Search

An analysis of the impact of simulation modelling in three diverse crop-livestock improvement projects in Agricultural Research for Development (AR4D) reveals benefits across a range of aspects including identification of objectives, design and implementation of experimental programs, effectiveness of participatory research with smallholder farmers, implementation of system change and scaling-out of results. In planning change, farmers must consider complex interactions within both biophysical and socioeconomic aspects of their crop and animal production activities. For this, whole-farm models that include household models of food, workload and financial requirements have the most utility and impact. The analysis also proposes improvements in design and implementation of AR4D projects to improve the utility of simulation modelling for securing positive agronomic and livestock outcomes and lasting legacies.

Approximately 15% of global wheat production (40% of the wheat produced by developing countries) occurs in irrigated, low-rainfall regions of the subtropics. Irrigated wheat is uncommon in sub-tropical Australia because cotton is more profitable, but high grain prices in 2008 triggered a record area of irrigated wheat production. Unfortunately, widespread lodging occurred and yields were lower than expected, although the losses directly attributable to lodging were unknown. Crop models have not been previously validated for calculating potential yield and yield gaps of mechanised, lodging-susceptible production systems. This study aimed to determine potential yield, farmer-realised yield, and water use requirements of lodging-prone irrigated wheat in sub-tropical Australia, using field monitoring and the APSIM simulation model. Yield data were obtained in 2008 and 2009 from 16 commercial irrigated wheat fields and multiple treatments from four irrigated experiments, and used to assess APSIM's ability to predict grain yield of irrigated wheat. APSIM predicted biomass production satisfactorily in 2008 but substantially over-predicted grain yield of lodged fields. The mean difference (yield gap) between APSIM-estimated potential yield and farmer-realised yield was 1.0 t ha−1 in non-lodged fields, and 2.7 t ha−1 in lodged fields. The average effect of lodging was therefore a decrease of 1.7 t ha−1; the difference between the yield gap calculated for lodged and non-lodged fields. Yield gaps in lodged fields were highly correlated with soil N content at sowing, with most of the lodged fields having more than 300 kg ha−1 of soil N at sowing to a depth of 90 cm. In 2009, management of commercial fields aimed to avoid lodging by using the canopy management technique of in-season N application. APSIM generally under-predicted biomass production and yield in these fields, suggesting that the N uptake parameters in APSIM may require adjustment. Observed yields from fields of a quick-maturing cultivar that experienced little lodging were simulated accurately when N was assumed to be non-limiting. The simulation study found that the potential yield of irrigated spring wheat in sub-tropical Australia is currently between 8 and 9 t ha−1, and average growing season evapotranspiration of such crops was approximately 500–550 mm. Potential yield of furrow-irrigated fields with one or two-metre wide beds were 16% and 8% lower, due to the area occupied by the 60 cm wide unsown furrows between irrigation beds. This study demonstrated the value of using validated simulation models for estimating potential yield and yield gap analysis. Yield gap analysis also enabled the identification of agronomic characteristics (high levels of soil residual N) which probably contributed to lodging. Further investigation is required to determine whether alternative varieties or the canopy management technique of in-season N application can increase the potential yield of irrigated spring wheat above the levels reported here.

World population growth, changing diets and limited opportunities to expand agricultural lands will drive agricultural intensification in the decades ahead. Concerns about the reliance of past agricultural intensification on non-renewable resources, about its negative impacts on natural resources both on and off farm and on greenhouse gas emissions, provide an imperative for future agricultural intensification to become ecologically efficient. We define ecological intensification of agriculture (EIA) as: producing more food per unit resource use while minimising the impact of food production on the environment. Achieving it will require increased precision in the use of inputs and reduction in inefficiencies and losses. It will also require a more holistic view of farming, going beyond efficiencies of single inputs into a single field in a single season to consideration of efficiencies of whole systems over decades. This paper explores the ecological intensification issues facing agricultural production in Australia where opportunities for agricultural intensification are centred on more efficient use of limited and unreliable water resources in both dryland and irrigated agriculture. Ecological efficiencies can be achieved by better matching the supply of nutrients to crops’ requirements both temporally and spatially. This has the added benefit of minimising the opportunities for excessive nutrients to impact on soil health (acidity and dryland salinity) and water quality (pollution of groundwater and eutrophication of lakes and rivers). Opportunities for ecologically efficient intensification are also identified through better integration of crop and livestock enterprises on mixed crop–livestock farms. We define nine desirable attributes of an EIA system: (1) increased agricultural production; (2) efficient use of limited resources; (3) minimal impact on global warming; (4) minimal negative on-site impacts; (5) minimal negative off-site impacts; (6) minimal risk and maximum resilience; (7) preservation of biodiversity in agriculture; (8) preservation of biodiversity in nature and; (9) positive social outcomes. We focus on four technologies and production systems emerging in Australian agriculture: climate risk management; precision agriculture; crop–livestock integration and deficit irrigation. For each of these systems we identify how well they are likely to match the nine desirable attributes of an EIA system. While it seems unlikely that any single technology can satisfy all nine desirable attributes, there is hope that in combination emerging and future technologies will progress Australian agriculture towards greater productivity and ecological efficiency.

The prospect that decision support systems (DSS) can help farmers adjust their management to suit seasonal conditions by putting scientific knowledge and rational risk management algorithms at farmers’ fingertips continues to challenge the science and extension community. A number of reviews of agricultural DSS have called for a re-appraisal of the field and for the need to reflect on past mistakes and to learn from social and management theory. The objective of this paper was to investigate whether there is an emerging consensus, among stakeholders in DSS for Australian agriculture, about the lessons learned from past experience with DSS tools. This investigation was conducted in three parts. The first part was a distillation of suggestions for best practice from the relevant literature. The second part was a reflection on what the champions of five current DSS development and delivery efforts in Australia learned from their recent efforts. The third part tested the level of support for the combined findings from the first and second approaches by surveying 23 stakeholders in the research, development, delivery and funding of DSS. The key propositions relating to best practice that were supported by the survey, listed according to the strength of support, were: 1. It is essential to have a plan for delivery of the DSS beyond the initial funding period. 2. DSS need to be embedded in a support network consisting of farmers, consultants and researchers. 3. DSS development requires the commitment of a critical mass of appropriately skilled people. 4. A DSS should aim to educate farmers’ intuition rather than replace it with optimised recommendations. 5. A DSS should enable users to experiment with options that satisfy their needs rather than attempt to present ‘optimised’ solutions. 6. DSS tools stand on the quality and authority of their underlying science and require ongoing improvement, testing and validation. 7. DSS development should not commence unless it is backed by marketing information and a plan for delivery of the DSS beyond the initial funding period. While the DSS stakeholders supported the proposition that it is essential to have a plan for delivery of a DSS beyond the funding period, the majority resisted the notion of DSS development being market-driven and especially commercial delivery of DSS. We argue that since public funding of the delivery of DSS for farmers’ management of climate risk is highly unlikely, reaping the benefits of lessons learned from past efforts will require that DSS stakeholders change their perception of the commercial delivery model or find an alternative way to fund the delivery of DSS beyond the R&D phase.

