African Crop Science Journal African Crop Science Society ISSN: 1021-9730 EISSN: 2072-6589 Vol. 6, Num. 2, 1998, pp. 205-214

African Crop Science Journal, Vol. 6. No. 2, pp. 205-214, 1998 - FORUM -

P. L. WOOMER, M. BEKUNDA¹ and D. BWAMIKI²

Department of Soil Science, University of Nairobi, P. O. Box 29053, Nairobi, Kenya
¹Makerere University Agricultural Research Institute Kabanyolo, P. O. Box 7062, Kampala, Uganda
²National Agricultural Research Organisation, P. O. Box 7065, Kampala, Uganda

(Received 12 September, 1997; accepted 30 March, 1998)

Code Number:CS98023
Sizes of Files:
      Text: 38K
      Graphics: Line drawings and tables (gif) - 39K

ABSTRACT

A major concern to food security in Uganda is banana yield reduction, a condition locally known as 'matoke decline'. Computer simulation is one means of understanding the complex interactions between plants, organic matter dynamics and soil but require that much be quantified within systems before models may be applied. We have initialised the CENTURY model to simulate a banana re-establishment experiment in Mukono District, Uganda. The initialisation required assumptions concerning assimilate allocation patterns between shoots, roots and bunches. The complex pattern of grass fallow, burning, maize, chemical fertilisation and banana receiving different managements could be simulated using schedule routines of the model, but some adjustments were necessary to allow for perennial growth. Two separate harvest files were constructed to account for harvest of bunches and the removal of basal offshoots. At the first harvest (18 months after planting) the simulation of treatments receiving no inputs and that having 10 t napier grass (Pennisetum purpureum) differed by 5.0 and 1.0 t ha^-1 in total aboveground biomass (dw) and fresh bunch yield. After six years of harvest, average annual yields were 5.1, 6.1 and 7.6 t ha^-1 in the control, livestock manure (10 t ha^-1 y^-1) and napier grass treatments, respectively. Simulated soil organic carbon content increased from 2.2% to 3.8% in the treatment receiving napier grass and harvest residues but increased only slightly in the complete control (2.5%). Model outputs suggest that a near 'steady-state' in banana biomass is achieved due to periodic senescence of mature pseudostems, bunch harvest and offshoots. Model outputs suggest that external organic inputs of 10 t ha^-1 y^-1 influence banana productivity to a greater extent than the retention of banana residues.

Key Words: Banana management, matoke decline, napier grass, organic inputs, soil organic carbon, Uganda

La réduction de rendement du bananier, situation localement appelée "le declin de Matoke" constitue une majeure préoccupation pour la securité alimentaire en Uganda. La simulation informatique est l'un des moyens permettant de comprendre les interactions complèxes entre les plantes, les dynamiques de matière organique et de sol, mais elle implique que des systèmes soient au préalable quantifiés avant que des modèles informatiques soient appliqués. Un modèle du siècle a été initié pour simuler le rétablissement d'une bananeraie dans une expérience conduite dans le district de Mukono en Uganda. Ceci a nécessité le développement des suppositions sur les voies d'allocation d'assimilats entre les pousses, les racines et les regimes de banane. Les successions complèxes de jachère herbeuse, brûlis, maòs, fertilisation chimique et de bananier recevant différents traitements peuvent Étre simulées en utilisant les routines programmées dans le modèle, mais quelques ajustements sont nécessaires pour favoriser une croissance pérenne. Deux dossiers distincts de récolte ont été montés pour noter la récolte des regimes et l'élimination de pousses de base. A la première récolte (18 mois après la plantation), la simulation de traitements sans intrants et ceux ayant reçus 10 tonnes d'herbes de napier (Pennisetum purpureum) a montré une différence allant de 5, 0 et 1, 0 ha^-1 de biomasse aérienne totale (dw) et de rendement en regimes. Après six années de récolte, les rendements annuels moyens ont été respectivement de 5, 1; 6, 1 et 7, 6 t ha^-1 pour le témoin, les déchets d'animaux (10 t ha^-1an ^-1) et le paillis de napier. La teneur simulée en carbone organic du sol a augmenté de 2,2% à 3,8% dans le paillis de napier et les résidus de récolte alors que l'augmentation n'a été que de 2,5% pour le témoin complet. Les résultats du modèle montrent que une situation stable de la biomasse de la bananeraie a été réalisée par la senescence périodique de pseudo-troncs mûrs, la récolte de regimes et de pousses. Le modèle informatique a également montré que l'addition d'un intrant organique extérieur de 10 t ha^-1 an^-1 influence plus la production de la bananeraie par rapport à l'usage seulement des résidues de la banane.

