PAGE 1
Factors that affect broca incidence in coffee plants Simon 1 What factors affect Broca beetle ( Hypothenemus hampei ) incidence in Coffea arabica plants at Life Monteverde? Ari Simon Department of Ecology, Behavior & Evolution University of California, Los Angeles EAP Tropical Biology and Conservation, Fall 2016 16 December 2016 ABSTRACT The goal of this study was to identify the factors that affect incidence of the coffee berry borer (CBB) or Broca beetle, Hypothenemus hampei I surveyed three sites within the Life Monteverde coffee plantation in Costa Rica. The three sites varied from High (C2 ) to Medium (C14) to Low (C18) historical incidence of Broca infestation. Six 6x3m plots per study site were cre ated and I selected four plants per plot. Per plant, I chose five branches with the one condition that the branch had a high density of berries (n>50). I conducted observations between 15 and 2 3 November 2016 and observed 39393 berries of which 3229 were infested with Broca (7.92%). Site C2 had an infestation of 18.88%, 3.83% were infested in Site C14, and 1.02% in Site C18. Using stepwise regression modeling, the resulting best fit model indicated that the most significant variables that affect the percentage of Broca infestation per study site were the distance from the back windbreak, the distance from each side windbreak, the distance from the road, the length of the branches, and the height of the tree. Understanding the factors that influence the habitat preferences of Broca beetles can help to increase natural and biological control methods, reducing the need for chemical and artificial control methods. Qu factores afectan la incidencia de Broca ( Hypothenemus hampei ) en plantas de Coffea arabica en Life Monteverde? RESUMEN El objetivo del presente estudio fue identificar los factores que afectan la incidencia de broca del caf, Hypothenemus hampei Realic un muestreo de tres sitios dentro de la plantacin de caf Life Monteverde en Costa Rica. Los tres sitios variaron de in cidencia histrica alta (C2) a media (C14) a baja (C18) de infestacin de Broca. Demarqu seis parcelas de 6x3m por sitio de estudio y seleccion cuatro plantas por parcela. Por planta, eleg cinco ramas con la nica condicin de que la rama tuviera una al ta densidad de frutos (n> 50). Hice mis observaciones entre el 15 y el 23 de noviembre de 2016 y cont un total de 39393 frutos, 3229 infestados con broca (7,92%). El sitio C2 tuvo una infestacin del 18,88%, C14 present 3,83% y C18 un 1,02% de incidencia Utilizando un modelo de regresin, el modelo de ajuste ptimo resultante indic que las variables ms significativas que afectan el porcentaje de infestacin de broca por sitio de estudio fueron la distancia del rompevientos trasero, la distancia del rom pevientos de lado, distancia de la carretera, longitud de las ramas, y la altura del rbol. Entender los factores que influyen en las preferencias de hbitat de la broca puede ayudar a aumentar los
PAGE 2
Factors that affect broca incidence in coffee plants Simon 2 mtodos de control natural y biolgico, reduciendo la nece sidad de mtodos qumicos y artificiales T he coffee berry borer (CBB) or Broca beetle, Hypothenemus hampei is one of the largest threats to global coffee production. This native African species has spread across coffee plantations worldwide decreasing crop yield by up to 80% and causing losses of close to $500 million every year to coffee growers (Cejas Navarro 2015). Broca beetles first a rrived in Costa Rica in 2000 and were present at 97% of farms by 2011 with an average incidence rate of 2.76% per farm (Rojas 2012). Female beetles, between 1 2 millimeters long, drill small holes into the immature or mature berries of the coffee plants wh ere they create special galleries to lay their eggs. The female lays between 31 119 eggs and after 4 8 days the eggs hatch. The resulting larvae feed on the berry as they matur e and after 12 15 days after reproducing via sibling mating or parthenogenesis (depending on the presence of males) the females emerge to infest new berries (Crdenas et al. 2007). Most of the life cycle is spent inside coffee berries (males never leave the berries) which makes this pest difficult to control. The CBB cau se the prem ature fall of young berries, increased vulnerability of infested ripe berries to fungus or bacterial infection and the reduction in b oth yield and quality of coffee College of Tropical Agriculture and Human Resources 2016). C ontro l of beetle populations has been difficult due to the protection given to the beetles inside the berries. Preventive measures such as the use of pheromone traps, the introduction of the fungus Beauveria bassiana and parasitoids, as well as chemical control measures such as insecticides, have aimed to eliminate the beetles prior to berry infe station These measures follow a strict schedule in order to maximize their effectiveness throughout the year and coincide with b reeding and emergence times of the beetles. The beetles attack when the dry weight of the berries is equal to or