1 Supplementary material

3 Materials and methods

5 Bat population estimates and roost locations

6 Our study area includes all U.S. counties producing either Pima or Upland cotton

7 varieties and within 50 km (conservatively, an individual bats nightly foraging distance

8 (Williams et al. 1973) of a major Mexican free-tailed bat roost. We used roost data

9 (location and bat population censuses) from the U.S. Geological Surveys Bat Population

10 Database (Ellison et al. 2003) and our own literature search (Supplementary material,

11 Table A1 & A2; Fig. 1). We only considered large summer roosts in our model (>7,000

12 individuals) because many smaller roosts lack accurate geospatial information and

13 because the combined populations of the largest summer colonies are thought to account

14 for the majority (>99%) of the migratory Mexican free-tailed bat population, which is

15 predominantly composed of females (McCracken 2003). As the roost population

16 estimates spanned a range of dates, we used only estimates obtained after 1970 to account

17 for concerns that bat populations may have decreased in the 1950s and 1960s due to DDT

18 exposure (Lewis 1995; Betke et al. 2008). In our model, we assumed that 90% of the

19 adult bats in each roost were female and 10% were male, which is consistent with field

20 data from breeding roosts in the southern United States (Federico et al. 2008). We were

21 not able to model changes in the bat population size over time, as the frequency and

22 precision of bat census data do not support time-series analysis (Federico et al. 2008).

23

1
24 Avoided crop damage calculation

25 Number of pests consumed. We first estimated the value of the cotton that would have

26 been lost without the presence of bats providing pest-control services. Conventional and

27 molecular analyses indicate that the Mexican free-tailed bats diet is quite varied,

28 encompassing 35 families of insects (Lee and McCracken, 2005). However, bats

29 capitalize on bollworms during outbreaks, and track and exploit local abundances of

30 bollworms within a region (Lee and McCracken 2005; Cleveland et al. 2006; McCracken

31 et al. 2012). These insects can comprise 3060% of the bats diet during times of peak

32 infestation, and each pregnant or lactating female bat consumes 510 female adult

33 bollworms per night during periods of peak bollworm infestation (Lee and McCracken

34 2005; Cleveland et al. 2006; McCracken et al. 2012). Since some bollworms infest other

35 crops in the area or migrate out of the region, we estimated that only 1020% of the

36 female moths, or approximately 1.5 individuals, consumed each night would have

37 dispersed into cotton and laid eggs (Cleveland et al. 2006). Because of high mortality

38 rates during insect development (9598%), the nightly consumption of 1.5 adult female

39 moths would prevent 5 larvae from developing and damaging cotton crops (Cleveland et

40 al. 2006; Sansone and Smith 2001). Bollworm consumption by non-reproductive female

41 and male bats was determined to be 32% lower than reproductive females due to the high

42 metabolic costs of lactation (Cleveland et al. 2006; Federico et al. 2008).

43 Over its lifetime, a single bollworm larva can damage 2 to 3 bolls of cotton

44 (Cleveland et al. 2006). We estimated values separately for each third of the cotton

45 growing season because the value of a cotton boll declines over the seasonbolls

46 produced during the first third of the season generate about 50% of the harvest while

2
47 bolls from the last third generate only 7% (Sansone et al. 2002). Bats prevent damage to

48 fewer bolls in Bt cotton versus conventional cotton. Therefore, compared to conventional

49 cotton, we assumed that bats prevented a lower number of larvae from developing in Bt

50 cotton (52.6% of the number of larvae in conventional cotton; Federico et al. 2008).

51

52 Cotton locations

53 We obtained data on the number of cotton hectares planted per county from the United

54 States Department of Agricultures National Agricultural Statistics Service (U.S.

55 Department of Agriculture 2012) and the National Cotton Council (National Cotton

56 Council 2013; Supplementary material, Table A2). Data on number of hectares planted

57 with cotton are available at the county level, so we approximated the locations of cotton

58 fields using potential crop soil maps in each county. This approach assumes that cotton

59 hectares are uniformly distributed over soils with high cotton potential. We used the U.S.