Over the past 20 years, farming systems modelling has become an accessible tool for developing intervention strategies targeted at smallholder farmers in southern Africa. Applying the Agricultural Productions Systems sIMulator (APSIM) to credibly simulate key soil and crop processes in highly constrained, low yielding maize/legume systems has led to four distinct modes of use: (i) to add value to field experimentation and demonstration; (ii) in direct engagement with farmers; (iii) to explore key system constraints and opportunities with researchers and extension agencies; and (iv) in the generation of information for policy makers, bankers and insurance institutions. Examples of application in each of these modes are presented. Despite being demonstrated as an excellent tool for developing intervention strategies and extension material, the use of simulation is limited by a lack of competent local users. Better co-operation within the simulation community, sharing of climate, soil and crop parameterisation and validation datasets, and focussing of efforts on using models to benefit smallholder farmers are suggested as ways of increasing the use and relevance of simulation. Substantial investment in the training of agriculturalists and the further science development of systems simulation is required to tackle the enormous challenges facing agricultural development in the region.

In recent years, interest in the policy-informing role of research organisations has grown considerably. It has been argued from within the science domain that the management of land, whether at the field, farm or regional scale, can benefit from computer-based systems analysis. This proposition was examined through 26 semi-structured interviews, by conducting a workshop on modelling support in policy decision-making and investigating 11 European-based cases of policy-oriented modelling research. On basis of these materials, we assess approaches towards policy-oriented modelling within the domain of agricultural production and other environmental services of land. First, we define the boundary arrangements for our field of interest, i.e. computer-based systems analysis of agricultural production and ecosystems services. Such ideotypical classification of boundary arrangements makes explicit the institutional space in which system scientists function. This space enables certain activities, and at the same time constrains other initiatives, dependent on who is perceived to initiate and/or control research activities and to formulate objectives, and whether it is considered appropriate to integrate scientific knowledge development with policy work. Five ideotypical boundary arrangements are proposed: ‘Civil mandate’; ‘Trickle out’; ‘Janus face’; ‘Critical participant’; and ‘Knowledge broker’. In a second step, a number of critical leverage points are identified based on the 11 cases and these are addressed in relation to the five boundary arrangements. The term ‘leverage point’ covers means, conditions and methods that foster use of modelling work that can be changed, and thus may be part of a strategy to increase the chances that a model is used. The leverage points were: reputation of research institute and/or scientists; raising and balancing expectations; communication about and investment in the scientific basis of the modelling work; participation in model development; heterogeneous and extensive social network in policy domain; institute mandate that secures availability of ‘stepping stones’. We conceive of two applications of our research for modellers who are interested in the use of their work in the policy sphere. First, the boundary arrangement classification helps to interpret the experiences of others and to assess the relevance of lessons and suggestions for their own context. Secondly, the combination of the boundary arrangement perspective and critical leverage points presents a basis to design an institutional pathway for enhancing impact of modelling research in the policy sphere. For those researchers functioning in a science-domain-oriented environment, the analysis in this paper suggests that there are more options than the frequently proposed ‘more participation’ for increasing the likelihood that their policy-oriented work is used. These include establishing contacts with research groups or institutes that are in a position to function as ‘stepping stones’, or engaging with others to develop a social network in the policy sphere.

The success of environmental and commercial forestry plantings in Australia's low to medium rainfall zones can be affected by drought-induced mortality during seedling establishment. An analysis of management options for climatic risk like those employed for agricultural industries requires both experimental data and simulation capacity. The growth and survival of two eucalypt species ( Eucalyptus melliodora and Eucalyptus populnea ) was studied under a range of drought conditions in a field trial near Toowoomba in Queensland, Australia (27°34′S, 151°52′E). The APSIM model was applied to this dataset to investigate the effects of various plant traits on the differing survival rates of each species under the imposed drought conditions. The model adequately described the growth, water use and mortality of both species. A simulation analysis of the long-term survival rates under a range of planting conditions was performed indicating that 90 mm of stored soil moisture was required for managing the risk of establishment losses across a large range of climatic conditions.

The Agricultural Production Systems Simulator (APSIM) is a modular modelling framework that has been developed by the Agricultural Production Systems Research Unit in Australia. APSIM was developed to simulate biophysical process in farming systems, in particular where there is interest in the economic and ecological outcomes of management practice in the face of climatic risk. The paper outlines APSIM's structure and provides details of the concepts behind the different plant, soil and management modules. These modules include a diverse range of crops, pastures and trees, soil processes including water balance, N and P transformations, soil pH, erosion and a full range of management controls. Reports of APSIM testing in a diverse range of systems and environments are summarised. An example of model performance in a long-term cropping systems trial is provided. APSIM has been used in a broad range of applications, including support for on-farm decision making, farming systems design for production or resource management objectives, assessment of the value of seasonal climate forecasting, analysis of supply chain issues in agribusiness activities, development of waste management guidelines, risk assessment for government policy making and as a guide to research and education activity. An extensive citation list for these model testing and application studies is provided.

The long-term benefits of retaining or planting trees on farms to rehabilitate land and protect the soil from erosion or salinity problems has to be traded off against the impact of tree competition on commercial crops, especially in the medium to low rainfall regions of Australia. The incentive to plant trees would increase if tree competition could be offset by economic returns gained from farm forestry products and by the beneficial impacts of tree windbreaks on crop yields and resource sustainability. The Agricultural Production Systems Simulator (APSIM) has a well-established capability to simulate cropping systems and this paper reports on progress in applying APSIM to agroforestry systems in order to quantify the potential benefits and risks of planting trees as windbreaks to cropping land in Australia. A simple case study indicating one possible model configuration is used to demonstrate this emergent capability for simulating tree and crop productivity and their interactions. The simulated agroforestry system consisted of the growth of a belt of trees (Eucalyptus argophloia) positioned as a windbreak on the edge of a field where a crop of winter chickpea (Cicer arietinum L.) is grown over a 30-year period. The example simulations quantify the yield and economic returns of annual chickpea crops in addition to the discounted economic return from timber production after 30 years of tree growth. This example demonstrates how APSIM can be used to quantify the economic tradeoffs in planting trees as windbreaks on a commercial farm in a low rainfall region of Australia.

