Identifying chickpea homoclimes using the APSIM chickpea model

CSIRO PUBLISHING www.publish.csiro.au/journals/ajar Australian Journal of Agricultural Research, 2008, 59, 260–269 Identifying chickpea homoclimes using the APSIM chickpea model Yash ChauhanA,B, Graeme WrightA, Nageswararao RachaputiA, and Kevin McCoskerA A B Department of Primary Industries and Fisheries, Kingaroy, PO Box 23, Qld 4610, Australia. Corresponding author. Email: [email protected] Abstract. Chickpea (Cicer arietinum L.) has been traditionally grown in India but is a relatively new export crop in Australia where its cultivation is expanding into new areas. The objective of this study was to identify homoclimes (i.e. similar chickpea-growing environments) in the major chickpea-growing areas of the 2 countries, using the Agricultural Production Systems Simulator (APSIM) chickpea model. The model, which processes climatic, soil, and plant information on a daily time step, was first validated and then used to simulate flowering, maturity, and grain yield of Amethyst, a midseason cultivar, and Barwon, a full-season cultivar, on low (100 mm), medium (150 mm), and high (190 mm) water-holding capacity soils, using historical climatic data of 67 Australian and 24 Indian locations. The mean of annual outputs of flowering, maturity, and grain yield of the 2 cultivars on 3 soils was then clustered using Ward’s hierarchical complete linkage clustering procedure. At a 90% level of similarity, all the locations could be grouped into 6 homoclime clusters. The Australian locations appeared more diverse as they were present in all the clusters, whereas the Indian locations were present only in clusters 1, 2, and 6. While there were clear geographical patterns of spread of these clusters, in Australia they were not entirely related to latitude. The cluster 1 and 2 locations, which represent the largest chickpea-growing area in Australia, had homoclime locations in common with northern India. The clustering of locations appeared generally consistent with the known adaptation of chickpea in different environments of the 2 countries and therefore suggests that the methodology could be potentially used for complementing conventional approaches of introducing or exchanging germplasm, as well as determining appropriateness of breeding/testing sites. Introduction Chickpea (Cicer arietinum L.) is a cool-season food legume, which is grown in Africa, West Asia, South Asia, and Europe. Its cultivation has recently expanded into non-traditional areas of North America and Australasia. As average grain yields of chickpea are generally <1 t/ha, growers, especially in new regions, are always seeking to adopt higher yielding and better adapted cultivars either bred locally or introduced from other national and international breeding programs or seed companies to improve their gross margins. In chickpea, large genotype
environment (G
E) interactions for seed yield have been recently reported (Berger et al. 2004, 2006) and an apparent association between germplasm origin and specific adaptation has also been uncovered (Berger et al. 2006). Such interactions make it difficult to make recommendations for new areas without conducting elaborate variety evaluation trials over several sites and seasons. However, conducting such large multilocation trials may not always be feasible due to logistical and other reasons. It would therefore be appropriate to complement current varietal testing approaches with other methods, such as identification of homoclimes (similar climatic environments), that could hasten the introduction of high-yielding cultivars into new areas. The homoclime analysis approach has previously been applied to determine adaptation ranges of perennial crops or trees, which can take many years to establish (Russell and Moore 1976; Booth et al. 1987; Smart 2003). It has also been used to identify climatic adaptation ranges of various pasture legume species (Russell and Webb 1976), as
CSIRO 2008 well as to examine the potential for invasion of annual weed species (Holt and Boose 2000). An underlying assumption in the homoclime approach is that the potential for expression on adaptive plant traits of a given cultivar will be similar within a homoclime and hence its performance. Indeed, Malhotra and Singh (1991) reported that genotype
