated with large precipitation events. Redberry Juniper and Mesquite in the North Concho River Watershed The Upper Colorado River Authority (UCRA) is leading several monitoring efforts in and around the North Concho River watershed in conjunction with a large state-funded brush-control project that is treating about 400,000 acres. Stream flows are being monitored with existing U.S. Geological Survey flow gauges on the North Concho River, on seven sites on the North Concho established by UCRA, and on several more sites on tributaries (UCRA 2003). Unfortunately, most of the new gauging stations were not established early enough to gather sufficient pre-treatment data to allow an unambiguous conclusion on whether any stream flow increases are due to brush control or to other factors, such as increased precipitation or changes in groundwater pumping. Most paired-watershed studies indicate that at least 5 10 years of data are needed prior to treatment. Groundwater levels are being monitored on a quarterly basis in about 61 privately owned wells in the North Concho River watershed (UCRA 2003). Well water levels for the past few years suggest a slight increase, and UCRA contends that these increases are due to brush removal and the commensurate elimination of water extraction from shallow aquifers by deep-rooted brush, primarily mesquite. A limitation is that the groundwater data are being collected in areas where brush control is under way, and the lack of wells in areas without brush control will make it difficult to determine whether an increase in water levels is due to brush control or to other factors. Paired study sites. Two sets of paired sites are being monitored near San Angelo in northern Tom Green County to quantify the hydrologic effects of brush control (UCRA 2003). One pair is in a flat, mesquite-dominated area with relatively deep soils. The other pair is in a hilly, redberry juniper dominated site with shallow soils. Both sets of watersheds consist of acres. Evapotranspiration is being measured from the mesquite watersheds, one of which was aerially sprayed with herbicide for mesquite control in June Runoff is being measured at the juniper sites, and one of the watersheds will be treated mechanically for juniper control in the future. No results are available at this time. The work of Dugas and Mayeux (1991) suggests that changes in evapotranspiration rates at the mesquite site may be small because of the compensatory growth and water use by herbaceous vegetation, the deep water table, and the deep soils. p. 14

17 NEEDED RESEARCH This review and analysis of scientific data makes it clear that there is considerable uncertainty about the relationship between shrub cover and water yields. These uncertainties will be resolved only through carefully crafted, comprehensive, and coordinated research that is conducted at the appropriate scales. The purpose of this section is to provide a prototype research program that would help to resolve many of the issues raised in this report. Ideally, research locations would be selected that represent the most important shrublands in Texas. At each location, studies would be conducted at the tree, hillslope, small catchment, and watershed scales. Tree-Level Studies Scientists have a relatively good understanding of interception and transpiration at the tree level, but relatively little is known about where trees access water. Fortunately, new technology using naturally occurring isotopes is now available for determining the source of water used for transpiration (Snyder and Williams 2000; Williams and Ehleringer 2000). Work of this type has been conducted in shrublands of Arizona, but little, if any, has been conducted in Texas. Isotope studies could be particularly important in determining whether mesquite trees routinely access groundwater, as was assumed in the North Concho River watershed brush-control study (UCRA 1998), or whether groundwater use by mesquite is limited to riparian corridors. Hillslope and Small Catchment Scales: Detailed Water Balance Studies Detailed water balance studies at the hillslope and small catchment scales could determine where the water goes and how it gets there. These studies are needed within all the shrub communities of concern. Water balance studies would make detailed measurements of runoff, soil moisture, evapotranspiration, and recharge. Technology is now available that allows continuous, high-resolution measurements of each of these components. For example, evapotranspiration may be determined at large scales using either the eddy covariance or Bowen ratio methods. One of the few peer-reviewed studies indicating higher recharge rates or decreased evapotranspiration resulting from shrub control is that of Dugas et al. (1998), in which large-scale evapotranspiration measurements were used. Additional studies of this nature are needed for a variety of conditions to confirm and expand on these findings. Small Catchment Scale: Monitoring of Springs It is remarkable how little is known about spring flow in the state of Texas. Few, if any, smaller springs have been monitored for any length of time. A long-term, comprehensive monitoring program of small-to-medium springs could foster a good understanding of the relationship between spring flow, rainfall, and vegetation cover. As mentioned in this report, increases in spring flow following shrub control have been commonly observed but rarely measured. Watershed Scale Studies Paired studies of headwater streams. The effect of shrub management on stream flow will most likely be evident at the stream s headwaters. To clearly establish the connection between shrub cover and stream flow, long-term and replicated experimental watershed research needs to be established. The most important and definitive work that could be done is the establishment of long-term watershed studies at several locations across the state. These studies would require measurements of stream flow and other components of the water budget, including evapotranspiration and recharge on replicated plots with and without woody plants, at multiple scales (10 10,000 acres). p. 15