The Agricultural Production Systems slMulator, APSIM, is a cropping system modelling environment that simulates the dynamics of soil-plant-management interactions within a single crop or a cropping system. Adaptation of previously developed crop models has resulted in multiple crop modules in APSIM, which have low scientific transparency and code efficiency. A generic crop model template (GCROP) has been developed to capture unifying physiological principles across crops (plant types) and to provide modular and efficient code for crop modelling. It comprises a standard crop interface to the APSIM engine, a generic crop model structure, a crop process library, and well-structured crop parameter files. The process library contains the major science underpinning the crop models and incorporates generic routines based on physiological principles for growth and development processes that are common across crops. It allows APSIM to simulate different crops using the same set of computer code. The generic model structure and parameter files provide an easy way to test, modify, exchange and compare modelling approaches at process level without necessitating changes in the code. The standard interface generalises the model inputs and outputs, and utilises a standard protocol to communicate with other APSIM modules through the APSIM engine. The crop template serves as a convenient means to test new insights and compare approaches to component modelling, while maintaining a focus on predictive capability. This paper describes and discusses the scientific basis, the design, implementation and future development of the crop template in APSIM. On this basis, we argue that the combination of good software engineering with sound crop science can enhance the rate of advance in crop modelling. Crown Copyright (C) 2002 Published by Elsevier Science B.V. All rights reserved.

Pearl millet landraces from Rajasthan, India, yield significantly less than improved cultivars under optimum growing conditions, but not under stressed conditions. To successfully develop a simulation model for pearl millet, capable of capturing such genotype×environment (G×E) interactions for grain yield, we need to understand the causes of the observed yield interaction. The aim of this paper is to quantify the key parameters that determine the accumulation and partitioning of biomass: the light extinction coefficient, radiation use efficiency (RUE), pattern of dry matter allocation to the leaf blades, the determination of grain number, and the rate and duration of dry matter accumulation into individual grains. We used data on improved cultivars and landraces, obtained from both published and unpublished sources collected at ICRISAT, Patancheru, India. Where possible, the effects of cultivar and axis (main shoot vs. tillers) on these parameters were analysed, as previous research suggested that G×E interactions for grain yield are associated with differences in tillering habit. Our results indicated there were no cultivar differences in extinction coefficient, RUE, and biomass partitioning before anthesis, and differences between axes in biomass partitioning were negligible. This indicates there was no basis for cultivar differences in the potential grain yield. Landraces, however, produced consistently less grain yield for a given rate of dry matter accumulation at anthesis than did improved cultivars. This was caused by a combination of low grain number and small grain size. The latter was predominantly due to a lower grain growth rate, as genotypic differences in the duration of grain filling were relatively small. Main shoot and tillers also had a similar duration of grain filling. The low grain yield of the landraces was associated with profuse nodal tillering, supporting the hypothesis that grain yield was below the potential yield that could be supported by assimilate availability. We hypothesise this is a survival strategy, which enhances the prospects to escape the effects of stress around anthesis.

FARMSCAPE (Farmers', Advisers', Researchers', Monitoring, Simulation, Communication And Performance Evaluation) is a program of participatory research with the farming community of northeast Australia. It initially involved research to explore whether farmers and their advisers could gain benefit from tools such as soil characterisation and sampling, climate forecasts and, in particular, simulation modelling. Its current focus is facilitating the implementation of commercial delivery systems for these same tools in order to meet industry demand for their access. This paper presents the story of what was done over the past decade, it provides performance indicators of impact, it reflects on what was learnt over this period and it outlines where this research is likely to head in the future. Over the past 10 years, the FARMSCAPE team employed a Participatory Action Research approach to explore whether farmers could value simulation as a decision support tool for managing their farming system and if so, could it be delivered cost-effectively. Through farmer group engagement, on-farm trials, soil characterisation, monitoring of crops, soils and climate, and sessions to apply the APSIM systems simulator, FARMSCAPE represented a research program on decision support intervention. Initial scepticism by farmers and commercial consultants about the value of APSIM was addressed by testing its performance both against measured data from on-farm trials and against farmers' experiences with past commercial crops. Once this credibility check was passed, simulation sessions usually evolved into participants interactively inquiring of the model the consequence of alternative management options. These ‘What if’ questions using APSIM were contextualised using local climate and soil data and the farmer's actual or proposed management rules. The active participation of farmers and their advisers, and working in the context of their own farming operations, were the key ingredients in the design, implementation and interpretation of the FARMSCAPE approach to decision support. The attraction of the APSIM systems simulator to farmers contemplating change was that it allowed them to explore their own system in a manner equivalent to learning from experience. To achieve this, APSIM had to be credible and flexible. While direct engagement of farmers initially enabled only a limited number of beneficiaries, this approach generated a commercial market for timely and high quality interactions based on soil monitoring and simulation amongst a significant sector of the farming community. Current efforts are therefore focused on the training, support and accreditation of commercial agronomists in the application of the FARMSCAPE approach and tools. The FARMSCAPE approach to decision support has come to represent an approach to guiding science-based engagement with farm decision making which is being tested nationally and internationally.

The development of cropping systems simulation capabilities world-wide combined with easy access to powerful computing has resulted in a plethora of agricultural models and consequently, model applications. Nonetheless, the scientific credibility of such applications and their relevance to farming practice is still being questioned. Our objective in this paper is to highlight some of the model applications from which benefits for farmers were or could be obtained via changed agricultural practice or policy. Changed on-farm practice due to the direct contribution of modelling, while keenly sought after, may in some cases be less achievable than a contribution via agricultural policies. This paper is intended to give some guidance for future model applications. It is not a comprehensive review of model applications, nor is it intended to discuss modelling in the context of social science or extension policy. Rather, we take snapshots around the globe to 'take stock' and to demonstrate that well-defined financial and environmental benefits can be obtained on-farm from the use of models. We highlight the importance of 'relevance' and hence the importance of true partnerships between all stakeholders (farmer, scientists, advisers) for the successful development and adoption of simulation approaches. Specifically, we address some key points that are essential for successful model applications such as: (1) issues to be addressed must be neither trivial nor obvious; (2) a modelling approach must reduce complexity rather than proliferate choices in order to aid the decision-making process (3) the cropping systems must be sufficiently flexible to allow management interventions based on insights gained from models. The pro and cons of normative approaches (e.g. decision support software that can reach a wide audience quickly but are often poorly contextualized for any individual client) versus model applications within the context of an individual client's situation will also be discussed. We suggest that a tandem approach is necessary whereby the latter is used in the early stages of model application for confidence building amongst client groups. This paper focuses on five specific regions that differ fundamentally in terms of environment and socio-economic structure and hence in their requirements for successful model applications. Specifically, we will give examples from Australia and South America (high climatic variability, large areas, low input, technologically advanced); Africa (high climatic variability, small areas, low input, subsistence agriculture); India (high climatic variability, small areas, medium level inputs, technologically progressing; and Europe (relatively low climatic variability, small areas, high input, technologically advanced). The contrast between Australia and Europe will further demonstrate how successful model applications are strongly influenced by the policy framework within which producers operate. We suggest that this might eventually lead to better adoption of fully integrated systems approaches and result in the development of resilient farming systems that are in tune with current climatic conditions and are adaptable to biophysical and socioeconomic variability and change. (C) 2001 Elsevier Science Ltd. All rights reserved.

