Bull. en/. Res. 74, 163-174 163 Published 19!14

The influence of overhead shade and fertilizers on the Homoptera of mature Upper-Amazon cocoa trees in Ghana

C. A. M. CAMPBELL * Cocoa Research Institute. P.O. Box 8. New Tafo. Ghana

Abstraet Homoptera were assessed by examining branches from mature cocoa trees in an Upper-Amazon 3 x 33 shade x nitrogen, phosphate and potash ferti­ lizer experiment at Tafo, Ghana. Trees in territories of the Crema togas­ ter c/ariventris Mayr and Acantholepis capensis Mayr were sampled. Planococcoides njalensis (Laing), Planococcus dtr; (Risso), Phenacoccus hargreavesi (Laing), Pseudococcus concavocerarii James, Maconellicoccus ugandae (Laing) and Mesohomotoma tessmanni (Aulm.) were more abundant on unshaded trees than on shaded ones. Steatococcus sp. and Stictococcus sjostedti Ckll. also occurred most frequently on unshaded trees. The abun­ dance of five other infrequently recorded species was unaffected by overhead shade. A species of Gascardia near G. zonata (Newst.) was more common as shade increased. Planococcus dtri increased with increasing nitrogen where A . capensis was dominant, and decreased with increasing potassiu~ ~r. heavily shaded trees where C. clariventris was dominant. M. tessmanni decreased with increasing potassium both on unshaded trees where A . capensis was dominant al'ld on trees which had received most phosphate where C. c/ariventris was dominant. No responses to fertilizers were detected among other species. An estimated 100, 99 and 83 %, respectively, of unshaded, Iightly and heavily shaded trees were infested with mealybugs, vectors of cocoa swollen shoot virus. These estimates presage little change in the rate of spread of virus due to the continued shade thinning over cocóa in Ghana.

Introduction Cocoa in Ghana has traditionally been grown under shade provided by mixed species of trees that are mainly survivors of the canopy trees of earlier forests. Johnson (1962) pointed out that the influence of shade trees on cocoa is complex. They not only affect the growth and physiology of the cocoa (Hutcheon, 1977) and its herbivores, but there is also a dynamic interchange between the faunas of the various species of shade trees and the cocoa beneath. Johnson foresaw exacerbated pest problems accompanying the reduction in shade over cocoa then proceeding in the Eastern Region and elsewhere in Ghana. His prediction of a change in the complex of mealybug species was confirmed (Bigger, 1981h; Campbell, 1983). Mealybugs are partieularly important pests in West Afriea as most species inhabiting cocoa trees are potential vectors of the virus causing cocoa swollen shoot disease (Roivainen, 1976).

• Present address: East Mailing Research Statíon, Maidstone. Kent ME 19 6BJ. UK. 164 C. A. M. CAMPBELL

Despite Johnson's warnings, the decline in the numbers of shade trees in the cocoa belt of the Eastern Region has continued due to the increased demands on land to provide food for the expanding populace and the aftermath of the devastation to cocoa farms caused by swoHen shoot disease (Legg, 1979). A stimulus for planned shade reduction was provided by research findings that linked improved yields from coco a with deshading and the use of fertilizers (reviewed by Alvim, 1977). In the present report, the influence of overhead shade and nitrogen, phosphate and potash fertilizers on the Homoptera of cocoa within an Upper-Amazon factorial shade­ and-fertilizer experiment at Tafo, Ghana, is assessed. FuH descriptions of the design and agronomic progress of the experiment are given by Ahenkorah & Akrofi (1968, 1977) and in annual reports of the Cocoa Research Institute at Tafo, between 1962-63 and the presento Unlike the shade by several tree species on farmers' plantations, the single species Terminalia ivorensis provided the shade treatments. Bigger (1981a) made a similar study on part of an Amelonado shade-and-fertilizer experiment shortly before the cocoa trees were grubbed up in 1972. Bigger (1973, 1976) also used multivariate statistical methods to assess the influence of various environmentai factors, including shade, on the fauna of cocoa. In aH of these studies, the species of dominant was the major variable affecting the abundance of Homoptera. Bigger (1981a) argued that the dominant ants probably obtain the bulk of their energy supplies from the produced by Homoptera. However, individual species of ants and their mutualistic Homoptera were often associated with particular shade regimes. Consequently, Bigger was unable to assess the influence of shade on the Homoptera within the territory of a single ant species. That problem was avoided in the present study.

Methods The 3 x 33 shade x nitrogen, phosphate and potash fertilizers experiment was repli­ cated as four complete block s (Fig. 1). The 13 ha were planted in 1959 with cocoa of open-poHinated Upper-Amazon parentage spaced at approximately 3 x 3 m. Fertilizer subplots each consisted of 20 trees (4 x 5) separated from adjacent subplots by single rows of guards, and from neighbouring shade plots by at least three rows of guards. The rates of fertilizer application are given in Table 1. The T. ivorensis shade trees were spaced at 17'2 x 17'2 m in lightly shaded plots and at 8'6 x 8'6 m in the heavily shaded ones. Fertilizers were not applied to Blocks 11 and 111 after 1971. Hence, any responses to fertilizers among Homoptera in those blocks were due either to differences in sap composition, caused by differences in residual effects of the fertilizer treatments, or to differences in the physiological status of the trees.