greater than 20%, which is achieved when the fruit reaches between 100 and 150 days of development after flowering (Camilo et al. 2003). The be etles target the penetrable berries and tend to gravitate towards ones that are in shadier and moister environments. Beetles are more abundant in h igher density pl antations and older larger tree. H ighest CBB infestations levels were observed on lower branc hes in contrast to the middle and top of the tree ( Aristizbal 2016). Main factors that lead to increased Broca incidence include (1) the w ay in which the berries are collected during harvest season, (2) the amount of berries that fall and remain on the gr ound, (3) the amount of berries that remain on the branch es from the previous year (4) the type of windbreaks surrounding the field and (5) the types of traps used ( J,Santamara, pers.comm ). Berry collection methods are especially important because r ipe and over ripe berries that are left on the trees after harvest and those that fall on the ground serve as a source of new CBB infestations ( Aristizbal 2016). The goal of this study is to further understand and explore what factors affect Broca infesta tion I aimed to explore the habitat preference of Broca on Co ffea arabica plants within three different sites of varying Broca incidence, on the Life Monteverde coffee plantation in Caitas, Guanacaste Costa Rica I hope that my data can be used to supplement biol ogical control measures not only utilized by this farm but also on othe r farms throughout Costa Rica.
PAGE 3
Factors that affect broca incidence in coffee plants Simon 3 MATERIALS & METHODS Study Sites I conducted my research on the Life Monteverde coffee plantation in Caitas, Guanacaste Province, C osta Rica between 15 November and 23 November 2016 This area has an annual mean temperature between 16 18 C (61 64 F) and an annual rainfall of ~3000mm. I selected three s ites within the farm selected based on their reported historical Broca incidence. The sites varied in size, vehicular traffic, elevation, windbreak type, and age of the coffee plants within them. However, because all sites were within the same farm, the preventative measures used for Broca were the same. On 2 April 2016 500cc of Muralla Delta 190 OD insecticide dissolved in 800L water was applied and in June 2016 1L of Beauvaria bassiana fungus dissolved in 800L water was applied throughout the entire farm The use of insecticides was minimized as much as possibl e on Life Monteverde in favor of biological controls. Life Monteverde has an overall Broca incidence of less t han 3%. T hough present in nearly every plot on the farm, t hrough integrated management techniques Broca numbers have been localized and minimized Site 1 (C2) known loc ally as La Paila, was a plot with an area of 4,531.23 m (48,773.76 ft ) at an elevation range of 1286 to 1296 m. I selected this plot because I had been told that it was the highest incidence area on the farm ( J. Santam ara pers. comm. ) distinguishing characteristic was a road with a high level of vehicular traffic on the road bisecting it collecting truck commonly use d this road as a throughway to the rest of the farm as well as a collection point for bags of picked be rries. The average age of Co ffea arabica plants in this plot was 16 years old. C2 contained a side windbreak and a back windbreak (Figure 1 ) comprised of natural secondary forest. Site 2 (C14 ) known locally as Los Pinos, was a plot with an area of 3,251 .79 m (35,001.98 ft) a t an elevation range of 1245 1255 m. This plot represented a medium incidence due to its incidence of slightly above the farm average (J. Santamara, pers. comm.) Los Pinos had less vehicular traffic than La Paila but more than Gua yabo The average age of Co ffea arabica plants in this plot was 12 14 years old. The windbreak s used were a mixture of natural secondary growth forest as well as Casuarina equisetifolia (Australian pine) and assorted C ypress trees that had been planted roughly 30 ye ars ago. This study site contained only a side windbreak. ( Figure 1 ) Site 3 (C18), known locally as Guayabo, was a plot with an area of 646.40 m (6,957.83 ft) at an elevation of 1241 m. This site was reported to have little to no Broca at any point of the year (J. Santamara, pers. comm.). It was representative of the incidence in the majority of the areas on the farm. There was almost no vehicular traffic to this site The average age of Co ffea arabica plants in this plot was approximately 10 years old. The windbreak used was Casuarina
PAGE 4