18 Groundwater surface water connections. For many rangeland watersheds, there is only a superficial understanding of where, how much, and how quickly aquifer recharge can occur. The Edwards Plateau region is probably the most studied and best understood because of the regional importance of the Edwards Aquifer, but even in this aquifer there are still major uncertainties. In most rangelands, there is uncertainty about the relative importance of large-but-infrequent versus small-but-frequent rainfall events. A limited understanding of the spatial and temporal nature of recharge limits the ability to understand and model how changes in woody plants may influence stream flow and recharge. A better understanding of the fate of recharge is also needed. Saltcedar along Lakeshores As noted in this report, large areas along the shorelines of reservoirs have been infested by saltcedar. As yet, there have been no careful analyses as to the hydrologic effect of this change in vegetation cover. For example, because no measurements have been made, it is not known whether total evapotranspiration is greater as a result of increased saltcedar cover at reservoirs such as Spence (Coke County), O. C. Fisher (Tom Green County), and Ivie (Concho, Coleman, and Runnels Counties). Research is needed to determine whether saltcedar roots extend to the water table along lakeshores and whether this water withdrawal affects reservoir levels. Dense monocultures of saltcedar have invaded the shorelines of Red Bluff Reservoir (left) and Ivie Reservoir (right), as well as those of many other lakes, ponds, and areas with shallow water tables in western Texas. p. 16

19 CONCLUSIONS 1. To date, increases in stream flow or recharge as a result of shrub control have not been demonstrated at scales larger than that of a small catchment. Because of limited data at large scales, there is considerable uncertainty in scaling up the small-scale measurements to large river basins. Thus, scientists are uncertain about whether shrub control can increase water yields anywhere in Texas and if so, how much. These uncertainties can be resolved only if state-funded brush control programs are continued in conjunction with carefully designed, well-maintained, and long-term monitoring studies. 2. The highest probabilities of water yield increases associated with brush control are likely for riparian areas, where saltcedar would be replaced by herbaceous plants, and in mesic karst or deep, sandy rangelands where groundwater recharge is rapid and substantial. Additional field research is required to determine the extent to which brush control has the potential to increase water yields on other rangelands in the state. 3. Models are useful to examine this issue, but modeling results should always be carefully examined and interpreted in light of data from field studies. At present, there are substantial discrepancies between field measurements and some model simulations models generally predict larger increases in water yields than have been observed. On the basis of this review and after careful consideration of this topic, the authors believe it is appropriate to broaden the issue from solely focusing on brush control for increasing water yields to best management practices for watershed health and sustainability. The encroachment of woody plants is only one of many ecological changes that have affected the vegetation and hydrological conditions of Texas watersheds (Hamilton and Ueckert 2004). The hydrological condition of natural plant communities and the partitioning of water within the hydrological cycle are determined by complex interactions between soil and vegetation factors. Clearly, in some cases, particularly for headwater streams fed by springs or riparian areas dominated by invasive phreatophytes, the integration of brush control with other best management practices has the potential to enhance base flow. It is unclear under what conditions these increased base flows will translate to greater stream flow in our larger rivers. However, the authors strongly believe the ecological condition of many of the headwater streams in Texas may be considerably improved through best management practices for overall improvement of rangeland health and ecosystem functioning, of which brush management is only one aspect. The integration of brush control with other best management practices has the potential to enhance base flow of headwater streams fed by springs or riparian areas dominated by invasive phreatophytes, but it is uncertain whether these increased base flows will translate to greater stream flow in larger rivers. The authors believe that the issue should be broadened from focusing solely on brush control for increasing water yields to best management practices for watershed health and sustainability. p. 17