TABLE 1. Annual rates of fertilizer application (kg/ha) * Triple super- Muriate of Urea phosphate potash (N) (P) (K) Level ... 1 2 1 2 1 2 1963-1971 84 140 34 67 84 168 1972 onwardst 84 140 67 134 112 224

• From Akenkorah & Akrofi (1977). t Only applied to Experimental Blocks I and IV; no fertilizer was applied lo Blocks 11 and II1 afler Ihe 1971 season.

A direct assessment of insect abundance foHowing the formal design of the experiment was inappropriate owing to the non-uniform patterns of the territorial mosaics of the dominant ants. During June and July 1974, aH the approximately 14 000 cocoa trees were examined from ground level, and all species of ants observed on the trees were recorded. Maps were prepared of the territories of the dominant ants, and lists were compiled of the numbers of trees in every fertilizer subplot with each dominant ant. The lists and maps were used to prepare a sampling plan encompassing aH replicates of the experiment within the territories of the smallest number of species of domihant ants (Fig. 1). Two species of HOMOPTERA OF AMAZON COCOA 165 A B .: ...... :.... : .. :: ...... : 'Ó1Ó ciii 21i • o',·· ".

.:'::' .::: : . '.~'.:. ',:. :

201 .'. 9'~l .. :'1'R? 1";"2' ~~. ~;{ ) 9W ?~'q boí .:.:.: lQQ (;iiq :. . ..: ..... : ..

.,.'.: 101....

fi2ÚO 600iii 01{ 661.: : 020110202 - 111= 021 201 210 :121102' : 200101" 100' - KEY 112022 21.? 2}J ~0.qr9.: 211:qQi 012 -: A. ======::======:: :======:: Shad.~ ------_ ... _------none 1~~ ?!cf J!1}: j1.0_112~~f~ . fl2.10} 1Z1: : O 100 ?~º 1Q2 ~ ~1c2 9?º ººQ. :201. 001101: - ligh t 201202 012 ~ 21261 f 20L : 000 220 022 : ~ poc oi~ ifº ~ ??Q1QQ !~~'. :200 002 021 ~ ~ heavy ~26 200061 ~ 001 202[ i L .222020120 : 02f 212 o:¡ó § 22f 126 012 •• 1fo· fo-o· Of2 ~ B. Ant territory - - . - A. capensis 221002101 :-200101211:011112 221 :- []-- 222121~ 022: 022121~ 010 21-0· 111: 202 ~ . '. C. clariventris 2111fi 020: 102 021 99~lt 21-1'lii 6fo" - D Fig. l.-A. plan ofthe Upper-Amazon shade-and·fertilizerexperiment at Tafo showing shade treatments; B. the same plan enlarged. showing territories of Acantholepis capensis and Crematogaster c1ariventris; sampled plots are indicated by the levels of nitrogen, phosphate and potash fertilizers applied. respectively (see Table 1). 166 C. A. M. CAMPBELL

ants were sufficiently widespread to permit assessment of the abundance of the Homoptera associated with them under most shade and fertilizer treatments. The fertilizer subplots sampled are identified in Fig. 1 by three digits corresponding to the amounts of nitrogen, phosphate and potash fertilizers they had received (Table 1). Using a long-handled pruner from ground level, a single branch was cut from one of the trees with the chosen species of ant on the day of sampling. Campbell (1983) provides details of sample preparation. Records were made of the identity and numbers of all Homoptera, and the numbers of lea ves, vegetative buds and shoots examined. AH examina­ tions were made in the laboratory using stereo-microscopes. Where Acantholepis capensis Mayr was dominant, the assessment was made during August and September 1974, which Gibbs & Leston (1970) describe as the second wet-dull season. T. ivorensis is deciduous in the dry-sunny season (October-Iate February). Consequently, sampling was stopped until T. ivorensis had developed new foliage. Sam­ pling of trees in the Crematogaster c/ariventris Mayr territories was from late February until early April during the first wet-sunny season in 1975.

ResuIts The territory of A. capensis was unusual as it was almost the only dominant ant in Block I and in the Iightly shaded plot of Block II (Fig. 1). That 3'6-ha rectangle of coco a was isolated from neighbouring cocoa plots on two sídes by roads and bordered a residential area on the other two sides. A. capensis was uncommon elsewhere on the site other than in the nearby part of the heavily shaded plot in Block 11. Repeat surveys in 1976 and 1978 (Campbell, unpublished) showed that A. capensis was eventually displaced almost everywhere by other genera more typical of the dominant ants found in cocoa farms in Ghana (Bigger, 1981a). C. c/ariventris was the dominant ant in the rest of the experimental site, as it had been in an Amelonado shade-and-fertilizer experiment (Bigger, 1981a). Frmrteen species "f HO'T'optera were encountered (Table 11). A. capensis was not intimately associated with any honeydew-producing homopteran species, although Toxopt­ era aurantii (Boyer de Fonscolombe) was found exc1usively in its territory, and a con­ tingency test of association showed that Steatococcus sp. also occurred more frequently with

TABLE II. The abundance o[ Homoptera on cocoa branches within the territories o[ the ants Acantholepis capensis and Crematogaster c1ariventris