Factors that affect broca incidence in coffee plants Simon 4 equisetifolia (Australian pine) and assorted C ypress trees that had been planted roughly 30 years ago. smaller size, the windbreaks surrounded the plants on 3 sides with the distance from the side windbreaks 18m from the center of the field. (Figure 1 ) Within each study site, planted windbreaks with native species Colpach ( Croton niveus ) and Tub ( Montanoa guatemalensis) were also used to supplement the other windbreak types. The age of these windbreaks was unknown. Figure 1: The map of the Life Monteverde coffee plantation. Study sites C2, C14 and C18 are marked. Monteverde Conservation League, 2010 Experimental Design Within each of my study sites, I created six 6 x3m plots Each plot contained 4 Coffea arabica plants The two varieties of Coffea arabica that were found within these sites were caturra and catua However, these varieties were not subject to separate anal ysis. Plot 1 at each site was the primary plot from which the other plots were created. Using a standard 30 meter tape measure, Plots 2 and 3 were set up 9 meters West (C2 ) or South ( C14/ C18) [Side #1] or 9 meters East (C2) or North ( C14/ C18) [Side #2] from Plot 1 within the same row. Plot s 4, 5, and 6 were set up 9, 18, and 27 meters (respectively ) uphill from Plot 1 ( see Appendix A ) The height of these plants was recorded and subsequently marked with orange flagging tape. This was done in order to pre vent coffee pickers from collecting berries within the selected plots. I selected f ive branches from each plant to sample I selected branches that contained a high density of berries (preferably 50 or more) Each individual branch was marked with orange f lagging tape and the distances of the branch from the ground as well as the length of each branch were recorded As there was a large amount of variation between the amount of berries
PAGE 5
Factors that affect broca incidence in coffee plants Simon 5 on each plant as well as the number of branches, by selecting the branch es with t he highest number of berries, I maximized the amount of berries able to be studied Once the plants and branches were flagged, I inspe cted each branch to locate Broca infeste ), which I then then removed and recorded. Next, I rescanned the branch and counted the total number of berries found on the branch (and branchlets) and collected a ny Broca infested berries missed during my first visual scan. After I scanned each branch I opened the Broca infested berries in order to verify if the berries were, in fact, infested A berry was considered infested if (1) a beetle was found and/or (2) an ob vious hole was present (Figure 2 ) and/or (3) damage to the berry was evident (Figure 3). Figure 2: The boring hole created by Hy pothenemus hampei (Broca) beetles in a mature Coffea arabica berry. L. Shyamal, 2013
PAGE 6
Factors that affect broca incidence in coffee plants Simon 6 Hypothenemus hampei (Broca) beetles in a mature Coffea arabica berry. Bob Nelson 2010 If a berry was not found to have Broca damage, it was simply added to the total count. E ach berry was checked individually to ensure the accuracy of my collection method. The accuracy at which I successfully identified Broca berries was 99.90%. (see Append ix A ) At each site, I recorded the t ype of windbreak used at the plot t he average a ge of the plants in the plot, and the elevation of the plot. Statistical Methods All data w ere entered into Microsoft Excel. I used the StatPlus:mac Professional software to run an ANOVA to analyze the variance between my three study sites and the percentage of Broca. I also used this program to run a Chi square test to the g oodness of fit of my observed Broca percentage between branches and trees in the three stud y sites. I used R Commander to run linear models to analyze the significance of my variables. I also ran forward and backward stepwise models as well as a forward/backward stepwise regression to find the best fit model for my variables. Google Maps 2016 was used in order to estimate the areas of the study sites. Linear Modeling Multiple linear regressions are used to assess the linear relationship between two or more continuous or categorical explanatory variables and a single continuous response variable (Lang 2007). The data set I used was one that included all the variables that I analyzed during my study period The following variables were used for a stepwise multiple regression: Dependent variable : Percentage of Broca infestation per branch (continuous) Independent variables : 1. Age of the plants (categorical) 2. Distance from the road (continuous) 3. D istance from the Back windbreak (continuous) 4. Distance from the Side #1 windbreak (continuous) 5. Distance from the Side #2 windbreak (continuous) 6. The elevation of the plots (categorical) 7. The height of the trees (continuous) 8. The height of the br anches (continuous) 9. The length of the branches (continuous) 10. The area of the sites (categorical) 11 The type of windbreak used (categorical) I ran a formal check as well as informal checks within R Commander to confirm and verify the assumptions of my model. The informal checks were an inspection of my linear model graphs for residuals as well as a descriptive statistics analysis to calculate mean and standard deviation of my
PAGE 7