Mots Clés: Gestion de la bananeraie, declin de matoke, paillis de napier, intrants organiques, carbone organique du sol, Uganda

The CENTURY model is a plant/soil environmental simulation model developed by Parton et al. (1987). The model was first applied to the grasslands and crops of the Great Plains of North America, but following the inclusion of additional routines, it has accurately simulated plant growth and organic matter dynamics in many temperate (Parton et al., 1987: Paustian et al., 1992) and tropical (Parton et al., 1989, 1994; Sanford et al., 1991; Woomer, 1993) environments. None of these reports, however, include the adaption of the model to simulate banana as a crop.

Uganda is the world's largest producer and consumer of bananas (FAO, 1995a). Per capita banana consumption was estimated to be 220-460 kg y^-1 (Hartman, 1989) where it is used as a dessert fruit, cooked as a staple starch or brewed as beverage (Stover and Simmonds, 1987). An ominous trend results from nutrient depletion of banana-based smallholdings. Approximately 2000 metric tonnes of fertilizer was consumed during 1994, all of which was imported (FAO, 1995b). Per capita fertilizer use is less than 0.1 kg y^-1 (calculated from FAO, 1995a, 1995b). One manifestation of nutrient depletion is matoke decline, where banana yields have declined to less than 10 kg bunch^-1 or 0.5 t ha^-1 y^-1 (Zake et al., 1994). Farmers who wish to re-establish banana in nutrient depleted soil without access to fertilizers must rely on external organic inputs. Bekunda and Woomer (1996) reported that two such inputs available to farmers are livestock manure and napier grass (Pennisetum purpureum).

Before a plant/soil simulation model can assist in our understanding of such complex phenomenon as matoke decline, it must first be demonstrated that the model is capable of initialising plant growth and current management practices. Six different simulations with the CENTURY Model were conducted based upon a banana re-establishment experiment being conducted in a degraded soil in Mukono, Uganda, as a test of that model's applicability to banana-based cropping systems.

The most recent version of Century version 4.0 (Metherell et al., 1993) contains an event scheduler which provides for rotational fallows, burning, removal of crop residues, addition of organic inputs and harvest/removal of both reproductive and vegetative structures which resemble the various land management strategies practised in banana cultivation. To initialise a simulation in CENTURY requires that three separate files be developed, CROP.100, SITE.100 and EVENT100.SCH files. A CROP.100 file describes the growth rate, allocation patterns and senescence of the plant. A SITE.100 file contains information on the climate, soil and many regulators of soil biological processes. An EVENT100.SCH file describes land management of a site and provides for many options such as tillage, fertilisation, harvest, burning and crop rotation. We based our simulation of banana growth on the treatments and land management at the Mukono Banana Experiment but relied on climate and soil data at the nearby, and better described Kabanyolo Farm (Yost and Eswaran 1990).

Banana as a CROP.100 file option. The BANANA option within the CROP.100 file was developed through trial and error because many of the data necessary to the construction of a CROP.100 file were not available. The initial approach was to simulate banana as a tree but this met with difficulties because tree sub-routines lack reproductive structures. No coarse wood, fine branches or woody roots exist in banana, a giant perennial herb (Purseglove, 1972), but these are important pools within tree sub-routines of the CENTURY model (Metherell et al., 1993). As a result, this approach was abandoned and no further information is provided on it in this report.