% branches Total infested Dominant ant ... A. cap. e. claro A. cap. e. claro X" Pseudococcidae Planococcoides njalensis 34 154 21 57 25.00 Planococcus citri 384 322 64 81 6'02 Phenacoccus hargreavesi 46 36 25 32 124 Pseudococcus longispinus 5 1 4 1 0'99 Pseudococcus concavocerarii 2 19 2 16 1212 Pseudococcus calceolariae 1 2 1 3 0'79 Maconellicoccus ugandae 10 23 7 12 154 Combined Pseudococcidae 482 557 70 86 593 Mar¡¡arodidae Steatococcus sp. 64 3 31 4 2038 Gascardia sp. nr G. zonata 22 7 15 6 135 Cribrolecanium andersoni 6 O 4 O 291 Stictococcidae Stictococcus sjostedti 39 1590 9 49 3760 Diaspididae Aspidiotus elaeidis 4 O 4 O 291 Psylhdae Mesohomotoma tessmanni 1409 1095 63 47 359 Toxoptera aurantii 74 O 10 O 8'34 • Contingency test that the 108 and 77 branches sampled from the territories of A. capensis and e. clariventris respectively, were infested with equal frequency.. . HOMOPTERA OF AMAZON COCOA 167

that ant (P < 0-001)_ Two relatively uncommon species, Cribrolecanium andersoni (News­ tead) and the diaspidid scale Aspidiotus elaeidis Marchal, were found only with Acantholepis capensis, but too infrequently to indicate ar.y positive associations. Dias­ pidid sea les are not solicited by ants for honeydew. Stictococcus sjostedti Cockerell was almost the only species of Homoptera found in the carton tents of Crematogaster clari ven tris _ However, contingency tests showed that mealybugs occurred more commonly with that ant (P < O-OS), particularly Planococcoides njalensis (Laing) (P < 0-001), Planococcus citri (Risso) (P < O-OS), and Pseudococcus con­ cavocerarii James (P < 0001)_ With the exception of P. longispinus (Targioni-Tozzetti), all species of Pseudococcidae occurred more frequently with C. clariventris than with A. capensis_ There were averages of 422 Homoptera individuals per branch with C. c1ariventris and 19A/branch with A. capensis_ However, about half of those with C. c1ariventris were individuals of S_ sjostedti. Discounting that species gave broadly similar average numbers of 21-6 and 19-1 individuals/branch with C. c1ariventris and A. capensis, respectively. Overall, P1anococcus átri was the mosr widespread species. Ir was found on 71 % of branches, whereas Mesohomotoma tessmanni (Aulmann), numerically the most abundant species, was found on 56 'M. of branches_ AII other species were detected on fewer than half the trees. The influence of fertilizers on the abundance of the Homoptera was in most instances relatively slight. Fewer Homoptera were encountered in experimental Blocks I and IV (which received regular fertilizer treatments), than in Blocks 11 and I1I, where fertilizers were last used in 1971. On lightly shaded trees with A. capensis there were averages of 11-8 individuals/branch in Block I and 28-1/branch in Block I\. Similarly, with C. c1ariventris there were averages of 39-2 individuals/branch in Block IV and 44-1/branch in Blocks 11 plus III. Nevertheless, sorne responses to specific fertilizers were evident for both P. citri and M. tessmanni (Table 111). No significant responses to fertilizers were detected among the remaining 12 species_ H0wever. for many of the <;pecie<; nnly very large fertilizer effects would have been detectable beca use of the high frequency of uninfested branches.

TABLE 111. The influence of nitrogen. phosphate and potash fertilizers on mean log numbers (x + 1) o f homopterous individuals per branch Level Dominant ant and homopteran Treatment O 2 s_e_d_ Acantholepis capensis Planococcus cüri N 0-324 0-396 0-526 0-0936 P 0-433 0-347 0-467 0-0936 K 0-360 0-393 0-494 0-0936 Mesohomotoma tessmanni* N 0514 0-643 0-520 0-1193 P 0-559 0-622 0-497 01211 K 0-674 0-563 0-441 0-1202 Sot:K 1-284 1-216 0-553 0-2066 Crematogaster clariventris Planococcoides njalensis N 0-298 0-298 0357 0-0902 P 0-329 0-361 0-263 0-0902 K 0-250 0-342 0-361 0-0902 Planococcus cüri * N 0-457 0-623 0-575 0-0837 P 0-549 0504 0-603 0-0826 K 0574 0-616 0-466 0-0845 S,t:K 0-652 0-559 0-275 0-1441 Stictococcus sjostedti N 0-682 0-648 0530 0-2140 P 0-743 0-384 0-733 0-2140 K 0-687 0-729 0-445 0-2140 Mesohomotoma tessmanni* N 0-573 0-619 0-494 02042 P 0-519 0-506 0-661 0-2061 K 0-720 0-428 0-538 0-2095 P,t:K 1307 OAOI 0274 03545 * Means adjusled for covariales. see lex!. t Significant interaclions were all wilh potash_ S. = unshaded; S, = Terminalia ivorenSIS shade trees at 8 Ó x 8-6 m- P, = highest level of phosphate_ - - , 168 C. A. M. CAMI'BELL