Tillering is an important adaptive feature of pearl millet (Pennisetum americanum [Pennisetum glaucum]) to the unpredictable growing conditions of dry areas of the semiarid tropics. Yet, this feature has largely been ignored in the development of simulation models for pearl millet. The objective of this paper is to parameterize and validate a leaf area module for pearl millet, which dynamically simulates crop leaf area from the leaf area of individual axes through simulating inter-axis competition for light. To derive parameters for the model, four cultivars (BJ 104, WRajPop, HHB 67 and RCB-IC 911), contrasting in phenology and tillering habit, were grown under well-watered and well-fertilized conditions across a range of plant densities in three experiments at two locations in India (Patancheru, Andhra Pradesh and Jodhpur, Rajasthan) during 1996 and 1997. For selected plants, observations on the number of primary basal tillers and on the number of visible, fully expanded, and senesced leaves on each axis were made twice a week throughout the growing season. Occurrence of panicle initiation (PI) was observed in two experiments only, but data were complemented by published and unpublished data, obtained for comparable cultivars. Parameters were obtained for the time from emergence to PI as a function of daylength, leaf initiation rate, ate of leaf and tiller appearance, and the leaf senescence rate. Parameters for leaf size were determined in a previous paper. Our parameter estimates compared well with published data and were, with the exception of time to PI and leaf size, mostly independent of cultivar, axis and density. Genotypic effects on productive tiller number could be attributed to differences in main shoot leaf size. Validation of the leaf are module showed that the module adequately reproduced the effects of density, photoperiod and genotype on the leaf area of individual axes and on productive tiller number. This was despite the fact that the reduction in leaf area of non-productive tillers was achieved in the module through a reduction in leaf size, whereas the crop reduced leaf area through a reduction in leaf number. Our results indicate that leaf area index (LAI) of a tillering crop can be simulated adequately by simulating LAI from individual leaf area and incorporating the effects of competition for light

Pearl millet (Pennisteum americanum L.) is an essential crop in farming systems of the dry areas of the semi-arid tropics and its tillering habit is an important adaptive feature. This is the first paper in a series aiming at developing and validating a pearl millet simulation model that recognises tillers as functional entities, analogous to intercrops. The objective of this paper is to quantify the effects of total leaf number per axis (TLN), cultivar, plant density and axis number on parameters that are used to simulate potential leaf area per plant. Four cultivars with different phenology and tillering habit were grown under well-watered and well-fertilised conditions at two locations in India, covering a range of daylengths and plant densities. For selected plants, the area of fully expanded leaves was measured non-destructively. A bell-shaped function adequately described the relationship between individual leaf area and leaf position on an axis. Its shape was determined by the position (X0) and area (Y0) of the largest leaf and by the breadth and skewness of the leaf area profile curve. TLN affected all four parameters, although the association with Y0 was weak. Cultivar only affected Y0, suggesting that parameterising new cultivars is straightforward. The observed density effect confirmed that competition for light between axes started during stem elongation. The results highlighted the consistent differences between leaf area profiles of main shoots and tillers. For a high-tillering crop like pearl millet, modelling leaf area dynamics through individual leaves is justified, as this approach can potentially deal with cultivar and environmental effects on tillering.

A simulation model of pigeonpea is described that is designed to simulate the development, growth, nitrogen accumulation and yield of a wide range of maturity types from extra-short to medium-duration in response to weather, soil conditions and agronomic management. Parameters of the model for phenological development, leaf area expansion, radiation interception, biomass accumulation and partitioning, crop water use, root growth and water extraction, and nitrogen accumulation are derived from published studies. In addition, the calibration exercise is described to derive the parameters accounting for the effect of plant population density on leaf area expansion. The model was tested on 38 data sets, not previously used to derive model parameters, collected at Patancheru, India. Data sets encompassed a wide range of plant type, sowing density, and seasons, grown on alfisol and vertisol soil types under dryland and irrigated conditions. The time to flowering and maturity from the extra-short to medium-duration types were simulated well, explaining 96 and 92% of the variance (RMSD=4.3 and 9.8 days), respectively. Total aboveground biomass was simulated with less accuracy, explaining 74% of the variance ( RMSD =2056 kg ha −1 ) and grain yield was simulated explaining 76% of the variance ( RMSD =332 kg ha −1 ) . There remains scope for model improvement in the areas of waterlogging and testing on crop N accumulation. This pigeonpea module, when coupled with other crop, soil and management modules can be used to address a range of cropping systems issues.

This paper quantifies the effects of N-stress on development and growth of sorghum by identifying critical values for stover N% (SNC) and specific leaf nitrogen (SLN) for a range of physiological processes (rates of leaf appearance, leaf expansion, leaf area increase, and biomass accumulation). It also compares the merits of the SNC and SLN approach for implementation in the N-routines of crop growth simulation models. Data from experiments covering a range of nitrogen treatments, grown at three contrasting locations in Australia, were used in the analyses. For the rate of biomass accumulation, the critical and minimum SLN were constant for most of the growing season, except for early stages prior to approximately panicle initiation. For SNC, however, the critical and minimum values continued to decline throughout the growing season, although the decline was limited at later development stages. During the stage of leaf growth, the critical N content was similar for all processes considered. Although critical SLN was less sensitive to development stage than SNC, identification of a critical SLN was hampered if environmental effects on specific leaf area were present. The merit of each approach in simulation models will depend upon their ability to capture genotypic differences in N-dynamics. (C) 2001 Elsevier Science B.V. All rights reserved.