environment interaction was minimal within a cluster formed using flowering and grain yield data of 2 international chickpea trials. Limited work conducted to describe the West Asia North Africa (WANA) chickpea region in terms of its climatic profile and nature of stresses experienced by the crop (Saxena et al. 1996) suggested that this approach, if applied on a wider scale, can lead to a better understanding of adaptation ranges in this crop as well. This was further supported by a recent eco-graphic analysis of chickpea by Berger and Turner (2007) which showed that chickpea is grown in a wide range of habitats characterised by different climates that exert different selection pressures on the crop. Two critical issues relevant to using the homoclime approach in chickpea are the definition and development of appropriate tools to identify homoclimes. Traditional homoclime approaches have often used physical variables to classify environments, and generally have ignored their effects on plant responses and hence may not be crop specific. To capture the crop specificity in a better way it would, however, be ideal if environments were characterised based on the crop’s response to environments. For example, environmental factors such as temperature, photoperiod and seasonal rainfall can have significant effects on crop phenology and yield. Defining a 10.1071/AR07380 0004-9409/08/030260 Identifying chickpea homoclimes homoclime only in terms of rainfall or temperature averages using pattern analysis may be less informative than defining it in terms of stresses resulting from these climatic variables, and also mechanisms the crop uses to cope with these stresses. Drought and cold stress are the two major abiotic stress factors that have been identified to affect chickpea adaptation in different regions (Berger and Turner 2007). The crop’s ability to tolerate these stresses in part is conferred via its phenology which is a cultivar specific characteristic (Berger et al. 2004, 2006) modulated by temperature and photoperiod (Roberts et al. 1985). Chickpea phenology is also affected by low temperature (<15˚C) and drought stress (Singh 1991; Clarke and Siddique 2004). For chickpea a homoclime, for example, could be a group of environments creating a similar degree of drought or cold stress, as well as modulation of phenology to cope with these stresses. There could be substantial application of this type of homoclime approach if it could characterise environments in such a way. In the past, flowering and grain yield data recorded in yield trials have been used to characterise chickpea environments with some success (Malhotra and Singh 1991). A new approach for this purpose could be to generate such data using a simulation modelling framework. The Agricultural Production Systems Simulator (APSIM) developed in Australia is one such modelling framework that has the ability process climatic, plant and soil information (McCown et al. 1996). The model been successfully used to characterise sorghum (Sorghum bicolour L.) drought environments (Chapman et al. 2002), and is being applied to decipher gene-to-phenotype relationships in order to improve plant breeding strategies (Chapman et al. 2003). The model can also simulate chickpea growth and grain yield (Robertson et al. 2002). It has, however, not been applied to identify chickpea homoclimes. The objective of this study was to explore if the APSIM chickpea model could be used to identify homoclimes, using locations in Australia where chickpea is a relatively new crop and covers a range of the temperate, sub-tropical and tropical environments, and in India where the crop has been grown for a long time in subtropical and tropical environments. In addition, historical daily climatic and validation datasets for the chickpea model were available for many locations of both countries. Australian Journal of Agricultural Research 261 radiation, ambient temperature, soil water, and nitrogen supply on a daily time step. The model was calibrated using data collected in central and south-east Queensland, and New South Wales, and has not been widely applied to winter sowings in Mediterranean type environments of Southern Australia, and to autumn sowings which typify Indian production systems. In the Mediterranean type environments of Australia and northern Indian environments, post-anthesis temperature of <15˚C inhibits podset (Srinivasan et al. 1998; Berger et al. 2004). This essentially means an increase in thermal time target