Factors that affect broca incidence in coffee plants Simon 7 test variables. My formal ch eck was a correlation matrix. This matrix was used to find the significant variables that would be used in order to run the stepwise regression models (see Appendix B). The factors that correlated significantly with the dependent variable were enter ed in the regression equation. The independent var iables ideally would not be very highly correlated with one another (hi gh multicolinearity). Thus, I removed the independent variables that correlated with one another at 0.5 or higher. From this correlation matrix, the variables that I was able to elimina te from the regression equation were the age of the plants, the area of the site, and the elevation of the plots. I only entered the variables that significantly contributed to the variance of the dependent variable in the equation. Using R Commander I ex ecuted both a forward and backward stepwise analysis as well as a forward/backward stepwise regression analysis with the significant variables to find the best fit model for my data. The model that was used in my regression analysis was: lm(formula = BrocadoPercentage ~ BackWindbreak + HeightBranch + LengthBranch + RoadDistance + Side1Windbreak + Side2Windbreak + TreeHeight + WindbreakType ) RESULTS In Site C2 I found 18.88 0. 138 (S D ) Broca. Of the 14612 berries scanned, 2556 were infested with Broca I n Site C14 I found 3.83 0. 069 Broca O f the 12432 berries scanned 564 were infested with Broca. In Site 3 I found 1.02 0.0 26 Broca. Of the 12349 berries scanned, 1 09 were infested with Broca ( Figure 4 ) The total number of berries counted was n=39393 with an n= 3229 infested with Broca
PAGE 8
Factors that affect broca incidence in coffee plants Simon 8 Figure 4 : P ercentage of Broca infe sted berries per study site ( S.D.) C2 does in fact have a much larger percentage of berries with Broca as compared to sites C14 and C18. There was statistical significan ce in the variation of Broca percentage between sites C2 is significantly different than sites C14 and C18 (F 2,357 =133.51, p<.0001). However there is no significant difference between sites C14 and C18 I measured the number of branches that were not infested with Broca. Per site, the number of branches analyzed was n = 120. Site C2, with the highest percentage of Broca infestation had 0 branches that were not infested Site C14, with a medium percentage of Broca infestation had 53 branches that were not infested Site C18, with a low percentage of Broca infestation had 69 branches that were not infested (Figure 5 ). The number of branches not infested by Broca was statistically significant between study sites. C2 C14 C18 Series1 18.88 3.84 1.03 0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00 18.00 20.00 Percent Brocado Site
PAGE 9
Factors that affect broca incidence in coffee plants Simon 9 Figure 5 : The n umber of branches not infested by Broca per plot 2 = 97.03, df = 2, p < 0.00001 ) Per site, the number of trees analyzed was n = 24. Site C2 had 0 trees without Broca infestation site C14 had 3 trees and C18 had 1 tree. T he number of trees not infested by Broca was statistically significant between study sites (Figure 6 ). Figure 6 : Number of trees not infested by Broca per plot. 2 = 3.72 df = 2, p = 0 .01 55 ) The resulting best fit model from my stepwise regression (Table 1) was: lm( formula = BrocadoPercentage ~ BackWindbreak + Side1Windbreak + Side2Windbreak + RoadDistance + LengthBranch + TreeHeight)
PAGE 10
Factors that affect broca incidence in coffee plants Simon 10 Table 1: The results from my stepwise regression. The F statistic and p value indicate a significant result. The adjusted R 2 value indicates that 54.06% of the unique variance in independence is accounted for by the above factors. This indicates that 45.94% of the var iance is due to chance, error, or other factors which I did not measure. Model Standard Error Multiple R 2 Adjust ed R 2 F p Best Fit Model 0.081 (dF = 345) 0.56 0.54 31.18 (dF = 14,435) 2.2e 16 This best fit model i indicates that the factors that most significantly affect the percentage of Broca infestation were: Distance from the Back windbreak Distance from the Side #1 windbreak Distance from the Side #2 windbreak Distance from the Road The Length of the Branches The Height of the Trees This model indicated that these factors be further analyzed. There was a significant negative correlation between the distance from the road and the percentage of Broca infestation in sites C2 and C14 (Figure 7 C2: p=0.02, C14: p=0.00004 ). Both of these correlations were statistically significant This data indicates that the higher Broca incidence was closer to the road and that the percentage of Broca declined in the plots further from the road. C18 did not show a significant correlation. Figure 7: The distance from the road and its effect on the percentage of Broca infested berries The distance from the back windbreak had a positive correlation on the percentage of Broca infested berries at Site C2 (p = 0.02) This indicates that the further from the back windbreak, or the closer towards the front of the plot, the higher the percentage of Broca infested berries (Figure 8). Site C18 did not have a significant correlation.