Data from sowing-date and other experiments conducted for nine cultivars at three locations ranging from 1°30′S to 27°15′N were analysed for photoperiod response. All cultivars were found to have a qualitative response to photoperiod. The results of the analysis show that cultivars previously reported to be “relatively insensitive” to photoperiod were, in fact, highly sensitive. Flowering in short-duration cultivars was delayed by up to a 100 days when daylength in the photoperiod-inductive phase exceeded a critical value. Medium- and long-duration cultivars delayed flowering by over 150 days in response to photoperiod. A model was able to predict this wide range in flowering dates.

Long-term historical weather data are needed to conduct crop simulation analyses. However, the network of weather recording stations which collect all necessary daily weather data (commonly rainfall, solar radiation, maximum and minimum temperature) for such analyses is sparse. Frequently only rainfall is recorded. Thus, weather data generation techniques are required for three situations: (i) where only rainfall data are available, (ii) where both rainfall and temperature data are available, but radiation is missing, and (iii) where records are otherwise complete, but techniques are cequired to fill short periods of missing data. Three weather generation techniques are compared, termed here (i) Bristow and Campbell's method, (ii) TAMSIM and (iii) WGEN. Methods (ii) and (iii) were used to generate temperature and radiation data to accompany recorded rainfall records, and methods (i)–(iii) to generate a solar radiation record to accompany recorded temperature and rainfall records. Data from four stations in tropical and subtropical Australia with long-term complete weather records were used to compare actual with generated data sets. Results were evaluated firstly by comparing the cumulative distribution function (CDF) of generated and actual values, and secondly by comparing CDFs calculated from the output of three crop simulation models used with the generated and actual data sets. Generally the distributions of radiation and temperature differed significantly. However, when the weather data sets were used by simulation models to estimate biomass, only 10 of the 50 CDFs differed significantly. When both temperature and radiation were generated, 30% of CDFs from TAMSIM and 20% of WGEN differed significantly. When only radiation was generated, 40% of CDFs generated by the Bristow and Campbell's method, 10% of WGEN and none of TAMSIM differed significantly. All methods simulated the more temperate sites with higher precision than the wet, tropical site. Simulated yields showed a similar pattern. It was concluded that where both temperature and radiation data have to be generated, WGEN is appropriate because it contains a stochastic element and thus simulates catastrophic events such as frosts. Where only radiation generation is required, both WGEN and TAMSIM performed adequately. Where temperature or radiation data sets are complete except for occasional missing days, TAMSIM was considered to be the most appropriate. In cases were the objective is not to conduct long-term simulation analyses, Bristow and Campbell's method appeared more appropriate because of its ability to better simulate the day to day variation in solar radiation.

The development and validation of simulation models requires access to extensive and detailed datasets on the subject to be modelled. The availability of such data is often limited. Consequently, this paper explores the use of resampling techniques in the estimation of parameters for and the assessment of accuracy of simulation models. A description of jackknife and cross-validation techniques is presented, as well as an application of these techniques to the fitting of a crop simulation model to a limited dataset. The example concerns the application of a simulation model for the growth of maize (Zea mays) in northern Australia. Jackknife techniques were applied to the estimation of the potential kernel number and the potential kernel growth rate of a specified maize hybrid, and the prediction accuracy of the estimation of grain yield was assessed by cross-validation. The jackknife estimates were found to differ from the estimates obtained from a single fit either to all the data or to subsets of half the data sampled from the datasets. In addition, a more reliable estimate of the prediction variance was found from the cross-validation step compared to that found in the more traditional validation method of using one half of a dataset independent of the estimation half. The use of the jackknife and cross-validation techniques permitted a limited dataset to be used in both the parameter estimation and validation processes of model development.

A more-efficient use of resources is a major reason advanced for the advantage of intercropping over alternative cropping systems. This paper reviews the interception and utilization of solar radiation in intercropping systems. To achieve this, a framework is presented on radiation interception and use in sole crops and elements of this framework are subsequently extended to intercropping systems. Improved productivity per unit incident radiation could be achieved by the adoption of an intercropping system that either increases the interception of solar radiation and/or has greater radiation-use efficiency. Minimizing the proportion of radiant energy reaching the ground is a simple means of promoting efficient utilization of incident solar radiation. Examples are discussed where advantages acrued from intercropping short- and long-duration species due to enhanced radiation capture over time. Frequent confounding with plant density effects has meant that the evidence for intercropping to enhance radiation capture in space is equivocal. Examples are also presented, demonstrating improved utilization of radiant energy, once intercepted, either by more-efficient production of biomass or increased proportion of biomass partitioned to yield. The potential variety and complexity of intercropping systems, means that recommendations on how to optimise radiation capture and use in intercrops could not readily be determined. Attempts to quantify and predict intercrop performance are examined and dynamic simulation models considered to offer a means by which intercrop performance could be optimised.

Severe water deficit is common during crop establishment in many areas of the semi-arid tropics. The influence of non-lethal water deficit imposed for 19 days after sowing was investigated in maize using a water gradient applied with an irrigation system mounted under a rainshelter. Water deficit had little effect on timing of emergence but delayed tassel initiation and silking by up to 4 days. Water deficit reduced the rate of leaf appearance and area of leaves 5 to 9, but did not greatly influence the number of leaves or total leaf area per plant. Marked differences in height and biomass, that were apparent on day 21, were not evident on day 51 and there was no influence of water deficit on mean ±s.e. grain yield (77 ± 6.3 g plant −1 ) or grain numbers (391 ± 32 grain plant −1 ). The threshold of available soil water below which water deficit affects phenology and leaf appearance appeared to be lower than for the effects on leaf expansion and accumulation of biomass above ground. We conclude that non-lethal water deficit during early vegetative growth of maize results in marked reduction in height and biomass and delays crop phenology but may not be associated with lower yield potential.

Routines controlling development of leaf area in the CERES-Maize model have been misinterpreted in recent papers, owing to an inaccuracy in parameter definition in the published documentation of the model. This paper provides the correct description of functions for leaf-area development in CERES-Maize, and tests the performance of these functions. Given the correct definitions of the variables cumph , as the cumulative number of visible leaf tips, and xn as the next leaf which will emerge on the plant, CERES-Maize simulates leaf-area expansion per plant as a function of the number of emerged leaf tips. A test of the leaf-canopy routines in CERES-Maize indicated that constants, used in the functions describing both rate of appearance of leaf tips and rate of leaf-area expansion per plant, needed calibration for predicting leaf-area development of the tropically adapted hybrid maize Dekalb XL82.