60 General Soil Maps (STATSGO data) from the U.S. Department of Agricultures Natural

61 Resource Conservation Service (NRCS) for locations of soil types suitable for cotton

62 production (U.S. Department of Agricultures Natural Resource Conservation Service

63 2012). For each year from 1990 to 2008, we assumed that the proportion of the cotton

64 hectares planted per county within foraging distance of the bats was equal to the

65 proportion of suitable cotton-growing soils for each county within the bats foraging

66 range of 50 km from each roost. Since 50 km is a conservative estimate of the foraging

67 range, and their patterns of foraging are unknown, we assume equal rates of pest

68 consumption across this distance. We also assumed that bats disperse randomly from

69 their roost, such that the percentage of the roosts bat population foraging in each cotton-

3
70 growing area was equal to the percentage of the area each cotton-growing region

71 composed of a roosts total foraging range. Since bats likely disperse non-randomly from

72 their roosts and concentrate on high quality foraging grounds, our calculation is

73 conservative.

74

75 Cotton prices

76 We used data from the National Cotton Council (National Cotton Council 2013) on U.S.

77 cotton prices from 1990 to 2008. The prices were adjusted for inflation and are reported

78 in 2011 USD (Supplementary material, Table A3).

79

80 Numbers of insecticide applications avoided

81 In general, insecticides are applied to cotton fields when bollworm infestations reach a

82 threshold of 20,00025,000 larvae per hectare. The date at which the threshold is

83 reached, which consequently triggers the first insecticide application, varies according to

84 region. The following cotton pest experts provided regional estimates of dates of first

85 insecticide application: C Sansone (Texas, Oklahoma, and Kansas), D Munier

86 (California), J Pierce (New Mexico), and P Ellsworth (Arizona). For fields planted with

87 Bt cotton, the threshold is reached later because the bollworm population growth rate is

88 lower (<10%) than that of conventional cotton (Jackson et al. 2003). This results in a

89 lower number of avoided insecticide applications in Bt cotton. We estimated the number

90 of insecticide applications that were avoided in the presence of bats by calculating the

91 number of times the threshold would have been reached without bat predation from the

92 first date of cotton flowering (and earliest significant economic susceptibility to

4
93 bollworms) to the first date of insecticide application. Even though bollworms can

94 damage cotton plants before flowering, we mark cotton flowering as the earliest

95 significant economic susceptibility. This is due to the fact that cotton plants can

96 compensate for early season fruit loss (Terry 1992; Wu et al. 2007), and there are

97 recommendations that early season insecticide applications may not be cost effective

98 (Terry 1992; North Carolina State University Extension 2016). We used a uniform

99 insecticide application rate of 0.29 kg/ha (Gianessi and Reigner 2006). Data on the cotton

100 season (e.g., mean planting and harvest dates) for different regions were obtained from

101 the U.S. Department of Agricultures National Agricultural Statistics Service (U.S.

102 Department of Agriculture 2012).

103

104 Avoided insecticide costs calculation

105 Private costs savings for insecticides represent the reduced cost to famers of purchasing

106 and applying chemicals. Information on mean costs of cotton insecticide applications per

107 hectare for each state from 1990 to 2008 were obtained from the Mississippi State

108 University Department of Entomology and Plant Pathologys databases on cotton losses

109 due to insects States (Mississippi State University Department of Entomology and Plant

110 Pathology). Social cost savings arise from lowered public health impacts to the

111 farmworkers who apply the insecticides, and reduced environmental damage due to loss

112 of beneficial pollinators, groundwater contamination, and damage to other species

113 (Pimentel et al. 1991). We determined the insecticides in the United States that are used

114 predominantly on cotton bollworms (Gianessi and Reigner 2006), and used data from

115 Kovach et al. (1992) and Cornell Universitys Integrated Pest Management Program

5
116 (Cornell Universitys Integrated Pest Management Program 2012) to estimate the

117 environmental and toxicological impacts of these particular cotton insecticides. We then

118 used a pesticide environmental accounting tool (Leach and Mumford 2008) to assign a

119 social-cost value in dollars for each insecticide according to estimates of its degree of

120 impact. The pesticide accounting tool attributes detrimental impacts in six different

121 categories: human health, groundwater contamination, aquatic systems (fish), birds, bees,

122 and other beneficial insects. Destruction of the latter category (beneficial, predatory

123 insects) can lead to secondary pest outbreaks in an agricultural setting (Pimentel et al.