for the crop during the reproductive period. As cultivar parameters in the original model did not have the ability to account for this effect, a modification in cultivar parameters in the model was considered necessary to account for the period during which pod set and filling will not occur (or will occur over a longer period), due to temperatures/ photoperiods being unfavourable for podset. This modification increases the thermal time target for periods between flower initiation and flowering, and between flowering and grain-filling under progressively shorter days to account for periods of low temperatures/photoperiods inhibiting pod filling, and has been calibrated against the observed data. For model validation, data on time to flowering, maturity and grain yield from three trials conducted throughout Australia were used, including studies by Thomas and Fukai (1995), Berger et al. (2004) and McCosker and Douglas (unpublished data). These trials covered sites at Emerald (23˚310 S and 148˚100 E), Biloela (24˚80 S and 150˚200 E), Roma (26˚340 S and 148˚470 E), Redland Bay (27˚370 S and 153˚190 E), Billa Billa (28˚120 S and 150˚210 E), Warwick (28˚130 S and 152˚60 E) in Queesland, Tamworth (31˚50 S and 150˚500 E) in New South Wales, Merredin (31˚480 S and 118˚160 E) in Western Australia, Minnipa (32˚500 S and 135˚100 E) in South Australia, and Walpeup (35˚70 S and 142˚E) in Victoria (Fig. 1). The agronomic details used for simulation are given in Table 1. These trials were conducted either under completely rainfed (Berger et al. 2004) conditions, or with full irrigation (Thomas and Fukai 1995), or irrigation was only given to establish a crop (C. Douglas, QDPI&F, pers. comm.). Soil depth data gathered from the literature was specified in the soil parameter file and starting soil water was set at 90 days before sowing to allow Materials and methods This study involved two stages; firstly to validate the APSIM chickpea model across a range of locations in Australia included in the homoclime analysis; and secondly to apply this model to generate outputs for several locations in Australia and India to identify chickpea homoclimes in both the countries. Model validation All simulations were conducted using the APSIM chickpea module (version 4) (Robertson et al. 2002) incorporating the chickpea model. As the APSIM chickpea model has been designed to simulate a uniform block of land and does not account for the confounding factors of pests, diseases and variable crop stands, its outputs represent the situations free from these confounding factors. The model simulates crop development, growth, and grain yield in response to inputs of Emerald Biloela Roma Billa Billa Merredin –23.2 S Redland Bay Warwick Tamworth Minnipa Walpeup Fig. 1. Australian locations used in validating the APSIM chickpea model. 262 Australian Journal of Agricultural Research Y. Chauhan et al. Table 1. Agronomic details used for validation of the APSIM chickpea model at different locations Location Date of sowing Cultivar Plants/m2 PAWC (mm) Starting water Water status Reference Billa Billa Billa Billa Biloela Biloela Emerald Emerald Warwick Warwick Roma Roma Merredin Merredin Merredin Merredin Minnepa Minnepa Minnepa Minnepa Walpepup Walpepup Walpepup Walpepup Tamworth Tamworth Warwick Warwick Warwick Warwick Redland Redland Redland 20/05/2003 1/06/2004 26/05/2003 26/05/2004 20/05/2003 21/05/2004 3/06/2003 7/06/2004 14/05/2003 19/05/2004 8/06/1999 8/06/1999 16/06/2000 16/06/2000 1/06/1999 1/06/1999 5/06/2000 5/06/2000 31/05/1999 31/05/1999 12/05/2000 12/05/2000 14/06/2000 14/06/2000 31/05/1999 31/05/1999 5/06/2000 5/06/2000 2/04/1990 10/07/1990 24/07/1991 Amethyst Amethyst Amethyst Amethyst Amethyst Amethyst Amethyst Amethyst Amethyst Amethyst Amethyst Barwon Amethyst Barwon Amethyst Barwon Amethyst Barwon Amethyst Barwon Amethyst Barwon Amethyst Barwon Amethyst Barwon Amethyst Barwon Amethyst Amethyst Warwick 30 30 28 30 34 17 30 30 30 30 53 53 28 28 45 45 41 41 40 40 27 27 39 39 45 45 37 37 35 35 35 190 190 240 240 150 150 240 240 190 190 190 190 190 190 190 190 190 190 190 190 190 190 190 190 190 190 190 190 140 140 240 2/3rd Full Full Full Full Full 2/3rd 2/3rd Full 2/3rd Full Full Full Full 1/3rd 1/3rd 1/3rd 1/3rd Full Full Full Full 2/3rd 2/3rd Full Full Full Full Full Full Full R R R R* R* R* R R R R R R R R R R R R R R R R R R R R R R I I I A A A A A A A A A A B B B B B B B B B B B B B