Temperature controls pre-emergent development of seedlings, and is generally the prime determinant of the duration from sowing to seedling emergence. This study was conducted to examine seed germination and the elongation of the hypocotyl and radicle of kenaf (Hibiscus cannabinus L.) under constant temperature conditions, and to provide the temperature parameters necessary to model the processes of seedling emergence. Seed of kenaf, cv. Guatemala-4, were germinated in incubators at 8 constant temperatures (15, 20, 25, 30, 35, 40, 43 and 45°C). The numbers of germinated seeds and lengths of the hypocotyl and radicle were measured over time in each temperature regime. Final germination percentage was unaffected by temperatures in the range 15–35°C, but declined sharply at the higher temperatures. The optimum (To) and maximum temperatures (Tm) for rate of seed germination, which was defined as the inverse of the half-time of germination, were 35°C and 45°C, respectively. Values of To and Tm for the lag and linear phases of hypocotyl elongation were about 31°C and 43°C, respectively. Similar values for To and Tm were calculated for elongation of the radicle. Above a base temperature (Tb) of 10°C, the thermal time requirement for 50% seed germination was 8.2°C d, the lag phase of hypocotyl elongation required 17.0°C d, and linear hypocotyl elongation required 0.45°C d mm−1. It is proposed that the emergence of kenaf seedlings can be modelled from a specified sowing depth by the summation of the thermal time required for germination and the lag and linear phases of hypocotyl elongation.

Phenology and leaf growth of the tropical sorghum ( Sorghum bicolor (L.) Moench) hybrid Dekalb DK55 were examined as functions of photoperiod and temperature. Data were collected on crops sown at different dates and grown under adequate water and nutrient supply in a semiarid tropical environment. The study was undertaken to develop mathematical relationships describing phenology and development and maintenance of leaf area to allow the prediction of green leaf area during growth. Photoperiod affected the timing of panicle initiation and, indirectly, the date of anthesis, but had no effect on the duration of grain-filling. Rather, the thermal time (base temperature of 7°C) for grain-filling was relatively constant across sowings (mean 680°C d). Final leaf number was linearly related to thermal time, with a leaf-initiation rate of 41.3°C d per leaf. A constant appearance rate of fully expanded leaves of 69.0°C d per leaf was determined over all sowings. The fully expanded area of each leaf was described as a function of leaf number, and when account was taken of the expanding leaf area, cumulative total leaf area could be estimated based on leaf number. Leaf senescence was better related to calendar time than to thermal time, although there was significant scatter in both relationships. The paper outlines a framework to allow the simulation of phenology and green leaf area that may be used for different sorghum genotypes in different locations. Essentially, it is necessary to (i) predict total leaf number as a function of photoperiod and thermal time, (ii) predict leaf appearance as a function of thermal time, (iii) predict total area per plant based on leaf number, (iv) predict the proportion of senesced leaf area, and (v) discount total leaf area according to senescence and calculate green leaf area during growth.

Crop simulation models relevant to real-world agriculture have been a rationale for model development over many years. However, as crop models are generally developed and tested against experimental data and with large systematic gaps often reported between experimental and farmer yields, the relevance of simulated yields to the commercial yields of field crops may be questioned. This is the third paper in a series which describes a substantial effort to deliver model-based decision support to Australian farmers. First, the performance of the cropping systems simulator, APSIM, in simulating commercial crop yields is reported across a range of field crops and agricultural regions. Second, how APSIM is used in gaining farmer credibility for their planning and decision making is described using actual case studies. Information was collated on APSIM performance in simulating the yields of over 700 commercial crops of barley, canola, chickpea, cotton, maize, mungbean, sorghum, sugarcane, and wheat monitored over the period 1992 to 2007 in all cropping regions of Australia. This evidence indicated that APSIM can predict the performance of commercial crops at a level close to that reported for its performance against experimental yields. Importantly, an essential requirement for simulating commercial yields across the Australian dryland cropping regions is to accurately describe the resources available to the crop being simulated, particularly soil water and nitrogen. Five case studies of using APSIM with farmers are described in order to demonstrate how model credibility was gained in the context of each circumstance. The proposed process for creating mutual understanding and credibility involved dealing with immediate questions of the involved farmers, contextualising the simulations to the specific situation in question, providing simulation outputs in an iterative process, and together reviewing the ensuing seasonal results against provided simulations. This paper is distinct from many other reports testing the performance and utility of cropping systems models. Here, the measured yields are from commercial crops not experimental plots and the described applications were from real-life situations identified by farmers. A key conclusion, from 17 years of effort, is the proven ability of APSIM to simulate yields from commercial crops provided soil properties are well characterised. Thus, the ambition of models being relevant to real-world agriculture is indeed attainable, at least in situations where biotic stresses are manageable.

In Australia, a land subject to high annual variation in grain yields, farmers find it challenging to adjust crop production inputs to yield prospects. Scientists have responded to this problem by developing Decision Support Systems, yet the scientists’ enthusiasm for developing these tools has not been reciprocated by farm managers or their advisers, who mostly continue to avoid their use. Preceding papers in this series described the FARMSCAPE intervention: a new paradigm for decision support that had significant effects on farmers and their advisers. These effects were achieved in large measure because of the intensive effort which scientists invested in engaging with their clients. However, such intensive effort is time consuming and economically unsustainable and there remained a need for a more cost-effective tool. In this paper, we report on the evolution, structure, and performance of Yield Prophet®: an internet service designed to move on from the FARMSCAPE model to a less intensive, yet high quality, service to reduce farmer uncertainty about yield prospects and the potential effects of alternative management practices on crop production and income. Compared with conventional Decision Support Systems, Yield Prophet offers flexibility in problem definition and allows farmers to more realistically specify the problems in their fields. Yield Prophet also uniquely provides a means for virtual monitoring of the progress of a crop throughout the season. This is particularly important for in-season decision support and for frequent reviewing, in real time, of the consequences of past decisions and past events on likely future outcomes. The Yield Prophet approach to decision support is consistent with two important, but often ignored, lessons from decision science: that managers make their decisions by satisficing rather than optimising and that managers’ fluid approach to decision making requires ongoing monitoring of the consequences of past decisions.

This paper reports on the design and performance of an irrigation system associated with a rain shelter. The shelter is used in the closed position as a platform for multiple spray lines. This system automatically supplies, at a specified time, water of specified depth and delivery rate, either uniformly or on a specified gradient across the sheltered area. Variations in crop water deficits and yields have conformed closely to imposed water gradients.