The data from the H I branches in Block I where A. capensis was dominant were used to assess the response of Homoptera to regular fertilizer applications. There was a positive association between nitrogen and the density of P. citri (P < 0·05), while the apparent positive response of P. cilri to potassium was not significant. The density of M. tessmanni was negatively associated with potassium, but only on the unshaded trees (P < 0·01). About seven times as many psyllids were present on control trees as on those which had received the highest rate of potash. In the plots dominated by C. clariventris, samples were variously from Blocks 11, 111 and IV. As Blocks 11 and III had last been treated with fertilizers four years previously, it was not surprising that the responses to fertilizers were less clear. As in Block 1, P. citri was more abundant on trees treated with nitrogen than on untreated ones, hut the differences were not significant. However, on the trees in the heaviest shade, P. citri was negatively associated with potassium (P < 0'05). M. tessmanni showed a similar negative response to potassium, but only on trees that had received the highest rate of phosphate (P < 0·01). The difference in psyllid density was about 20-fold. The results for Planococcoides njalensis and S. sjostedti were not slgnificant. The effects of several potential covariates were investigated. M. tessmanni prefers f1accid extension growth (Entwistle, 1972). The presence of such tissue, when used as a covariate, significantly reduced the error variances in both analyses (P < 0·01 in both cases). No covariate was effective with the counts of Planococcus citri in the analysis from Block 1, but in the other analysis the number of vegetative buds on the sample (trans­ formed to square roots) was effective (P < 0·01). The numbers of leaves or shoots (similarly transformed) were alternative measures of sample size and were almost equally effective covariates. In the analyses of data from the territory of C. clariventris, acode identifying samples as either from Blocks 11 and 111 or from Block IV gave no significant reduction in the residual variance. A statistical analysis of the influence of shade on den sities of the Homoptera was inadvisable owing to non-orthogonality of the "ampling plan. itself dictated \"Iv the oattern of the ant mosaico Although samples were taken from every shade treatment for all four block s of the experiment, the number of samples varied between two and 27 in the different plots. With such variability in sampling frequency, it was not possible to estimate standard errors for mean densities. Nevertheless, the mean log (x + 1) densities are informative and are presented in Table IV, together with the percentage of branches infested with each species. Contingency tests were carried out on the frequency with which branches were infested. The relative favourability of a shade treatment for each species was assessed from the percentage which specimens of that species contributed to either the total mealybug population or to the total of other Homoptera within each shade regime. The frequency with which branches were infested declined with increased shade, both for mealybugs (P < 0-(01) and other Homoptera (P < 0'05). With the exception of a species of Gascardia near G. zonata (Newstead), all common species were more numerous on unshaded trees than on those most heavily shaded (Table IV). Mealybugs were on average 3'5 times as abundant on unshaded as on heavily shaded trees, with intermediate numbers on lightly shaded trees. Similarly, other Homoptera were 2'7 times as numerous on unshaded as on heavily shaded trees. The percentage of branches infested dec1ined as shade increased for P. citri (P < 0·001), Pseudococcus concavocerarii (P < 0·05), Maconellicoccus ugandae (Laing) (P < 0'01), Steatococcus sp. (P < 0·01) Mesohomotoma tessmanni (P < 001). Phenacoccus hargreavesi (Laing) showed the same trend, but the differences were not significant. The frequency with which branches were infested with the species of Gascardia followed the opposite pattern, with heavily shaded trees most often infested, but that result was not significant. Only part of the differences between shaded and unshaded trees was caused simply by the greater numbers of branches infested in the latter. The levels of infestation were also higher on unshaded branches; for example, Planococcus citri was found on lJ4

TABLE IV. The injluence 01 overhead shade on the abundance and frequency 01 occurrence 01 Pseudococcidae and other Homoptera Geometric mean Percentage Percentage no. of individuals of totals of branches per branch (within shade) infested , . , , , ~ Shadet So S, S2 So S, S2 So S, S2 X,. a. Pseudococcidae Planococcoides njalensis 071 034 056 11 15 31 50 25 40 890 Planococcus cilri 477 128 101 75 56 55 94 66 55 2125 Phenacoccus hargreavesi 061 049 021 10 22 12 40 25 21 568 Pseudococclls longispinus 001 004 000 O 2 O 2 5 O 320 Pseudococcus concavocerarii 012 004 0'04 2 2 2 15 5 4 639 Pseudococclls calceolariae 000 003 000 O 1 O O 4 O 400 Macone//icoccus ugandae 015 006 001 2 3 O 19 5 4 \031 Total 637 228 183 \00 \01 \00 96 74 62 1761

b. Other Homoptera Slealococcus sp. 026 029 0'04 4 7 2 27 25 6 962 Gascardia sp. nr G. zonala 009 007 016 1 2 6 \O i5 19 4'31 Cribrolecanium andersoni 000 003 002 O 1 1 O 4 2 212 Sliclococcus sjosledli 1 14 077 0'97 16 19 36 31 28 19 211 Aspidiotus elaeidis 000 003 001 O 1 O O 4 2 212 Mesohomoloma tessmanni 565 268 147 78 66 55 73 55 42 1071 Toxoptera aurantii 011 017 001 1 4 O 4 \O 2 432 Total 725 404 268 100 100 \00 90 80 68 842 • Contingency tests that the 52 branches examined from So, 80 from S, and 53 from S2 were infested with equal frequency. t So = unshaded; S, = Terminalia ivorensis at 17'2 x 17'2 m; S2 = T. ivorensis at 8'6 x 8'6 m.

unshaded branches and 42 % of heavily shaded ones, yet it too was about four times as numerous on unshaded branches. Other species were less affected by the absence of shade. For example, Planococcoides njalensis and Stictococcus sjostedti were most numerous on unshaded trees. Nevertheless, their numerical responses were considerably less pronouDcec than those of Planococcus citri and M. tessmanni. As percentages of the total insects within a shade treatment, both Planococcoides njalensis and S. sjostedti were relatively more common on heavily shaded trees than on unshaded ones. P. njalensis constituted aImost a third of the mealybugs on heavily shaded trees, but only one mealybug in nine was this species on unshaded trees. Similarly, S. sjostedti formed 36 % of the total of other Homoptera on heavily shaded trees and just 16 % on unshaded ones. By contrast, Pseudococcus concavocerarii constituted about 2 % of the numbers in all three shade regimes.