B B B B B C C C PAWC, Plant-available water-holding capacity. Water status: R, rainfed; I, Irrigated; *irrigated to full profile at planting; A, C. Douglas, QDPI&F, pers. comm.; B, Berger et al. (2004); C, Thomas and Fukai (1995). pre-sowing rainfall to be accounted for. Sowing date, plant population and the amount of applied irrigation were specified in the manager module. Weather data were downloaded from the ‘SILO’ weather site (SILO 2005). Where available, rainfall and temperature data collected from the trial sites’ were patched onto the weather data obtained from ‘SILO’. Separate runs were made using both original and modified cultivar parameters. For the Indian locations, maturity data from the chickpea trials by the All India Coordinated Program conducted at Coimbatore, Hisar, Jabalpur, Ludhiana, Patancheru and Rahuri during 1986–2002 were averaged. These were regressed against the mean time to maturity simulated by the APSIM chickpea model. Homoclime analysis Homoclime analysis was conducted on 67 locations in Australia, including those used for validating the model, and 24 locations in India (Table 2). The Indian locations represented a range of chickpea growing regions; however there were only a limited number of sites, especially in central and northern India that could be included, due to a paucity of quality climatic data. Climatic data of a larger number of locations were available from Australia to adequately cover the diversity of chickpea growing environments. Daily climatic databases of Australian locations were obtained from the Queensland Department of Natural Resources (SILO 2005), and for India from the National Ocean and Atmospheric Administration in the USA, and the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) in India. Weather data of all the Australian locations covered the period from 1957 to 2005, however for the Indian locations time spans varied from 8 to 100 years. The APSIM chickpea model was run for three soils of 100, 150 and 190 mm plant available water holding capacities (PAWC). Since chickpea grain yield is known to vary linearly in response to water supply (Johansen et al. 1994; Zhang et al. 2000) these three soils were expected to cover a range of crop responses to variation in PAWC. For each location, a plant density of 35 plants/m2 was used. The two cultivars chosen for this analysis represented the extremes of slow and quick maturity types that have so far been parameterized. In order to reduce the complexity that could arise due to different crop rotations practiced in India, the winter chickpea-summer fallow system, which is the most common chickpea cropping system across the two countries, was simulated. Sowings were simulated Identifying chickpea homoclimes Australian Journal of Agricultural Research 263 Table 2. Australian and Indian locations and their latitudes (degree decimals) and longitudes (degree decimals) in different states included in the homoclime analysis Location Lat. (S) Australian locations Long. Location A Queensland Banana Bendee Billa billa Biloela Bundaberg Capella Cecil Plains Clermont Condamine Dalby Dysart Emerald Gindie Gogango Goondiwindi Jimbour Moonie Moura Orion Roma St George Surat Thallon Theodore Warwick New South Wales Coleambally Coonamble Griffith Hay Moree Narrabri Narromine Tamworth W. Wagga Walgett A 24.5 23.8 28.2 24.4 24.9 23.1 27.5 22.8 27.7 27.2 22.6 23.5 23.7 23.7 28.5 27 27.6 24.6 24.3 26.6 28 27.2 28.6 25 28.2 150.1 148.4 150.4 150.5 152.4 148 151.2 147.6 151.3 151.3 148.4 148.2 148.1 150.1 150.3 151.2 150.4 150 148.4 148.8 148.6 149.1 148.9 150.1 152.1 34.8 31 34.3 34.5 29.5 30.3 32.2 31.1 35.2 30 145.9 148.4 146.1 144.9 149.8 149.8 148.2 150.9 147.5 148.1 Victoria Bendigo Beulah Kalkee Mildura Shepparton Swan Hill Walpeup South Australia Eyre Peninsula Gawler Gladstone Hart Horsham Minnipa Nonning Port Lincoln Rosedale Western Australia Bindi Bindi Cunderdin Esperance Geraldton Gnowangerup Kununurra Lake Grace Merredin Mullewa Northam Nyabing Pingaring Three Springs Varley West Miling Wyndham Lat. (S) Long. 36.8 35.9 36.3 34.2 36.4 35.3 35.1 144.3 142.4 142.1 142.1 145.4 143.6 142 33.6 34.6 33.3 33.5 36.7 32.8 32.5 34.7 34.6 135.9 138.7 138.4 138.3 142.1 135.2 136.5 135.9 138.8 30.6 31.7 33.6 28.8 33.9 15.8 33.1 31.5 28.5 31.6 33.5 32.8 29.5 32.8 30.4 15.5 116.4 117.3 121.8 114.7 118 128.7 118.5 118.2 115.5 116.7 118.2 118.6 115.8 119.5 116.3 128.1 Location