PAGE 11
Factors that affect broca incidence in coffee plants Simon 11 Figure 8: The distance from the back windbreak and its effect on the percentage of Broca infested berries. I did not measure a back windbreak at Site C14. T he distance from the Side #1 Windbreak showed a significant positive c orrelation in sites C2 and C14 (C2: p = 0.00005, C14: p = 0.00004) This indicates that the further towards the middle of the plot, the higher the percentage of Broca infested berries (Figure 9). Site C 18 did not have a significant correlation between the percentage of Broca infested berries and the distance from either Side #1 or Side #2 windbreaks. Figure 9: The distance from the Side #1 windbreak and its effect on the percentage of Broca infested b erries. C2 and C14 did not have Side #2 windbreaks. DISCUSSION As expected, I found the highest Broca infestation in site C2 La Paila, medium Broca infestation in site C14 Los Pinos, and low Broca infestation in site C18 Guayabo. Furthermore, in sites C14 and C18 a larger number of branches had no Broca
PAGE 12
Factors that affect broca incidence in coffee plants Simon 12 whatsoever. This corresponded with the overall percentage of Broca infested trees and plots within each site. Significant Factors Distance from the Road I found that t he distance from the road was statistically significant in my regression modeling. The distance from the road is an important factor due to the fact that it serves as an access point for Broca beetles to enter the farm and, once on the farm, to move further into the farm. Most Bro ca will fly short distances and have been found to fly up to 500 m, befo re infesting another plant but long distance dispersal may be achieved via wind, animals, and humans, with international dispersal occurring inadvertently via the coffee trade (Arisiti zbal et al 2016). Broca can easily attach to a vehicle or to a coffee picker and use them as a method to infest an area with desirable conditions for boring. The distance from the road showed a negative correlation indicating that the further from the road, the lower the percentage of Broca infested berries. This lends support to the idea that roads serve an integral role in the spread of Broca. To further understand this factor, it would be beneficial to quantify the amount of traffic on a road as well as analyze the materi als and workers that utilize roads. Distance from the Windbreak s I found that th e distance of t he windbreaks from the plots were statistically significant. The distance from the back windbreak, as well as the distance from the Side #1 and Side #2 windbreaks. Windbreaks are used in order to prevent the effects of wind and soil erosion on the coffee plants. It also p rovides shade, wh ich serves to protect young coffee plants from drought stress and over exposure to sun and tree overbearin g and/or dieback in older trees ( Food and Agricultural Organization of the United Nations ). Shade can maintain a better environment (higher humidity) that is more desirable by the beetles. Increased transpiration rates in forests slightly raise relative humidity of the air by shading the soil Thus, taller and denser windbreaks provide ideal conditions for coffee growth but also conditions for Broca inf estation. However, coffee plantations grown with shade trees have an advantage in that Beauvaria bassiana may work better. Most importantly, windbreaks provide habitats for predators of Broca such as birds, ants, beetles, lizards, and parasitoids. These ca n serve as natural biological control methods that can be used instead of or in conjunction with human methods. There is a direct correlation between the amount of forest cover and the number of Broca eating birds. Resea rch done by Karp et al (2013) found that borer consuming birds increased in abundance and exerted stronger control on borer populations on plantations with higher forest element cover He found a reduction in Broca of up to 50% during his study. Thus, it seems as if it would be beneficial to increase the amount of windbreaks and natural forest on coffee farms in order to encourage natural predation instead of using chemical and non native biological controls. Karp et al (2013) found at least five bird species shown to predate on Broca in Cos ta Rica Rufous capped Warblers, Yellow Warbler s, Rufous breasted Wren s, Buff throated Foliage Gleaner s, and White tailed Emerald s By creating suitable habitats for these species on Costa Rican coffee plantations, it would be possible to reduce Broca incidence as well as maintain native bird populations. Though