Discussion The trend to more open woodland conditions in the cocoa belt of the Eastern Region of Ghana has continued since Johnson (1962) first wamed about the likely effects of this policy on cocoa pests. Undoubtedly the widespread removal of shade trees and the considerable loss in numbers of cocoa trees caused by swollen shoot disease (Legg, 1979) will have had sorne impact, as yet unreported, on the c1imate of the regíon. The whole plots used here were relatively large (over 1 ha for each replicate of a shade treatment) and were probably adequate for assessing the comparatively local influence of overhead shade on the Homoptera within a regional environment that has altered progressively over several decades. The experiment provided an opportunity to examine shade effects within the territories of two species of ants. It also made possible assessment of the influence of fertilizer treatments on the abundance of the Homoptera associated with each of those ants. Most species of mealybug occurred more frequently with C. c1ariventris than with A. capensis, as did S. sjostedti (Table JI), the only homopteran found in the carton tents of C. c1ariventris. Clearly mealybugs benefited from the mutualism between S. sjostedti and C. clariventris. Possibly they derived sorne protection from natural enemies, and sorne individuals may have been tended casually by the ants for honeydew. Nevertheless, mealybugs were not abundant with either species of ant. There were averages of 7'2 mealybugs/branch with C. clariventris and 4'5/branch with A. 170 C. A, M, CAMPBELL

capensis. Both figures are low compared with the overall average of 15'9 mealybugs/branch obtained in a concurrent study of the branch sampling technique (Campbell, 1983). Twenty of the 24 sites examined in that study were from the Amazon shade-and-fertilizer experiment. Only three relatively uncommon mealybugs (P. concavocerarii, P. calceolariae (Maskell) and Maconellicoccus ugandae) were slightly more abundant during the present study than in the earlier one, and all the common species were less numerous. The low density of Steatococcus sp. with C. clariventris may be unrelated to the presence of that ant, as Bigger (1976, 1981a) recorded positive associations between these species. In Trinidad, Fennah (1959) investigated the effects of nitrogen, phosphate and potash fertilizers and overhead shade on the abundance of mealybugs (mostly Planococcus cilri) on trees of three lCS clones 2,5 years after planting. He argued that differences in mealybug abundance were related to nutritional suitability of the sap, particularly to soluble nitrogen levels. The influence of plant nitrogen levels on herbivorous insects has received considerable attention (McNeill & Southwood, 1978; Mattson, 1980). As a general rule, higher levels of plant nitrogen are beneficial for many species of Homoptera. Exceptions occur, for example when specific amino acids are limiting, or when antifeedants render the sap unpalatable. More recentiy, Prestidge (1982) showed that sorne grassland Homoptera responded only to relatively narrow ranges of nitrogen concentrations in sapo Other than for P. citri, which showed a positive numerical response to increased nitrogen here and in Fennah's (1959) study, there is no evidence that any other species among the Homoptera on cocoa responded to nitrogen. Mesohomotoma tessmanni prefers young tissues (Entwistle, 1972), which are sites of protein synthesis and are consequently naturally ríeh sources of nitrogen. Possibly little additional enrichment with soluble nitrogen occurs at those sites following the use of a nitrogenous fertilizer. The present results indicate that levels of potassium may be relatively more important than nitrogen for M. tessmanni and probably also for P. citri. Fennah (1959) reported that P. citri responded negatively to potassium. In the present study too (Table IlI), a negative response to potassium was observed, but only on heavily shaded trees where C. clariventris was the dominant ant. Just eight individuals of P. citri were recorded on branches from the heavily shaded trees where A. capensis was dominant, too few to reveal any response. Fennah (1959) presented his data in full, thereby enabling a reassessment of his results (Table V) that revealed a previously overlooked but significant interaction between potassium and shade (F4 105 = 4'11, P = 0'005). As in the present study, the negative response of P. citri to potassium was only evident on heavily shaded trees (Table Vb; 15 and 25 % offull sunlight). EIsewhere, Fennah (1954) observed that the soils on the si te of the trial were " ... markedly sub-optimal for both N and K ... and ... that trees in the NK plots ... had received the least nutrient amendments compatible with correction of known major deficiencies ...". This suggests that a natural deficiency of potassium on untreated plots may have led to the higher numbers of P. citri on the heavily shaded trees. In the C. clariventris territory, many of the samples were from Blocks II and III (Fig. 1) which had received no fertilizer on earlier treated subplots since 1971. Ahenkorah & Akrofi

TABLE V. The influence of overhead shade and fertilizers on the abundance (mean score per tree) of mealybugs (mostly Planococcus citri) on 2 '5-year-old Criollo and Trinitario cocoa in Trinidad (data from Fennah (1959))

(a) Main effects of nitrogen. phosphate and potash Level N P K O 0'563 0,675 0,796 1 0'758 0'646 0'525 S.e.d. = 0'085 (b) Interaction between potash and overhead shade % of full sunlight ~I • , of K 15 25 50 75 100 O 1396 1'583 0'125 0812 0'063 1 0'625 1'000 0'146 0708 0146 S.e.d. vertical comparisons = 0'191; other comparisons = 0'403. HOMOPTERA OF AMAZON COCOA 171