The next approach was to treat banana as a perennial cereal grain and the bunches as harvest products. One advantage of this approach is that leaves and stems are not distinguished within model sub-routines, which is fortunate because the pseudostem of banana is actually a modified leaf structure, while the true stem remains low to the ground and serves as a storage organ (Stover and Simmonds, 1987). Very little empirical data exist on the allocation of recent assimilates between shoots and roots, the proportion of pseudostems to true stems, the proportion of true stems to reproductive structures (bunches) when growth conditions are favourable or the effects of shoot removal (cutting the pseudostem at harvest) on root turnover were available, yet these are crucial parameters within the CROP.100 file. As a result, it was necessary to make many assumptions when developing the file, run simulations, observe the model outputs of above-ground, below-ground and bunch biomass, adjust assumptions and rerun the model in an iterative manner. Trial-and-error may be too kind a term for this process, perhaps it is better compared to learning to play a very difficult computer game without benefit of the instruction manual. Earliest simulations resulted in excess root biomass or die-off of above-ground biomass, but eventually a realistic-appearing simulation was generated. Some of the key parameters in the BANANA option of the CROP.100 file used in later simulations are presented in Table 1. When plotting model outputs, all plant tissues were assumed to contain 45% carbon and banana bunches to contain 75% moisture.

TABLE 1. Selected CROP.100 parameters, their definitions and values within the BANANA option of the CENTURY model

Mukono as a SITE.100 file. Construction of the MUKONO.100 file required much less innovation but was frustrated by a lack of long-term climate records for the site and detailed soil measurements of the original degraded grass fallow prior to land clearing. Ten year averages of monthly precipi-tation and maximum and minimum monthly temperature were available for the Kabanyolo Experimental farm, which is similar to Mukono in terms of elevation and geographical position within the northern Lake Victoria Basin as well as for other longer-term sites near Kampala (see Hargreaves and Samani, 1986). Climate data were entered as monthly means without standard deviations, which then disallows the stochastic weather generation option available within the CENTURY Model. Monthly climate data entered into the MUKONO.100 file appear in Table 2.

Similar approximation was necessary with the input of soil data. The Mukono Farm site was originally selected as being representative of the severely degraded matoke decline areas situated to the east of Kampala rather than for the availability of data on soil physical and chemical properties. A detailed description and analysis is available for an oxisol at the Kabanyolo Experimental Farm (Yost and Eswaran 1990) and the data for the 0-17 cm soil layer (texture, pH, bulk density, water holding capacity, soil organic C, mineral N) were entered into the MUKONO.100 file. A lack of information on soil organic matter fractions required that total soil C be speculatively divided among three SOM pools, with a majority of the soil C being assigned to the PASSIVE pool (see Parton et al., 1987). A lack of information on phosphorus pools and dynamics required that P sub-routines were not included within the simulations. Selected soils data incorporated into the MUKONO.100 file are presented in Table 3.

TABLE 2. Estimates of monthly precipitation and mean monthly maximum and minimum temperatures of the Mukono Farm

TABLE 3. Selected soil parameters and values entered to initialise the CENTURY Model for banana production

The Mukono experiment as an EVENT100.SCH file. As was previously stated, the Mukono site was selected on the basis of its degraded condition and land management required that the banana plants be provided with sufficient inputs to become established but not in excess as to interfere with subsequent treatment effects resulting from organic matter transfers. A schedule file was developed with the following features. The simulation starts in January 1994 with a grass fallow resembling C-4 tropical grasses (see Metherall et al., 1993), the year ends with grass senescence in November (due to herbicide application) and then is burned in December. During 1995 a hybrid maize crop is grown without external inputs (representing the initial maize uniformity trial) between March and July, harvested, and then the soil is hand tilled during September to prepare for planting banana.