Indian locations Lat. (N) Punjab Ludhiana Amritsar Haryana Hisar Delhi Madhya Pradesh Indore Jabalpur Andhra Pradesh Nandyal Patancheru Maharashtra Parbhani Solapur Mohol Jeur Rahuri Aurangabad Karnataka Banglore Annigeri Bellary Hagari Dharwad Raichur Bheemar Bijapur Gulbarga Tamil Nadu Coimbatore Long. 30.9 31.6 75.9 74.9 29.2 75.7 28.7 77.2 22.7 23.2 75.8 80 15.3 17.3 78.4 78.2 15.1 17.7 17.8 18.2 19.4 20 75.1 75.9 75.5 75.2 74.7 75.3 13 15.1 15.2 15.2 15.4 16.2 16.6 16.8 17.4 77.6 75.1 76.9 77.1 75.1 77.4 76.8 75.7 76.9 11.7 77.1 State names are in bold letters. to take place whenever the soil had accumulated 90 mm of extractable soil water (ESW) between 15-May and 14-June in Australia, and 15-Oct and 14-Nov in India. If this condition was not met until the last day of the respective sowing window in each country, then sowing was initiated with 30 mm irrigation on the following day, disregarding the ESW constraint. Pre-sowing irrigation is not uncommon in India where gaps in the timing of the monsoon withdrawal and chickpea sowing can be large leaving inadequate seedbed soil moisture. However, in Australia pre-sowing irrigation is generally rare; but it was included to capture the climatic effect of all seasons, as for India, in the event of insufficient seedbed moisture which would be equivalent to about a week’s moisture supply. A range of outputs was simulated. However, only time to flowering, maturity and grain yield, which are normally recorded in any breeding trial and were expected to capture the integrated effect of plant, soil and weather interactions, were used for homoclime analysis. The means of the annual outputs of these characters comprising 18 variables (3 soil types
2 cultivars
3 observations), were clustered using hierarchical complete linkage clustering (Wards method). The number of meaningful clusters was determined on the basis of scree-plot pattern (Cattell 1966). To determine climatic and biological characteristics of different homoclimes, the averages of pre-season rain, preanthesis rain, post-anthesis rain, lowest minimum temperature, minimum temperature at germination, minimum and maximum temperatures at 1st pod set, moisture availability index (water supply/water demand) for the last 60 days, times to first flowering, first pod set and maturity, grain yield and biomass were compared among different clusters. The differences among clusters were analysed using GENSTAT’s (version 9.2, Lawes Agricultural Trust, Rothamsted Experimental Station, UK) unbalanced analysis of variance procedure with clusters as 264 Australian Journal of Agricultural Research Y. Chauhan et al. factors and values for different soil types as replications. This computed maximum, minimum and average least significant difference at 5% probability level, but only average values were retained for comparison. account for the delay in pod set caused by the crop’s exposure to chilling temperatures. The chilling sensitivity of chickpea was better captured by the modified model without affecting yield simulation. This was evident by improvement in the prediction of maturity (R2 = 0.65) because it increased thermal time targets of the post-flower-initiation phases when mean ambient temperature was <15˚C (Fig. 2). With the modified model, simulated maturity tended to be slightly more than the observed maturity in some locations, which to some extent could be related to subjectivity introduced in assessing maturity based on the external pod or crop colour by different individuals across sites. Using the modified cultivar parameters, the model was still able to predict grain yield with a similar level of accuracy as achieved by the original parameters; hence these parameters were retained for subsequent homoclime analysis. The model, with the modified cultivar parameters accounted for 92% of the total variation in time to maturity for six Indian locations (Ludhiana, Hisar, Jabalpur, Rahuri, Patancheru, and Coimbatore) for which mean maturity data were available from various variety evaluation trials (data not shown). Results and discussion Model validation for the target environments The APSIM chickpea model was reported to predict flowering time and grain yield reasonably well, but was not able to predict maturity accurately when ambient temperature was <15˚C during the reproductive phase (www.apsim.info/apsim/ publish/apsim/chickpea). Indeed in the present study, the ability