PAGE 13
Factors that affect broca incidence in coffee plants Simon 13 Broca beetles are a relatively recent phenomenon, I think that it could be possible that more native bird species will adapt to include Broca within their diets. Ultimately, the use of traditio nal coffee farming with agroforestry techniques has be en shown to help mitigate the effects of Broca. The Height of the Trees and t he Length of the Branches I found that t he height of trees was statistically significant. The height of trees can correlate with the age of plant. Though age of the plants was not consi dered significant in my modeling other workers found that older larger trees provide more habitats for CBB and make control more difficult (Arisitizbal et al 2016) The taller and big ger trees within my study sites could potentially lead to increased Broca incidence. The trees within C2 were the oldest and biggest trees overall and had a larger percentage of Broca infestation. Furthermore, female Broca have large eyes and respond to mo vement indicating they use their sight to find berries. T he length of the branches could lead to higher Broca incidence because they increase the visibility of coffee berries to the beetles. (Arisitizba l et al 2016) Furthermore, t he longer branches could also indicate an older branch, which have been shown to have higher Broca infestation In fact, 1 and 2 year old branches showed between 52 and 65% less attack by Broca than 3 year old branches (Rojas 2012) These findings indicate the need for strict and careful pruning schedules to reduce the amount of vectors available for Broca infestation. Other Factors Age of P lants The age of the plants was statistically significant This was due in part to the fact that I only had the age of the plots and the average age of the plants within those plots I knew the planting date of each plot but I did not know the specific age of each plant. This could indicate that the average age of the plot does not significantly affect the Broca inc idence. Instead age data must be take n on an individual plant level. An other indicator of likely Broca infestation would be the flowering period and t he development of the berries a s CBB females prefer older berries 150 240 days (>20% dry weight) ove r younger berries < 90 days ol d (Arisitizbal et al 2016). The E levation of the P lots E levation was not a significant factor My plots only varied at most by 50m in elevation, which would explain the lack of significance in my data. However, geographic elevation is i mportant because at high elevations co ffee berry borer populations are less abundant due to lower temperatures (Avelino 2012) It has been shown that Broca reach a maximum reproductive rate at an average daily temperature of 26.7 C Due to the fact that Monteverde has an annual average of 16 18 C it would make sense that there would be a reduced incidence overall The H e ight of the B ranches T his height of the b r anches was not significant in my regression modeling. I expected to see a significant difference in the percentage of Broca infested berries
PAGE 14
Factors that affect broca incidence in coffee plants Simon 14 based on the height at which the branch was located. Arisitizbal et al (2016) found that h ighest CBB infestations levels were observed on lower branches in contrast to the middle and top of the tree I did not find the same trend in my data but instead found that the height of the tree and the length of the branches we re significant. The Area of t he Sites The size of the study sites was not significant in my regression modeling As Broca do not nece ssarily require large distances of travel between boring site s, a small site could have a large concentration of Broca incidence. This is supported by the fact that the i nitial infestations of Broca in coffee plantations assume an aggregated distribution ( Arisitizbal et al 2016) The females release