(1968) reported that after seven years of continuous cropping there were no signs of potassium deficiency among untreated controls, although the mean levels of exchangeable potassium were then at the threshold of the undesirably low level. They also noted that pod husks, removed in harvesting. cOlltained more than 3 'x, of potassium. It seems likely, therefore, that a further seven years of cropping led to suboptimallevels in untreated plots. Reduced yields from previously treated subplots similarly point to likely partial exhaustion of potassium. It is also noteworthy that the rates of potash fertilizers were doubled from 1972 onwards (Table 1). A shortage of potassium in the soil could provide an explanation for the higher average densities of Homoptera in Blocks 11 and III compared with Blocks I and IV, to which fertilizer was applied regularly. Compared with the wealth of information on the importance of nitrogen to herbivorous insects, relatively little has been reported on the influence of potassium. However, van Emden et al. (1969) reported that most workers found that low levels of potassium supplied to plants were favourable to , whilst high levels reduced them. Schoene (1941) suggested that increased multiplication of the mealybug Pseudococcus comstocki (Kuwana) on apple, was associated with low plant potassium. Ahenkorah et al. (1977) pointed out that as shade is reduced over cocoa, phosphate and potash fertilizers may be needed for sustained improvements in yields. In contrast, application of nitrogenous fertilizer reduced yields. Thus the recommended fertilizer treatments might contribute to lower incidences of both Planococcus citri and M. tessmanni. However, whereas P. citri apparently responded to potassium only on heavily shaded trees, the response of M. tessmanni was on unshaded trees in Block I and as an interaction with phosphorus e1sewhere. The damage caused to mature trees by the psyllid is not usually important, aIthough de Mire (1975) found that it was consistently more severe on unshaded trees than on shaded ones. The results of the present study (Table IV) show that psyllid numbers were higher on unshaded trees, but no damage was observed. The interactions between potassium and heavy shade on P. citri here (Table I1I) and in Trinidad (Table V) may be attributable to under-utilisation of potassium by heavily shaded trees so creating sap concentrations related to the rates at which potash was applied. The !!reater demand for potassium among lightly shaded and unshaded trees (Murray, 1953) seemingly resulted in concentrations too low to be detrimental toPo citri, even at the highest rates of application. If that is the case, then the 89'6 kg/ha of potash (approximately equivalent to treatment level 1 (Table 1» proposed by Ahenkorah et al. (1977), for mature partially deshaded Upper-Amazon cocoa, would be unlikely to have a noticeable impact on mealybug den sities. (Note: in what is presumably a misprint, Ahenkorah et al. (1977) state 89'6 kg/acre whereas their other recommended rates were in kg/ha.) The preponderance of many species on more insolated trees (Table IV), may also reflect potash levels. Murray (1953) found that as shade over coco a was reduced, sO too were the proportions of both potassium and nitrogen in the leaf tissues. At the same time, he iIIustrated results obtained by T. Tanada who showed similar reductions for both potassium and nitrogen in the leaves of coffee as shade wa~ reduced. De F1uiter (1937) recommended that, on coffee, P. citri could be controlled by increasing overhead shade. If the abundance of P. citri is governed principally by the levels of plant nitrogen, then a decline would be expected as shade was reduced, whereas the opposite occurred. In contrast to the present results, and to de Fluiter's on coffee. Fennah (1959) recorded the highest numbers of P. citri on heavily shaded plots, while unshaded trees were almost uninfested (Table Vb). He pointed out that the virtual absence of mealybugs on the unshaded trees could be explained by assuming that it was due to a deficiency of soluble nitrogen in the lea ves resulting from a high rate of protein synthesis, which also limited growth and yield ofthe trees at that time (Murray, 1955). The nitrogen deficiency was somewhat ameliorated by the fertilizer, and as in the present study (Table IV), Fennah found that mealybug numbers were higher on treated trees than on untreated ones. Analysis of leaf samples supported the conclusion that total nitrogen was less on unshaded trees (Murray, 1953; Fennah, 1954) than on shaded ones. However, Murray (1955) found that the response of cocoa to nitrogen ended once the canopies of adjacent trees met and they became self-shading. That event occurred at about four years after planting. Murray noted that a c10sed nitrogen cycle then ensued, and subsequently it was the availability of phosphGrus and potassium which limited 172 e A. M. CAMPBELL