of the model to predict time to maturity with the original cultivar parameters was poor (R2 = 0.49), as it was unable to simulate delayed maturity in longer season environments (Fig. 2). A conspicuous outlier was that of Redland Bay, which was a long season site with mean ambient temperatures frequently falling to <15˚C during the reproductive phase. The unmodified APSIM model did not simulate maturity very well at this site because it did not Original model Simulated days to maturity 220 Modified model y = 0.5638x + 76.639 R 2 = 0.4926 200 y = 0.6306x + 68.671 R 2 = 0.6505 180 160 Redland Bay 140 120 100 80 80 100 120 140 160 180 200 220 80 100 120 140 160 180 200 220 Observed days to maturity Simulated grain yield (kg/ha) 4000 y = 1.0285x + 19.551 R 2 = 0.788 y = 1.0233x + 8.37.93 R 2 = 0.7965 Warwick Warwick 3000 2000 Emerald Emerald 1000 0 0 1000 2000 3000 4000 0 1000 2000 3000 4000 Observed grain yield (kg/ha) Fig. 2. Observed and simulated time to maturity and grain yield of chickpea in sowings across Australia (see Table 1 for details) during 1999–2005 using original (left) and modified (right) cultivar parameters. A few outliers in maturity and yield are also shown. Identifying chickpea homoclimes Australian Journal of Agricultural Research Homoclime analysis The cluster analysis of means of the APSIM simulated flowering and maturity and grain yield separated 91 locations of the two countries into 6 clusters at a 90% level of similarity (Fig. 3). The membership of clusters 1, 2, and 6 was relatively large and that of clusters 4 and 5 the smallest. The Australian locations were present in all the six clusters whereas the Indian locations were only present in clusters 1, 2 and 6. This suggests that Australian chickpea growing environments are relatively more diverse than Indian environments. The inclusion of locations of both countries in three clusters suggests that some of the Australian locations should have similar growing environments as experienced by chickpea in some locations in India. Such locations can be considered as homoclimes because their environments produced similar outcomes of flowering, maturity, as well as grain yield using the APSIM model. Malhotra and Singh (1991) have used flowering and grain yield from two international yield trials to characterise chickpea growing environments. However, in many environments, flowering commences whenever its thermal time target is reached. However, in reality pod setting is delayed due to cooler temperatures which may affect maturity. To account for this variation in the period between commencement of flowering and pod-set, the inclusion of maturity was considered appropriate in the present study. The memberships of individual clusters followed a systematic geographic pattern in both countries (Fig. 4). In India, clusters 1 and 2 covered 4 subtropical northern Indian locations, and cluster 6 all the rest of the locations in the tropics (Fig. 4). In Australia, distribution of locations also appeared to be influenced by different agro-ecological conditions. For example in cluster 1, locations were found mainly in the semi-arid tropical, subtropical and temperate slopes and plains; cluster 2 locations were mainly in the subtropical and tropical slopes and plains; cluster 3 mainly in the temperate highlands on the eastern coast; cluster 4 on the wet-subtropical eastern and western coast; cluster 5 on wet southern temperate coast; and, cluster 6 in the north-west wet/dry tropics (Fig. 4). The spread of clusters 2 and 3 0.3 0.5 0.6 0.7 Level of similarity 0.4 0.8 0.9 1.0 Cluster Cluster 1 Ludhiana Amritsar Kalkee Beulah Swanhill Walpeup Coleambally Hay Griffith Mildura Gnowangerup Eyre Peninsula Nyabing Hart Gladstone Lake Grace Variey Pingaring Nonning Norham Merredin Coonamble Narrabri Cluster 2 Walgett Moree Thallon Goondiwindi Billa billa Condamine Moonie Cecil Plains Dalby Surat Jimbour Roma Delhi Hisar Minnipa Cunderdin Bindi Bindi West Milling Three Springs Mullewa St George Theodore 265 Moura Banana Biloela Orion Bendee Gindie Gogango Emerald Capella Clermont Dysart Cluster 3 Bendigo Horsham Shepparton W. Wagga Gawler Rosedale Narromine Tamworth Warwick Cluster 4 Geraldton Bundaberg Cluster 5 Port Lincoln Esperance Cluster 6 Banglore Annigeri Parbhani Bellary Hagari Nandyal Dharwad Raichur Bheemar. Bijapur Coimbatore Patancheru Gulbarga Solapur Mohol Jeur Rahuri Aurangabad Indore Jabalpur Kununurra Wyndham Fig. 3. Hierarchical cluster analysis of 67 locations in Australia and 24 locations in India based on Ward’s method, for flowering, maturity and grain yield. The Indian locations are in italics. The clustering was done at 90% level of similarity. 