hor mones after initial colonization, which causes the flocking of ot her females to that area. As the population increases however the spatial distribution of Broca become more regular ( Arisitizbal et al 2016). This indicates that the size of a coffee plot does not necessarily affect whether or not Broca choose to infest it but rather, other factors have a stronger influence on their infestation behavior. The Type o f Windbreak Used The type of windbreak was not found significant in my regression modeling. The most important factors related to windbreak seemed to be the distance of the plot from the windbreak. An intere sting study would be to see how different ages of windbrea k affect the Broca incidence on coffee farms. It is difficult to say which of these factors affect the percentage of berries infested with Broca the greatest. The significant results indicate that the aforementioned factors are the ones that have the lar gest effect on my study sites and indicate the factors that comprise the habitat preference of Broca beetles on the Life Monteverde coffee plantation. A further experiment could analyze the differences between the three sites to understand why C2 had the l a rgest amount of Broca infestation in comparison to the other two sites. ACKNOWLEDGMENTS I would like to thank my primary advisor Sofia Arce Flores for helping me with my experimental set up as well as helping me with my statistical analyses. I would lik e to thank my secondary advisor Frank Joyce and Andres Camacho for helping me with my stepwise regression models. Most importantly, I would like to thank Life Monteverde for allowing me to use their farm as my study site. Thank you to my advisor there, Dan iel Vargas, and to Guillermo Vargas, Juan Manuel Santa Mara, and Jose Luis Vargas for all their help with getting me data from the farm. Finally special a cknowledgements to Jerson Santam ar a and Alfredo Mairena for their help showing me how to find Broca and find variables for my study
PAGE 15
Factors that affect broca incidence in coffee plants Simon 15 LITERATURE CITED Ar istizbal, Luis F., Alex E. Bustillo, and Steven P. Arthurs. "Integrated Pest Management of Coffee Berry Borer: Strategies from Latin America That Could Be Useful for Coffee Farmers in Hawaii." INSE CTS 7.1 (2016): n. pag. MDPI. 3 Feb. 2016. Web. 07 Nov. 2016. . Borbn, O. 1994. Manejo integrado de la broca del fruto del cafeto: acciones a desarrollar. ICAFE. San Jos, Costa Rica Camilo, Jos Efran, Frank F lix Olivares, and Hctor Antonio Jimenz."Fenologa Y Reproduccin De La Broca Del Caf(Hypothenemus Hampei Ferrari) Durante El Desarollo Del Fruto." Agronoma Mesoamericana 14.1 (2003): 59 63. 2003. Web. 07 Nov. 2016. . Crdenas, Mara Del Carmen, Rodolfo Valentino Marcano Brito, Humberto Giraldo, and Angel Aquino. "Biologa De La Broca Del Caf, Hypothenemus Hampei Ferrari (Coleoptera: Curculionidae) Bajo Condiciones De Campo, En El Estado Tchira, Venezuela. ENTOMOTROPICA 22.2 (2007): 49 55. Web. Ceja Navarro, Javier A., Fernando E. Vega, Ulas Karaoz, Zhao Hao, Stefan Jenkins, Hsiao Chien Lim, Petr Kosina, Francisco Infante, Trent R. Northen, and Eoin L. Brodie. "Gut Microbiota Mediate Caffeine Detoxification in the Primary Insect Pest of Coffee." Nature Communications 6 (2015): 7618. Web. 7 Nov. 2016. "Chapter 3 Field Management & Planting Trees." Arabica Coffee Manual for Lao PDR. Food and Agricultural Organization of the United Nations, n.d. Web. 07 Dec. 2016. . "Coffee Berry Borer (CBB)." Coffee Berry Borer (CBB) University of College of Tropical Agriculture and Human Resources, n.d. Web. 07 Nov. 2016. Karp, Daniel S., Chase D. Mendenhall, Randi Figueroa Sand Nicolas Chaumont, Paul R. Ehrlich, Elizabeth A. Hadly, and Gretchen C. Daily. "Forest Bolsters Bird Abunda nce, Pest Control and Coffee Yield." Ecology Letters 16.11 (2013): 1339 347. ResearchGate. Web. 6 Dec. 2016. Lang, Tom. "Documenting Research in Scientific Articles: Guidelines for Authors: 3. Reporting Multivariate Analyses." CHEST Journal 131.1 (2007): 317. Web. Rojas, Mainor. "Manejo Sostenible De La Broca Del Caf (Hypothenemus Hampei) Mediante Poda Sistemtica Del Cafeto En Costa Rica." Agronoma Costarricense 36.2 (2012): 71 79. Centro De Investigaciones Agronmicas Web. 6 Dec. 2016.