124 1992; Gross and Rosenheim 2011). We used a weighted mean cost of insecticide

125 applications per hectare over time. Finally, we estimated the total value of these avoided

126 insecticide applications by calculating the number of avoided insecticide applications and

127 the costs of insecticide use, and then summing both the social and private costs

128 (Supplementary material, Table A4; Kovach et al. 1992; Cleveland et al. 2006.

129

130 Bt cotton

131 Bt cotton was introduced for the Upland variety of cotton in 1996. Bt cotton is not

132 available for the Pima variety, which composes less than 5% of the total cotton

133 production in the United States (Mississippi State University Department of Entomology

134 and Plant Pathology). Data on the timing of adoption of transgenic Bt Upland cotton was

135 obtained from the Mississippi State University Department of Entomology and Plant

136 Pathologys database on cotton crop losses (Supplementary material, Table A4;

137 Mississippi State University Department of Entomology and Plant Pathology).

138

6
139 Sensitivity analysis

140 To better understand factors influencing bat ecosystem service values, we analyzed the

141 sensitivity of the annualized mean pest-control value from 1990 to2008. We altered the

142 following parameters by 10%: total area planted with cotton, Mexican free-tailed bat

143 population size, and price of cotton. We measured the effect of these perturbations on the

144 annualized mean pest-control value. The mean pest-control values were equally sensitive

145 to ecological and economic factors: both a 10% change in the bat population size and a

146 10% change in cotton prices caused a 9.1% change in the mean pest-control values.

147 However, altering the total area planted with cotton by 10% only caused a 0.9%

148 change in the mean pest-control value over time.

149

150

151

152

153

154

155

156

157

158

159

160

161

7
 162 Table A1. Summer breeding roosts and winter regions for migratory Mexican free-tailed
163 bats.
Site/Region Country State Roost Abundance Source
type Estimate
Eagle Creek Cave USA AZ Summer 300,000 (Mohr 1972; Reidinger 1972)
Cosumnes River USA CA Summer 60,000 (Cayton et al. 2015),
Preserve unpublished data
Yolo Bypass Bridge USA CA Summer 250,000 (Cayton et al. 2015),
unpublished data
Orient Mine USA CO Summer 100,000 (Freeman and Wunder 1988)
Carlsbad Caverns USA NM Summer 341,026 (Betke et al. 2008)
Merrihew Cave USA OK summer 100,000 (Arganbright 1989)
Read Cave USA OK summer 500,000 (Elliot 1994)
Vickery Cave USA OK summer 1,000,000 (Humphrey 1971)
Bracken Cave USA TX summer 4,000,000 co-author GFM, unpublished
data
Congress Bridge USA TX summer 1,500,000 (Wahl 1993; Keeley and Tuttle
1999)
Davis Cave USA TX summer 431,205 (Betke et al. 2008)
Devil's Sink Hole USA TX summer 2,000,000 co-author GFM, unpublished
data
Eckert James River Cave USA TX summer 1,312,027 (Betke et al. 2008)
Fern Cave USA TX summer 250,000 (BCI 2003)
Frio Cave USA TX summer 2,000,000 co-author GFM, unpublished
data
McNeil Bridge USA TX summer 600,000 (Allen et al. 2010)
Ney Cave USA TX summer 397,846 (Betke et al. 2008)

Stuart Bat Cave USA TX summer 500,000 (Texas Parks Wildlife 2007)
Waugh Bridge USA TX summer 250,000 (Texas Parks Wildlife 2007)
Cuatrocinegas de Mexico Coahuila summer 1,000,000 (MaNIS 2011)
Carranza

Cueva de Consuelo Mexico Coahuila summer 800,000 (BCI 2003)

Cueva La Boca Mexico Nuevo summer 2,000,000 (Lopez-Damian 2009)