266 Australian Journal of Agricultural Research Y. Chauhan et al. Cluster 1 2 6 23.2 N Cluster 1 2 3 4 5 6 2 1 1 3 2 5 5 –23.2 S 6 11 7 11 6 10 10 8 9 7 9 10 10 Agro-ecological regions 1 2 3 4 5 6 7 8 9 10 11 8 9 8 North-west wet/dry tropics Northern wet/dry tropics North-east wet/dry tropics Wet tropical coasts Semi-arid tropical/subtropical plains Subtropical slopes and plains Wet subtropical coast Wet temperate coasts Temperate highlands Temperate slopes and plains Arid interior 8 9 8 Fig. 4. Geographic spread of different clusters across India and Australia. See Fig. 3 for the names of locations within each cluster. Locations within and across the two countries with a similar cluster symbol are homoclime. covered the widest range of subtropical (northern), temperate and the Mediterranean type environments in Australia where much of the current chickpea production occurs. The clusters differed in biological and physical characteristics (Table 3). Biomass was at its maximum in cluster 5 and lowest in cluster 6 locations. The crop flowered and matured the earliest in cluster 6 and the latest in cluster 3, with this difference being about 2-fold. The period of ineffective flowering of 34 days (difference in anthesis and pod-set) was the longest in cluster 3 locations which spread in the cool sub-tropical and temperate highlands on the eastern coast of Australia. Grain yields were the lowest for cluster 6 and were ~27% of cluster 5, which had the highest grain yield. The mean of cluster 1 grain yield was ~40% higher than the grain yield mean of cluster 2. Regan et al. (2006) reported that grain yield of crops grown on northern latitude locations in south- Western Australia that belonged to cluster 2 in our study (see Fig. 4) were generally higher than those on cooler southern latitudes that belonged to cluster 1, especially in autumn sowings. This was not supported by the model simulations in our study although duration of the crop simulated was indeed longer in cluster 1 locations. As noted earlier, the model does not account for local specific constraints related to pests and diseases or lodging and hence there could be some discrepancy in the observed and simulated performance of the crop for these reasons. On the basis of climatic analysis Berger and Turner (2007) grouped the world’s chickpea growing region into four common rainfall and temperature categories namely, Mediterranean type – cool or warm climate; and summer dominant rainfall – cool or warm climate. According to their analysis Indian environments largely fall in to the summer dominant rainfallcool and warm environments and Australian environments into both summer-dominant and the Mediterranean type – cool and warm environments. Regan et al. (2006) recently reported that within the narrow range of the Mediterranean type environments of south-western Australia, chickpea growing regions could be divided into warm northern, intermediate central and cooler southern region. Our analysis confirms the separation of the northern locations from the southern region by placing them in different clusters. The northern locations within the Mediterranean type environments were in the same cluster as the central Queensland locations. The analysis grouped two locations such as Merredin and Northam in cluster 1, but a nearby location at Cunderdin in cluster 2. Regan et al. (2006) had placed all the three locations in a separate ‘central’ group which in their study behaved somewhat similar to northern locations in early sowings, and to southern locations for the late sowings. As different sowings were not simulated in our study, this was not assessed. The locations in subtropical Queensland and New South Wales with a summer dominant rainfall pattern were in cluster 1, 2 and 3. Also, the locations with warmer central Queensland were in cluster 2 and those of cooler south-eastern Queensland in clusters 1 and 3. A somewhat surprising result of this analysis, as noted above, was that several locations in central Queensland homoclimed (present in the same cluster 2) with locations in the Mediterranean type environments of south-western Australia and Minnipa in South