PAGE 16
Factors that affect broca incidence in coffee plants Simon 16 Tarazi, Reem, and Catherine McKeever. "Stepwise Multiple Regression." Schatz.sju.edu. Ph ilip Schatz, n.d. Web. 6 Dec. 20 16. . A ppendix A Table 1 A Site 1 (C2) La Paila Site Specifications Plot Distance from Road (m) Distance from Back Windbreak (m) Distance from Side #1 Windbreak (m) Distance from Side #2 Windbreak (m) 1 9 33.5 31.5 2 9 33.5 22.5 3 9 33.5 .5 4 18 24.5 18 5 27 15.5 18 6 36 6.5 18 Table 2 A Site 2 (C14) Los Pinos Site Specifications Plot Distance from Road (m) Distance from Back Windbreak (m) Distance from Side #1 Windbreak (m) Distance from Side #2 Windbreak (m) 1 15.5 9 2 6.5 18 3 24.5 0 4 15.5 9 5 15.5 9 6 15.5 9 Table 3 A Site 3 (C18) Guayabo Site Specifications Plot Distance from Road (m) Distance from Back Windbreak (m) Distance from Side #1 Windbreak (m) Distance from Side #2 Windbreak (m) 1 3 36 18 18 2 3 36 9 27 3 3 36 27 9
PAGE 17
Factors that affect broca incidence in coffee plants Simon 17 4 12 27 18 18 5 21 18 18 18 6 30 9 18 18 Table 4 A : The number of misidentified berries as Broca infested within each plot per study site Area Plot Misidentified Broca C2 1 3 C2 2 5 C2 3 2 C2 4 1 C2 5 0 C2 6 3 C14 1 2 C14 2 3 C14 3 1 C14 4 0 C14 5 0 C14 6 6 C18 1 0 C18 2 0 C18 3 1 C18 4 2 C18 5 3 C18 6 0 Appendix B From T arazi, Reem, and McKeever (2016) : Assumptions to be met in Stepwise Multiple Regression : A. There is no multicolinearity among predictors. That is, there may not be any correlation between predictor variables. 1. Multicolinearity: (def.) Moderate to high interco rrelations among predictors. 2. Multicolinearity can lead to three problems: a. It severely limits the size of R, as the predictors are going after much of the sa me variance in Y (predictant). b. It makes determining the importance of a given predictor difficult because the effects of the predictors are confounded due to the correlation among them. c. It increases the variance o f the regression coefficient. The greater the variance, the more unstable the prediction equation will be.
PAGE 18
Factors that affect broca incidence in coffee plants Simon 18 B. There must be at least 10 15 (conservative count) or 20 (the limit for discriminant analyses as well) subjects per predictor variable as a small N loaded onto each variable will yield a weaker prediction equation. C. One must guarantee that the results are not skewed by outlier values, therefore ensuring that the results are indeed a reflection of the relationship between predict or and predictant variables. 1. will cause the range of data to change thus creating new outliers, however a higher amount of the varianc e can now be accounted for. 2. O ne is advised to remove outl qualities that naturally account for variance (in data obtained from humans) must be accounted for. Otherwise, removal of too many outlying cases can leave a le to the population. Figure 1B : The informal check ran prior to stepwise regression. It is a sca tter plot of residuals on the y axis and fitted values (estimated responses) on the x axis The plot is used to detect non linearity, unequal error varian ces, and outliers. As seen in my residuals, outliers 31, 33, 38 occur which indicates the need for a correlation matrix to remove insignificant variables prior to regression analysis. Plotting done in R Commander
PAGE 19
Factors that affect broca incidence in coffee plants Simon 19 Figure 2B : The following provides the means and standard deviations for the test variables. This can indicate skewed distributions as well as outliers within my data. This is used in conjunction with my other informal check. Data created by R Commander
PAGE 20
Factors that affect broca incidence in coffee plants Simon 20 Figure 3B : This is my correlation matrix. It was used to determine which factors should be included in my stepwise regression modeling. Data created by R Commander i The results for the variables I included in my analysis can be biased by the significant variables that I didn Stepwise regression is a great tool and can get me close to the correct model. However, because the stepwise algorithm follows simple rules and it knows nothing about the underlying process or subject area.