Leon
Maviri Mexico Sinaloa summer 100,000 co-author RAM & Ejido Juan
Aldama, unpublished data
Cueva del Tigre Mexico Sonora summer 2,000,000 co-author RAM, unpublished
data

Presa de Amistad Mexico Tamaulipas summer 1,000,000 co-author RAM, unpublished

data
Chiapas Mexico Chiapas winter 1,000,000 (Lopez-Damian 2009)

8
Hildago Mexico Hildago winter 20000 (MaNIS 2011)
Michoacan & Jalisco Mexico Michoacan winter 20000 (Clark et al. 1995)
& Jalisco
Quertaro Mexico Quertaro winter 10000 co-author RAM, unpublished
data
164
165

9
166 Table A2. Cotton extent and Mexican free-tailed bat population size per county.
Mean
Bat pop. Mean cotton Bat pop. cotton
County State size hectares* County State size hectares*
Pima AZ 1,012 4,731 Denton TX 24,117 641
Colusa CA 115 1,632 Donley TX 19,754 5,677
Sutter CA 19,471 223 Floyd TX 69,445 63,771
Yolo CA 50,310 421 Fort Bend TX 34,092 17,666
Barber KS 2 97 Guadalupe TX 758,067 228
Eddy NM 9,263 2,897 Hall TX 120,955 28,777
Beckham OK 91,594 4,063 Hardeman TX 2,314 4,236
Greer OK 133,153 3,233 Harris TX 167,276 243
Harmon OK 114,044 8,877 Hays TX 356,558 2,119
Harper OK 1,553 37 Hidalgo TX 362,081 27,265
Jackson OK 47,544 22,425 Kinney TX 59,430 2,353
Kiowa OK 1,153 7,652 Mason TX 353,907 30
Pawnee OK 18,103 175 McCulloch TX 186,304 1,937
Payne OK 373 326 Medina TX 415,902 2,391
Woods OK 36,261 2,700 Montague TX 225,287 831
Armstrong TX 13,955 285 Motley TX 64,320 10,526
Bastrop TX 203,968 399 Starr TX 3,727 2,879
Bexar TX 1,705,727 228 Swisher TX 13,003 26,706
Brazoria TX 22,780 2,195 Travis TX 662,059 2,147
Briscoe TX 184,253 13,636 Uvalde TX 1,063,973 3,204
Caldwell TX 95,654 1,832 Wheeler TX 1,490 2,866
Cameron TX 2,615 26,441 Williamson TX 223,004 13,165
Childress TX 15,260 13,631 Wilson TX 245,298 426
Clay TX 43,359 603 Wise TX 148,727 318
167 Note: *From 1990 to 2008.
168
169

10
170 Table A3. Upland and Pima cotton price over time.
Upland Cotton Pima Cotton
Year $/kg $/kg
1990 0.34 0.51
1991 0.26 0.44
1992 0.24 0.37
1993 0.30 0.40
1994 0.40 0.50
1995 0.38 0.68
1996 0.32 0.49
1997 0.31 0.45
1998 0.27 0.41
1999 0.24 0.37
2000 0.24 0.39
2001 0.15 0.36
2002 0.21 0.36
2003 0.27 0.44
2004 0.21 0.51
2005 0.22 0.51
2006 0.22 0.42
2007 0.28 0.38
2008 0.22 0.43
171
172
173
174
175
176
177
178
179
180
181

11
182 Table A4. Compound spatial-temporal approach variables, their mean values, and ranges.
Variables Mean* Minimum Maximum
Bat population size 214,045 6,199 516,713
Hectares of cotton 160,641 3 1,835,222
Hectares of Pima cotton 18 0.00 208
Hectares of upland cotton 6,078 0.06 63,771
% Bt cotton 24% 0% 100%
Private insecticide costs per hectare
($) 15.91 3.71 43.24
Hectares of suitable soil covered by
bats 109,059 18 465,022
Social insecticide cost $ per kg/ha# 7.47 5.59 9.57
Upland cotton price $ per kg# 1.32 0.82 2.06
Pima cotton price $ per kg# 2.15 1.73 2.92
183 Note: *Mean annualized values across all counties except where noted, #Mean across
184 time (19902008).
185
186
187
188
189
190

191

192

193

194

195

12
196 References

197 Allen L, Richardson C, McCracken G, Kunz T. 2010. Birth size and postnatal growth in

198 cave- and bridge- roosting Brazilian free-tailed bats. J Zool. 280:8-16.