yields. Murray's findings, therefore, agree with those of Akenkorah & Akrofi (1977) at Tafo. The apparent contradiction between Fennah 's results and those from the present study (Table IV) on the effects of shade on P. citri seems largely attributable to the variable amounts of mobilisable nitrogen in the sap associated with the different stages of growth of the trees in the two studies. Fennah (1959) reported that the trees werc re-surveyed for mealybugs 3'5 years after they were planted, and no differences were found from the previous year. However, no further surveys were reported after the canopy c1osed, and when trecs no longer responded to nitrogen applications. Bigger (19H la) sampled the insect fauna including the Homoptera on the southernmost part of the Amelonado shade-and-fertilizer experiment shortly before the trees were grubbed up in 1972. His plots were c10se to Block IV of the present experiment. Although the two shade- and-fertilizer experiments were planted on adjacent sites, and were contemporary for 13 years, the opportunity was lost for comparing the faunas simultaneously. U ndoubtedly a contributory cause for that omission was the erstwhile lack of suitable sampling techniques for Homoptera on mature coco a trees (Campbell, 19H3). Differences between the two shade-and-fertilizer expcriments included the species of shade tree and the degraded foliar canopy of unshaded plots in the Amelonado experiment. In addition, C. c/ariventris and Oecophylla /onginoda (Latreille), the two species of dominant ants, were unevenly distributed among the five plots, each of nine trees, that Bigger destructiveiy sampled for Homoptera. C. c/ariventris was present on one shaded plot and was co-dominant with O. longinoda on the half-shaded plot, whereas O. /onginoda was dominant on the two unshaded plots and the other shaded plot. Planococcoides njalensis constituted 90 %, Planococcus citri 7 % and Phenacoccus hargreavesi 3 % of the mealybugs on the Amelonado trees, compared with 7, 80 and 10 %, respectively, for these species on the Upper-Amazon trees sampled two years later (Table 11). These figures suggest that the change of cocoa variety from Amelonado to Upper-Amazon types was at least partially responsible for the changed composition of the mealybug fauna at Tafo (Bigger, 1981b). Indeed, Firempong (1982) found that Amelonado was a more suitable host for Planococcoides njalensis than sorne Upper-Amazon hybnds. However, {he figures from the Amelonado shade-and-fertilizer experiment are deceptive, as 89 % of the individuals of P. njalensis were on the two unshaded plots where O. longinoda was the dominant ant and C. castanea F. Smith a subdominant ant. C. castanea preferentially tends mealybugs for honeydew, and P. njalensis is the species of mealybug most commonly tended on cocoa. lt is equally possible, therefore, that it was the presence of a particular species of ant that caused the greater abundance of P. njalensis on the unshaded trees of the Amelonado experiment. Stictococcus sjostedti, which is preferentially tended both by C. c/ariventris and O. /onginoda, was the most abundant species of Homoptera encountered by Bigger (1981a). He reported that S. sjostedti and the species of Gascardia showed a gradient from shaded to unshaded plots. Therefore, for S. sjostedti, the trend was opposite to that found on the Upper-Amazon cocoa in the present study (Table IV), whereas for the species of Gascardia the trends of abundance were the same in both investigations. Bigger (1981a) sampled sorne other species of Homoptera using a pyrethrum knockdown technique. He found that, as in the present study, T. aurantii was uninfluenced by overhead shade. Surprisingly, considering its abundance here (Table 11), just five individuals of M. tessmanni were counted. Gibbs & Leston (1970) similarly used pyrethrum knockdown. With the exception of two samples, they too obtained quite low counts of M. tessmanni, particularly in the wet dull season, when compared with numbers in the present study (Table 11) and earlier (Campbell, 1983). The pyrethrum knockdown technique may be less suitable for assessing Homoptera if they are killed with their mouthparts inserted in the plant tissues. The species recorded in the present study seldom cause direct damage to mature cocoa trees (Entwistle, 1972), and they are only of economic significance in relation to cocoa swollen-shoot disease. AII the mealybug species recorded here are vectors of the causal virus (Roivainen, 1976), and sorne other species of Homoptera are important because they influence the composition of the ant mosaic (Bigger, 1981a). Ca,mpbell (1983) found that by examining on average one branch sample per tree, 74 'y., of those trees actually infested with HOMOPTERA OF AMAZON COCOA 173

mealybugs were detected. In the present study, 96, 74 and 62 'M, of samples from unshaded, lightly shaded, and heavily shaded trees, respectively, were found to be infested with one or more species of mealybug. Correcting these figures for the one in four trees infested with mealybugs but not identified as such from sampling, gives estimated true percentages of trees infested of 100, 99 and 83. respectively. Neither of the dominant species of ants tended mealybugs other than casually, and consequently the densities of mealybugs were relatively low when compared with those associated with mealybug-tending ant species (Campbell, 1983). As yet, there is no evidence that the relatively low densities of mealybugs usually occurring on cocoa trees limit the spread of swollen shoot virus. The present results suggest that further reductions in numbers of shade trees could increase population densities and the number of trees infested with mealybugs, and favour the generally more mobile species (Bigger, 1981b). However, the increased abundance of potential vectors following reduced shade may have little impact on the rate of spread of swollen shoot disease, as even under the heaviest shade 4 of every 5 trees are probably already infested with mealybugs. The interlocking crowns of adjacent trees create abundant natural bridges between the trees which facilitate the dispersal of perambulating mealybugs. In addition, the momentum for de-shading is probably irreversible in Ghana.

Acknowledgements I am grateful to K. J. Martin for statistical advice, and to Messrs Adjani, Ampomah, Bonku and Gyamfi for technical assistance. I thank Professor A. F. Posnette, Dr J. M. Thresh and K. J. Martin for critically reading the script. This paper is published with the permission of the Director of the Cocoa Research Institute of Ghana. The work was sponsored by the British Overseas Development Administration.