Areas of brigalow (Acacia harpophylla) dominated landscapes in north-eastern Australia have declined drastically due to major clearing and agricultural expansion during the late 1940s and early 1960s. The inherently high salt content of the soils of this region present a potential downstream salinity hazard from groundwater recharge. Chronosequence analysis using paired chloride profiles from soil cores taken beneath brigalow remnants and adjacent pasture or cropping lands provide a tracer for quantifying historic recharge rates as a consequence of vegetation management and agricultural practice. Present day chloride levels are the direct result of past land management. In this paper we present the results of simulation studies used to benchmark historic management practice since clearing in terms of chloride leaching and drainage. These simulations estimated that 15.3 t/ha of chloride leached from the top 150 cm in 7 major drainage events (>15mm) over a 34-year period, and that these leaching events corresponded with peaks in rainfall cycles. Use of virtual experiments to investigate alternative cropping systems found significant increases in the frequency and magnitude of drainage events of no-tillage wheat compared with sorghum grown in a summer-rainfall region. Systems simulation can provide guidelines for designing cropping systems which best balance production with drainage objectives in dryland farming systems.

Yield advantages of crops grown behind windbreaks have often been reported, but underlying principles responsible for such changes and their long-term consequences on crop productivity and hence farm income have rarely been quantified. Physiologically and physically sound simulation models could help to achieve this quantification. Hence, the APSIM systems model, which is based on physiological principles such as transpiration efficiency and radiation use efficiency (termed here APSIMTE), and the Soil Canopy Atmosphere Model (SCAM), which is based on the Penman–Monteith equation but includes a full surface energy balance, were employed in developing an approach to quantify such windbreak effects. This resulted in a modified APSIM version (APSIMEO), containing the original Penman equation and a calibration factor to account for crop- and site-specific differences, which were tested against field data and simulations from both the standard APSIMTE and SCAM models. The APSIMEO approach was tested against field data for wheat and mungbean grown in artificial enclosures in south-east Queensland and in south-east Western Australia. For these sheltered conditions, daily transpiration demand estimates from APSIMEO compared closely to SCAM. As the APSIMEO approach needed to be calibrated for individual crops and environments, average transpiration demand for open field conditions predicted by APSIMEO for a given site was adjusted to equal that obtained using APSIMTE by modifying a calibration parameter β. For wheat, a β-value of 1.0 resulted in best fits for Queensland, while for Western Australia a value of 0.85 was necessary. For mungbean a value of 0.92 resulted in the best fit (Qld). Biomass and yields simulated by APSIMTE and the calibration APSIMEO for wheat and mungbean grown in artificial enclosures were generally distributed around the 1:1 line, with R2 values ranging from 0.92 to 0.97. Finally, APSIMEO was run at 2 sites using long-term climate data to assess the likely year-to-year variability of windbreak effects on crop yields. Assuming a 70% reduction in wind speed as representing the maximum potential windbreak effect, the average yield improvement for the Queensland site was 13% for wheat and 3% for mungbean. For wheat at the WA site the average yield improvement from reduced wind speed was 5%. In any year, however, effects varied from negative, neutral to positive, highlighting the highly variable nature of the expression of windbreak effects. This study has shown how physical and biological modelling approaches can be combined to aid our understanding of systems processes. Both the environmental physics perspective and the biological perspective have shortcomings when issues that sit at the interface of both approaches need to be addressed. While the physical approach has clear advantages when investigating changes in physical parameters such as wind speed, vapour pressure deficit (VPD), temperature or the energy balance of the soil–plant–atmosphere continuum, it cannot deal with complex, biological systems adequately. Conversely, the crop physiological approach can handle such biological interactions in a scientific and robust way while certain atmospheric processes are not considered. The challenge was not to try and capture all these effects in 1 model, but rather to structure a modelling approach in a way that allowed for inclusion of such processes where necessary.

This overview paper presents a description of the National Windbreaks Program (NWP) — its objectives, the main methods used to achieve these objectives and a summary of the key results. It draws these from the individual papers appearing in this special issue, which provide detailed descriptions and discussion about the specific research sites and research methods used, in addition to interpreting and discussing the results. The key findings were the following: (i) Two broad areas of crop and pasture response can be identified downwind of a porous windbreak: a zone of reduced yield associated with competition with the windbreak trees that extended from 1 H to 3 H, where H is the windbreak height, and a zone of unchanged or slightly increased yield stretching downwind to 10 H or 20 H. (ii) Averaged over the paddock, yield gains due to the effect of shelter on microclimate were smaller than expected — especially for cereals. Yield simulations conducted using the APSIM model and 20 years of historical climate data confirmed this result for longer periods and for other crop growing regions in Australia. Larger yield gains were simulated at locations where the latter part of the growing season was characterised by high atmospheric demand and a depleted soil water store. (iii) Economic analyses that account for the costs of establishing windbreaks, losses due to competition and yield gains as a result of shelter found that windbreaks will either lead to a small financial gain or be cost neutral. (iv) Part of the reason for the relatively small changes in yield measured at the field sites was the variable wind climate which meant that the crop was only sheltered for a small proportion of the growing season. In much of southern Australia, where the day-to-day and seasonal variability in wind direction is large, additional windbreaks planted around the paddock perimeter or as closely-spaced rows within the paddock will be needed to provide more consistent levels of shelter. (v) Protection from infrequent, high magnitude wind events that cause plant damage and soil erosion was observed to lead to the largest yield gains. The main forms of direct damage were sandblasting, which either buries or removes seedlings from the soil or damages the leaves and stems, and direct leaf tearing and stripping. (vi) A corollary to these findings is the differing effect that porous windbreaks have on the air temperature and humidity compared to wind. While winds are reduced in strength in a zone that extends from 5 H upwind to at least 25 H downwind of the windbreak, the effects of shelter on temperature and humidity are smaller and restricted mainly to the quiet zone. This means that fewer windbreaks are required to achieve reductions in wind damage than for altering the microclimate. (vii) The wind tunnel experiments illustrate the important aspects of windbreak structure that determine the airflow downwind, and subsequent microclimate changes, in winds oriented both perpendicular and obliquely to porous windbreaks. These results enable a series of guidelines to be forwarded for designing windbreaks for Australian agricultural systems.