Bananas are planted in 1996 with the addition of 34 kg N and P ha^-1. Hand weeding is conducted at 3 month intervals. Organic inputs are first applied in September 1996, and every September thereafter (no inputs, 10 t manure ha^-1 y^-1, 10 t napier grass ha^-1 y^-1). Crop residues (banana pseudostems and leaves) either remain as surface litter or are removed. The experimental design (3 inputs by two residue managements) required that six different schedule files be developed. The first harvest occurs in May 1997 and then every six months thereafter. The entire bunch is removed (considered to be grain in the model routines) and 20% of vegetative shoots (the bunch-bearing pseudostem) dies as a result of harvest, which in turn is programmed to accelerate root senescence. A second harvest file was designed to replicate the practice of 'sucker (basal offshoot) removal' starting in September 1996 and conducted at six month intervals thereafter. The 'sucker harvest' was not intended to disturb reproductive structures but rather removed 5% of the shoots and 5% of the roots with an additional 3% of the roots dying afterwards. EVENT.100 normally designates the crop harvest month to be the last month of growth. This feature was cancelled to allow the crop to perennialise. The schedule file ends the simulation in December 2002.

The simulated effects of two land management strategies at the banana re-establishment experiment at Mukono are presented in Figure 1. The simulation starts with the growth of a natural grass fallow, obtaining a biomass of 3.4 t ha^-1 after 11 months, which is then treated with herbicide and burned at the end of the year (1994). The following year (1995), a maize crop is planted and produced 1.4 t ha^-1 and then land is prepared for bananas. Bananas are planted in January 1996 and produce 12.0 t ha^-1 of above-ground biomass after 9 months. Up to this point the simulations for the different land managements are identical because treatment effects have not yet been imposed. In September 1996, selected treatments receive 10 t organic inputs ha^-1 and the first bunches are ready for harvest eight months afterward. The pseudostems bearing harvested bunches and their leaves are either removed from the banana plots (residues removed) or chopped into large pieces and uniformly distributed over the soil surface (residues retained). At the first harvest the two treatments compared in Figure 1 differ by 5.0 and 1.0 t ha^-1 in total above-ground biomass (dw) and fresh bunch yield. Model outputs suggest that a near 'steady-state' in banana biomass is achieved due to periodic harvest and vegetation removal.

Figure 1. Century model simulation for two contrasting treatments for the Mukono Farm banana experiment, no organic imputs, banana residues removed (a) and napier grass applied at 10 t ha^-1y^-1 (b)

The simulated above-ground biomass of the six treatments is presented in Figure 2. The simulated addition of organic inputs at 10 t ha^-1 had greater effect on productivity than the retention or removal of banana residues. Note that the biomass of those treatment simulations receiving greatest inputs appear to increase with time, while the complete control is stable or decreasing.

Figure 2. Century model simulation of banana receiving three levels of external inputs with or without harvest residue applied.

Figure 3. Simulated dynamics of soil organic matter of Mukono Farm banana experiment by Century model.

Simulated changes in soil organic carbon resulting from the treatments (Fig. 3) suggest that the retention of harvest residues has a greater effect on the maintenance of soil organic matter than on above-ground productivity (Fig. 2). All treatments contained 2.25% soil carbon when bananas were planted in January 1996. Simulated soil C levels increased by 0.3 and 1.54% after seven years in the complete control and napier grass with residues treatments, respectively. Soil organic carbon in all treatments increases with time, although the complete control increases least. The simulated aggradation of soil carbon may be an artifact of excess root turnover (data not shown) because we are uncertain of the 'protection' from root turnover by photosynthate reserves in the stem and rather we relied upon published values for annual crops to run the simulation (Metherell et al., 1993).

Averages of simulated fresh bunch yields (t ha^-1) are presented in Figure 4. These data result from combining the two annual harvests and then averaging those harvests across the six years of simulated banana production. Note the 3.0 t ha^-1 difference between the least and most productive treatments, and that organic inputs continue to influence yields to a greater extent than the removal or retention of harvest residues.