References AHPIKOR"-H v & -'\K RO!" , 0. S. ( ! 95:::). Amazon cacao (Theoóroma cacao L) shade and manurial experiment (K2-0 1) at the Coco a Research Institute of Ghana. l. First five years.-Agron. J. 60, 591-594. AHENKORAH, Y. & AKROFI, G. S. (1977). Amazon cocoa (Theobroma cacao L) shade and manurial experiment (K2-01) at Cocoa Research Institute of Ghana. III: cumulative yield analysis.­ pp. 291-301 in Atanda, O. A., Olaniran, Y. A. O., Omotosho, T. 1., Youdeowei, A., Adelusi, l. O. & Daramola, A. M. (Eds.). Proceedings of the 5th International Cocoa Research Conference, Cocoa Research Institute of Nigeria, Ibadan, Nigeria ¡-9th September, 1975.-642 pp. Ibadan, Nigeria. AHENKORAH, Y, HALM, B. J. & AKROFI, G. S. (1977). Sorne agronomic factors affecting coco a rehabilitation in Ghana.-pp. 199-203 in Atanda, O. A., Olaniran, Y. A. O ., Omotosho, T. l., Youdeowei, A., Adelusi, 1. O. & Daramola, A. M. (Eds.). Proceedings of the 5th International Cocoa Research Conference, Cocoa Research Institute of Nigeria, Ibadan, Nigeria 1-9th September, 1975.-642 pp. Ibadan, Nigeria. ALVIM, P. DE T. (1977). Ecological and physiological determinants of cacao yield.-pp. 25-38 in Atanda, O. A., Olaniran, Y. A. O.,Omotosho, T. I., Youdeowei, A., Adelusi, l. O. & Daramola, A. M. (Eds.). Proceedings of the 5th International Coco a Research Conference, Coco a Research Institute of Nigeria, Ibadan, Nigeria 1-9th September, 1975-642 pp. Ibadan, Nigeria. -BIGGER, M. (1973). Association analysis.-Rep. Cocoa Res. Inst.. Ghana 1971-72, 120--125. BIGGER, M. (1976). Virus outbreak survey.-Rep. Cocoa Res. Inst.. Ghana 1973-74. 101-107. ,BIGGER, M. (191\ la). Observations on the insect fauna of shaded and unshaded Amelonado cocoa.-Bull. enr. Res. 71, 107-119. BIGGER, M. (1981b). The relative abundance of the mealybug vectors (: Coccidae and Pseudococcidae) of cocoa swollen shoot disease in Ghana.-Bu/l. enr. Res. 71, 435-448. CAMPBELL, e A. M. (1983). The assessment of mealybugs (Pseudococcidae) and other Homoptera on mature cocoa trees in Ghana.-Bulf. ene. Res. 73, 137-151. DE FWITER, H. J. (1937). Waarnemingen omtrent de witte luisbestrijding.-Bergculrures 11,995-999. DE MIRE, P. B. (1975). Comportement de cultivars de cacaoyers a I'égard des attaques d'homop­ teres.-pp. 176-180 in Kumar, R. (Ed. ). Proceedings of the 4th Conference of West African 174 C. A. M. CAMPBELL

Cocoa Entomologists. Zoology Department, University of Ghana, Legon, Ghana, 9-13th December. 1974-202 pp. Legon. Ghana. Zool. Dep .. Univ. Ghana. ENTWISTLE. P. F. (1972). Pests of cocoa-779 pp. London, Longman. FENNAH. R. G. (1954). Nitrogen partition in the normal cacao leaf.-Rep. Cacao Res. 1953.45-52. FENNAH, R. G. (1959). Nutritional factors associated with the development of mealybugs on cacao.-Rep. Cacao Res. 1957-1958. 18-28. FIREMPONG. S. (19H2). The performance of P/anococcoides nja/ensis (Homoptera: Pseudococcidae) on some cocoa cultivars.-Ann. appl. Biol. 100 (Supplement: Tests of agrochemicals and cultivars. 3). 100-10 1. GIBoS. D. G. & LESTON. D. (1970). Insect phenology in a forest cocoa-farm locality in West Africa.-J. app/. Eco/. 7.519-54H. HUTCHEON, W. V. (1977). Physiological aspects of cocoa agronomy.-pp. 39-48 in Atanda. O. A.. Olaniran. Y. A. O .• Omotosho. T. l.. Youdeowei. A .• Adelusi.1. O . & Daramola. A. M. (Eds.). Proceedings 01' the 5th International Cocoa Research Conference. Cocoa Research Institute of Nigeria. Ibadan. Nigeria 1-9th September, 1975-642 pp. Ibadan, Nigeria. JOHNSON, C. G. (1962). The ecological approach to cocoa disease and heaJth.-pp. 348--352 in WiIIs. J. B. (Ed.). Agriculture and land use in Ghana.-504 pp. Oxford, Univ. Press. LEGG, J. T. (1979). The campaign to control the spread of cocoa swollen shoot virus in Ghana.­ pp. 285-293 in Ebbels. D. L. & King, J. E. (Eds. ). Plant health. The scientific basis for administrative control of plant diseases and pests.-322 pp. Oxford, Blackwell. MATTSON. W. J. (1980). Herbivory in relation to plant nitrogen content.-Ann. Rev. Eco/. & Syst. 11, 119-161. McNEILL. S. & SOUTHWOOD. T. R. E. (1978). The role of nitrogen in the development of insect/plant relationships.-pp. 77-98 in Harborne. J. B. (Ed.). Biochemical aspects of plant and coevolution.-435 pp. London, Academic Press. MURRAY. D. B. (1953). A shade and fertiliser experiment with cacao-progress report-con­ tinued.-Rep. Cacao Res. 1952. 11-21. MURRAY. D. B. (1955). A shade and fertiliserexperiment with cacao. IV.-Rep. . Cacao Res. 1954. 32-36. PRESTIOGE, R. A. (1982). The influence of nitrogenous fertilizer on the grassland (Homoptera).-J. app/. Ecol. 19. "35-74Q .ROIVAINEN. O. (1976). Transmission of cocoa viruses by mealybugs (Homoptera: Pseudococcidae).­ J. scient. agric. Soco Fin/and 48. 203-304. ,SCHOENE. W. J. (1941). Plant food and mealybug injury.-J. econ. Ent. 34.271-274. VAN EMDEN. H. F .• EASTOP. V. F .• HUÚHES. R. D. & WAY. M. J. (1969). The ecology of Myzus persicae. -A. Rev. En/. 14. 197-270.

(Received 6 May /983)

(Z') Commonwealth Agricultural Bureaux. 1984