199 Arganbright K. 1989. Population dynamics in a colony of Mexican free-tailed bats

200 (Tadarida brasiliensis) in Bandelier National Monument. 1989 USGS Final Report.

201 Fort Collins, CO, USA.

202 BCI, Bat Conservation International. 2003. Bat Conservation International Annual

203 Report. Austin, TX.

204 Betke M, Hirsh DE, Makris NC, McCracken GF, Procopio M, Hristov NI, Tang S,

205 Bagchi A, Reichard JD, Horn JW, et al. 2008. Thermal Imaging Reveals Significantly

206 Smaller Brazilian Free-Tailed Bat Colonies Than Previously Estimated. J

207 Mammal.89:18-24.

208 Cayton HL, Haddad NM, Gross K, Diamond SE, Ries L. 2015. Do growing degree days

209 predict phenology across butterfly species? Ecology. 96:1473-1479.

210 Clark DR, Moreno-Valdez A, Mora MA. 1995. Organochlorine residues in bat guano

211 from nine Mexican caves, 1991. Ecotoxicology. 4:258-265.

212 Cleveland CJ, Betke M, Federico P, Frank JD, Hallam TG, Horn J, Lopez JD,

213 McCracken GF, Medellin RA, Moreno-Valdez A, et al. 2006. Economic value of

214 the pest control service provided by Brazilian free-tailed bats in south-central

215 Texas. Front Ecol Environ. 4:238-243.

216 Cornell Universitys Integrated Pest Management Program. 2012. [Internet]. [cited 2012

217 Mar 12]. Available from

218 http://www.nysipm.cornell.edu/publications/eiq/equation.asp

13
219 Elliot W. 1994. Conservation of Western Oklahoma Bat Caves. Ok. Undergr. 17:44-53.

220 Ellison L, O'Shea T, Bogan M, Everette A, Schneider D. 2003. Existing data on colonies

221 of bats in the United States: summary and analysis of the US Geological Survey's bat

222 population database. USGS/BRD/ITR--2003-0003. Fort Collins, CO, USA.

223 Federico P, Hallam TG, McCracken GF, Purucker ST, Grant WE, Correa-Sandoval AN,

224 Westbrook JK, Medellin RA, Cleveland CJ, Sansone CG, et al. 2008. Brazilian free-

225 tailed bats as insect pest regulators in transgenic and conventional cotton crops. Ecol

226 Appl. 18:826-837.

227 Freeman J, Wunder L. 1988. Observations at a colony of the Brazilian free-tailed bat

228 (Tadarida brasiliensis) in southern Colorado. Southwest Nat. 33:102-104.

229 Gianessi L, Reigner N. 2006. Pesticide use in U.S. crop production. Washington D.C.

230 Gross K, Rosenheim JA. 2011. Quantifying secondary pest outbreaks in cotton and their

231 monetary cost with causal-inference statistics. Ecol Appl. 21:2770-80.

232 Humphrey SR. 1971. Photographic estimation of populataion size of Mexican free-tailed

233 bat, Tadarida brasiliensis. Am Midl Nat. 86:220-223.

234 Jackson RE, Bradley Jr JR, Van Duyn JW. 2003. Field performance of transgenic cottons

235 expressing one or two Bacillus thuringiensis endotoxins against bollworm,

236 Helicoverpa zea (Boddie). J Cott Sci. 7:57-64.

237 Keeley BW, Tuttle MD. 1999. Bats in American Bridges. Resource Publication No. 4.

238 Austin, TX, USA.

239 Kovach J, Petzoldt C, Degni J, Tette J. 1992. A method to measure the environmental

240 impact of pesticides. NY Food & Life Sci Bull. 129:1-8.

14
241 Leach A, Mumford J. 2008. Pesticide Environmental Accounting: A method for assessing

242 the external costs of individual pesticide applications. Environ. pollut. 151:139-147.