A major issue for the sustainability of cropping systems is the maintenance of soil fertility and especially the supply of nitrogen to cereal crops. Choice of appropriate management strategies, including the role of legumes, is problematic, especially where climatic variation is large. Simulation models provide the means of extrapolation from the site- and season-specific bounds of experimental data to permit scenario analyses that can explore alternative management options. This paper is a status report on the capabilities of the APSIM modelling framework to simulate legume-cereal systems. APSIM deals with water and nitrogen constraints to crop growth and is well suited to the task of modelling whole systems involving crop rotations. The components that are not yet fully developed are modules for growing the legume crops and coupling these with the module describing the dynamics of soil organic matter to obtain sensible predictions of nitrogen supply to subsequent crops. Evidence is provided that those parts of the system that can be represented by current APSIM modules are predicted satisfactorily. The closest approach to a whole system that has been simulated to date is grass or legume (Stylosanthes hamata cv. Verano) leys followed by crops of maize or sorghum grown in experiments at Katherine, NT. Predictions of the yields of the leys and the cereal crops, especially the benefit from the legume leys to a second crop, were sufficiently close to measured yields to suggest that there are good prospects for developing useful models of other systems involving legumes and cereals. A simulation scenario exploring a chickpea-wheat system demonstrates how models can be used to analyse both productivity and sustainability aspects of the system.

An innovative ley farming system, involving cereal crops grown in rotation with pasture legumes, has been tentatively adopted by farmers in the semi-arid tropics of northern Australia. Yet, after more than a decade of experimental research, the long-term

The feasibility of utilizing kenaf (Hibiscus cannabinus L.) as a fibre crop and potential source of paper pulp is being investigated. This paper is the first in a series which develops and validates a simulation model of kenaf, and applies it to assessing the potential for dryland production in regions of tropical Australia. The duration from sowing to flowering is an important determinant of fibre yield in kenaf. Accordingly, the effect of temperature and photoperiod on the phenology of kenaf was examined. Data were collated from sowing-date experiments on kenaf cultivars Guatemala 4 and Everglades 71 at six locations in tropical Australia, ranging in latitude from 12� 54'S. to 19� 32'S. Data from one site, Kununurra, W.A. (15� 39'S.), was used to develop a model which described the duration from sowing to flowering in kenaf based on four stages: (1) sowing to emergence; (2) emergence to the end of the basic vegetative phase (BVP); (3) a photoperiod-induced phase (PIP) which ends at floral initiation; and (4) a flower development phase (FDP). Both kenaf cultivars responded as qualitative short-day plants where flowering did not occur above a critical photoperiod of 12.9 h. Assuming a base photoperiod of 12.0 h, the thermal time required to complete the BVP for the two cultivars was similar. However, photoperiod sensitivity during PIP, and the thermal time required for FDP, were greater for Guatemala 4 than for Everglades 71. Validation of the model against independently observed data for both cultivars at other sites resulted in close predictions in the thermal time required for flowering. This phenological model for kenaf can be used to assess the effect of sowing date on phenology at different locations in tropical environments, and is a key component of a crop growth simulation model to assess the environmental constraints to productivity in these regions.

Crop simulation models are proposed as tools for agricultural risk analysis in order to explore potential cropping locations and appropriate farming systems in the semi-arid tropics. This study takes the initial step of independently validating the STANDARD CERES-Maize simulation model in the semi-arid tropics, and reports some modifications made to improve its performance. The CERES-Maize model did not accurately predict grain-yield of cultivar Dekalb XL82 which was grown over a range of sowing dates and water regimes at Katherine, N.T. Experimental yields (at 15.5% moisture) ranged from 0 to 9840 kg ha −1 . Calibration of CERES-Maize reduced the root mean square deviation ( rmsd ) for observed grain-yields from 3480 to 2015 kg ha −1 . Functions describing phenology, leaf growth and senescence, assimilate production and grain growth were revised and validated against field data. The revisions to CERES-Maize not only provide a model more applicable to the semi-arid tropics but also identify the parameters that may require calibration for other maize genotypes and locations in this climatic zone. Further validations of the functions describing nitrogen cycling and rainfall infiltration and runoff are required to increase the model's applicability to risk-analysis studies.

Pearl millet (Pennisetum americanum (L.) Leeke) was grown at ICRISAT, Hyderabad, India at 20 densities ranging from 50 000 to 400 000 plants ha−1 using a Nelder fan design. Studies were made on the effect of population on the distribution of plant dry weight, leaf area, grain yield and yield components throughout the season. The first effect of increasing population was evident at panicle initiation (

Phenology and leaf growth of the tropical maize hybrid Dekalb XL82 were examined as functions of photoperiod and temperatures. Data were collected from fully irrigated crops sown on three dates. The study was undertaken to obtain mathematical relationships describing phenology, and development and maintenance of leaf-area for use in crop-simulation models describing phenology, and development and maintenance of leaf-area for use in crop-simulation models describing the growth of maize in semi-arid tropical environments. The direct effect of photoperiod was on the thermal time from germination to tassel initiation, and subsequently on final leaf-number. Leaf appearance increased exponentially with thermal time, and the relationship was coincident for all sowings. The fully expanded area of each leaf was described as a function of leaf number, and the same relationship described each sowing, provided that account was taken of differences in final leaf-number across sowings. When account was taken of the leaves still expanding, the relationship between total per-plant leaf-area and leaf-number was similar across sowings. The relationship between leaf-area senescene and thermal time was significantly different among sowings, with the main difference accurring near maturity. The effects of water deficit on these relationships were also examined. Water deficit during early vegetative growth affected the rate of leaf appearance and leaf size, but not the final leaf-number nor leaf senescence. The main effect of water deficit during later growth was on leaf senescence.

Pennisetum americanum (L.) Leeke, hybrid cv. BJ 104 was examined at ICRISAT Hyderabad, India over a range of populations (50 000 to 400 000 plants ha−1) and, during early vegetative growth, three photoperiods (13.5, 14.5 and 15.5 h) were imposed. A growth analysis was carried out at weekly intervals. As the photoperiod lengthened from 13.5 to 14.5 and 15.5 h, the time taken to panicle initiation of the main axis increased from 16 to 23 and 34 days after emergence respectively. Grain yield was not affected by photoperiod at high populations but yield was reduced (up to 35%) when photoperiod was extended at low populations on account of a lesser contribution from tillers. The total number of tillers emerged per plant was unaffected by photoperiod but the number of productive tillers per plant at maturity was reduced by up to 38% as photoperiod was extended. Grain number was constant on the main axis. Grain number per tiller decreased in plants under extended photoperiods although there was some compensation through increased individual seed weights. Plants under extended photoperiods had greater vegetative dry weight and lower harvest indices than plants under natural photoperiods.