Figure 4. Century model simulation of average banana yield over six years of production with or without banana harvest residues.

The CENTURY model contains sufficient flexibility to simulate the complex patterns of crop growth and land management involved in banana production but the accuracy of model outputs were not tested, owing in large part to a paucity of 'systems-level' information on banana in Uganda. Development of a banana option in CENTURY CROP.100 routines was also speculative, in large part because of the extreme difficulty in collecting whole-plant and plant allocation data for banana due to its massive size compared to annual field crops. Nonetheless, we have produced first-generation simulations of banana which resemble reports from the literature. Average banana yield in Uganda are reported at 5.8 t ha^-1 y^-1 (Rubaihayo, 1992), similar to the overall mean of the simulated complete control (5.6 t ha^-1 y^-1, Fig. 4). Bekunda and Woomer (1996) reported that 35% of banana farmers in Uganda's Lake Victoria Basin applied livestock manure to bananas and their estimated yields were 6.9 t ha^-1 y^-1, compared to 6.8 t ha^-1 y-1 simulated for the treatments receiving manure in this study. The simulations establish a near steady state in banana biomass through inclusion of periodic banana harvest and sucker removal, similar to that observed in well-managed plantations where each mat is maintained to consist of three or four pseudostems at different stages of development for many years.

Some features of the CENTURY model restrict the application of the CENTURY Model to banana-based cropping systems. The model allows for crop rotation but not intercropping, other than for combinations of trees and crops. We first sought to enter banana into model routines as a tree, allowing for other intercrops to be simultaneously cultivated, but this attempt failed because of inherent incompatibilities between the anatomy of banana and the compartmentalisation of tree sub-routines. Bekunda and Woomer (1996) report that 69% of banana farmers in Lake Basin Uganda practice intercropping, a practice which cannot be replicated by the modelling approaches employed in this paper. An annoyance, rather than obstacle, is the model's harvest routine feature which signals an end to crop growth at harvest, which is inappropriate for a perennial crop and must be corrected within the schedule file for each harvest. Furthermore, it must be stressed that the results of this study offer speculation that the CENTURY Model is capable of simulating banana production practised in Uganda rather than confirmation, because model outputs are not being comprehensively tested with an independent data set. One feature that was not considered within the simulation was the occurrence of black sigatoka disease, which severely defoliates banana (Stover and Simmonds, 1987) and appeared more severe in treatments receiving no external inputs.

It is our intent to continue developing the application of the CENTURY Model to banana production in Uganda and to test the model outputs with data collected from the Mukono Farm. We are presently collecting total mat biomass and components of yield through destructive harvest at two month intervals, and these data will offer opportunity to improve the CROP.100 files. Collection of additional soils information from the Mukono Farm will allow for a refined SITE.100 file, including the initialisation of phosphorus sub-routines. Lekasi (1998) reported that phosphorus was the nutrient most limiting to cabbage productivity at the Mukono Farm. The accuracy of the Century model phosphorus routines in highly weathered tropical soils and the ability of the model to account for many of the factors which limit productivity in those soils was challenged by Gijsman et al. (1996) and our future contributions may assist in resolving this controversy. Finally, it is important that we refine CENTURY model application to banana-based systems and run longer-term simulations as a means of pre-selecting a wide assortment of technically-feasible management interventions intended to counter the matoke decline syndrome.

The authors thank Bill Parton, coauthor of the CENTURY model, for providing useful advise concerning this study. The first author was hosted by the Department of Soil Science, Makerere University while model simulations were run and interpreted. The authors gratefully acknowledge the Forum for Agricultural Resource Husbandry of The Rockefeller Foundation for funding this research.

Bekunda, M.A. and Woomer, P.L. 1996. Organic resource management in banana-based cropping systems of the Lake Victoria Basin, Uganda. Agriculture Ecosystems and Environment 59:171-180.