243 Lee Y-F, McCracken GF. 2005. Dietary variation of Brazilian free-tailed bats linked to

244 migratory populations of pest insects. J Mammal. 86:67-76.

245 Lewis SE. 1995. Roost fidelity of bats - a review. J Mammal. 76:481-496.

246 Lopez-Damian L. 2009. Dieta de Tadarida brasiliensis mexicana en el noreste y sur de

247 Mxico en el contexto de la fenologa del maz (Zea mays) [The diet of Tadarida

248 brasiliensis mexicana in the Northeast and south of Mexico in the context of mais Zea

249 mays phenology]. [Thesis]. Mexico City, Mexico: Universidad Nacional Autnoma

250 de Mxico.

251 Mammal Networked Information System (MaNIS). 2012. [Internet]. [cited 2012 Mar 12].

252 Available from: http://manisnet.org.

253 McCracken G. 2003. Estimates of population sizes in summer colonies in Brazilian free-

254 tailed bats (Tadarida brasiliensis). USGS/BRD/ITR--2003-0003. Fort Collins, CO,

255 USA.

256 McCracken GF, Westbrook JK, Brown VA, Eldridge M, Federico P, Kunz TH. 2012.

257 Bats Track and Exploit Changes in Insect Pest Populations. PLoS One. 7:e43839.

258 Mississippi State University Department of Entomology and Plant Pathology. 2012.

259 [Internet]. [cited 2012 Mar 12]. Available from

260 http://www.biochemistry.msstate.edu/resources/cottoncrop.asp

261 Mohr CE. 1972. The status of threatened species of cave-dwelling bats. Bull Nat Spel

262 Soc. 34:33-47.

15
263 National Cotton Council. 2012. http://www.cotton.org [Internet]. [cited 2012 Mar 12].

264 Available from http://www.cotton.org

265 North Carolina State University Extension. 2016. Cotton Information. Raleigh, NC,

266 USA.

267 Pimentel D, Acquay H, Biltonen P, Rice S, Silva M, Nelson J, Lipner V, Giordano S,

268 Horowitz A, D'Amore M. 1991. Environmental and economic effects of reducing

269 pesticide use. BioScience. 42:750-760.

270 Reidinger RF. 1972. Factors influencing Arizona bat population levels [Dissertation].

271 Tucson, AZ, USA: University of Arizona.

272 Sansone C, Smith J. 2001. Natural mortality of Helicoverpa zea (Lepidoptera:

273 Noctuidae) in short-season cotton. Biol Cont. 30:113-122.

274 Sansone CG, Isakeit T, Lemon R, Warrick B. 2002. Texas cotton production:

275 Emphasizing integrated pest management. College Station, TX.

276 Terry IL. 1992. Effect of early season insecticide use and square removal on fruiting

277 patterns and fiber quality of cotton. J Econ Entomol. 85:1402-12.

278 Texas Parks Wildlife. 2007. Bat Watching sites in Texas. PWD BK W7000-1411.

279 Austin, TX, USA.

280 U.S. Department of Agricultures National Agricultural Statistics Service: U.S.

281 Department of Agriculture. 2012. [Internet]. [cited 2012 Mar 12]. Available

282 from http://www.usda.nass.gov

283 U.S. Department of Agricultures Natural Resource Conservation Service. 2012.

284 [Internet]. [cited 2012 Mar 12]. Available from

16
285 http://www.nrcs.usda.gov/wps/portal/nrcs/detail/soils/survey/geo/?cid=nrcs142p2

286 _053629

287 Wahl R. 1993. Important Mexican free-tailed bat colonies in Texas. 1989 National Cave

288 Management Symposium Proceedings.

289 Williams TC, Ireland LC, Williams JM. 1973. High altitude flights of the free-tailed bat,

290 Tadarida brasiliensis, observed with radar. J Mammal. 54:807-821.

291 Wu G, Chen FJ, Ge F. 2007. Impacts of early-season square abscission on the growth

292 and yield of transgenic Bt cotton under elevated CO 2. Field Crops Res. 102:239-

293 43.

294

295

296

297

298

299
300

301

17