Food and Agriculture Organization of the United Nations (FAO). 1995a. FAO Yearbook: Production. Volume 48. FAO, Rome. 243 pp.

Food and Agriculture Organization of the United Nations (FAO). 1995b. FAO Yearbook: Fertilizer. Volume 44. p. 119. FAO, Rome.

Gijsman, A.J., Oberson, A., Tiessen, H. and Friesen, D.K. 1996. Limited applicability of the CENTURY model to highly weathered tropical soils. Agronomy Journal 88:894-903.

Hargreaves, G.H. and Samani, Z.A. 1986. World Water for Agriculture. International Irrigation Center, Utah State University. 617 pp.

Hartman, E. 1989. Five Year Food Plan, 1989-1994, and Recommendations for Streng-thening Research and Extension Linkages. USAID/MFAD Project, Makerere University, Kampala.

Lekasi, J.K. 1998. Biological Management of Soil Fertility in Banana-based Cropping Systems of Uganda. MSc. Thesis, Makerere University, Kampala. 86 pp.

Metherell, A.K., Harding, L.A., Cole, C.V. and Parton, W.J. 1993. CENTURY: Soil Organic Matter Model Environment. Colorado State University.

Parton, W.J., Schimel, D.S., Cole, C.V. and Ojima, D.S. 1987. Analysis of factors controlling soil organic matter levels in Great Plains grasslands. Soil Science Society of America Journal 51:1173-1179.

Parton, W.J., Sanford, R.L., Sanchez, P.A. and Stewart, J.W.B. 1989. Modeling soil organic matter dynamics in tropical soils. In: Dynamics of Soil Organic Matter in Tropical Ecosystems. Coleman, D.C. Oades, J.M. and Uehara, G. (Eds.), pp. 153-171. NifTAL Project, University of Hawaii, Honolulu.

Parton, W.J., Woomer, P.L. and Martin, A. 1994. Modelling soil organic matter dynamics and plant productivity in tropical ecosystems. In: The Biological Management of Tropical Soil Fertility. Woomer, P.L. and Swift, M.J. (Eds.), pp. 171-188. John Wiley and Sons, Chichester, U.K.

Paustian, K., Parton, W. and Persson, J. 1992. Modelling soil organic matter in organic-amended and nitrogen-fertilized long-term plots. Soil Science Society of America Journal 56:476-488.

Purseglove, J.W. 1972. Tropical Crops: Monocotyledons. John Wiley and Sons, New York. 607 pp.

Rubaihayo, P.R, 1992. Banana Based Cropping Systems Research. Bulletin No. 3. Makerere University, Kampala, Uganda. pp 2.

Sanford, R.L. Jr., Parton, W.J., Ojima, D.S. and Lodge, D.J. 1991. Hurricane effects on soil organic matter dynamics and forest production in the Luquillo Experimental Forest, Puerto Rico: Results of simulation modelling. Biotropica 23:364-372.

Stover, R.H. and Simmonds, N.W. 1987. Bananas. 3rd Edition. Longman, London. 468 pp.

Woomer, P.L. 1993. Modelling soil organic matter dynamics in tropical ecosystems: Model adoption, uses and limitations. In: Soil Organic Matter Dynamics and Sustainability of Tropical Agriculture. Mulongoy, K. and Merckx, R. (Eds.), pp. 279-294. John Wiley and Sons, Chichester, U.K.

Yost, D. and Eswaran, H. 1990. Major Land Resource Areas of Uganda. Soil Management Report Services, USAID, Washington. 218 pp.

Zake, Y.K., Nkwine, C., Sessanga, S., Bwamiki, D.P., Tumuhaire, J.K., Najjuma, C. and Kagole, H. 1994. Identification of suitable soils and soil mangement for banana productivity. In: Banana Based Cropping Systems Research. Final report. Department of Crop Science, Makerere University, Kampala, Uganda. pp. 25-30.

Copyright 1998, African Crop Science Society

The following images related to this document are available:

Line drawing images