Australia and in northern India (Figs 3 and 4). It is in this region of Queensland that the chickpea area is currently expanding. This suggests that a cultivar adapted to one of these homoclimes could find adaptation in the central Queensland regions. Supportive evidence for this hypothesis came from a recent release of the variety ‘Moti’ in central Queensland, which was originally bred and selected in south-western Australia (Berger et al. 2004). For locations of this cluster, terminal drought may be a major issue as suggested by the low moisture availability index during the reproductive period and hence development of early maturing cultivars that can escape terminal drought may be advantageous. Two Australian locations in cluster 3, Warwick and Tamworth, where much of the chickpea breeding work is currently concentrated, did not cluster with any of the locations from Western Australia, or any other locations in Identifying chickpea homoclimes Australian Journal of Agricultural Research 267 Table 3. Means of biological and physical characteristics in different clusters Characteristics Clusters #1 #2 #3 #4 #5 #6 l.s.d. Biological Anthesis Biomass (t/ha) Grain yield (t/ha) Days to first pod-set Days to maturity 100 4 1.25 130 170 80 3.82 0.89 108 146 112 5.01 1.9 144 187 77 5.86 1.86 105 145 95 6.66 2.66 127 172 54 2.83 0.74 78 112 0.9 0.101 0.051 0.9 0.9 Physical Germination Min T. (˚C) Min. Crop Temp. (˚C) Max. T at podding (˚C) Min. T at podding (˚C) Average Max. T (˚C) Average Min. T (˚C) Pre-season rain (mm) Post-anthesis rain (mm) Pre-anthesis rain (mm) Moisture availability index 7.2 –1.7 25.2 10.2 21.1 7.4 119 83 121 0.54 11.2 0.2 26.4 10.7 24.6 9.4 146 54 69 0.43 6.2 –1.8 23.4 9.6 18.7 6.8 127 112 188 0.68 13.2 4 23.2 11.2 22.5 11.2 219 73 179 0.59 8.9 3.2 21.2 10.2 18.7 9 133 97 209 0.72 19.5 9.5 29.6 14.9 30 16.1 338 17 75 0.56 0.22 0.15 0.29 0.28 0.11 0.11 9.6 3.5 5 0.014 l.s.d. = average least significant difference at 5% probability. south-eastern or central Queensland. Based on the analysis of genotype
environment interaction for grain yield in chickpea trials conducted in Australia, Berger et al. (2004) suggested that Tamworth may not be a representative site for developing cultivars that are better adapted to other environments in Western Australia and Queensland. Our homoclime grouping therefore supports their suggestion and further indicates that Warwick may also not be a representative site for selecting material adaptation in many other Queensland locations in cluster 1 and 2, as it does not homoclime with them. Implications for chickpea improvement The earlier work on chickpea has recognised that the grouping of homogenous environments can minimize genotype
environment interactions in chickpea (Malhotra and Singh 1991). Hence a genotype developed at one location can be expected to perform well at other locations within the homoclime group. However, the usefulness of this type of homoclime analysis based on simulated outputs to breeders/ agronomists will be more easily apparent if it can be shown that germplasm adapted to one homoclime location will indeed do well in other homoclime locations. A few examples of past releases tend to support this indirectly. For example, ‘Tyson’ which originated from Ludhiana in N. India (Beech and Brinsmead 1980) which fell in cluster 1 environment, was released in cluster 1 – south-east Queensland locations in 1978. Within India, several lines developed in central India, e.g. JG 62 and ICC 4958 have been found to be better adapted to southern India (Saxena 2003; Berger et al. 2006). Berger et al. (2006) reported that several chickpea cultivars that did well in locations in both central and southern India were in the same cluster, suggesting similar adaptation strategies, e.g. high harvest index and early flowering employed by cultivars adapted to these regions. Some chickpea cultivars could be suitable in more than one homoclime, if they have been bred in two different homoclime environments. For example, two recent G
E studies conducted in Australia and India (Berger et al. 2004, 2006) suggested that the chickpea line ICCV 10, which was bred through a shuttle breeding and selection program carried out at cluster 2 (Hisar) and cluster 6 (Patancheru) locations in India by ICRISAT, was found to have wider adaptation in locations of clusters 1 and 2 in Australia, as well as in cluster 6 in India. A similar strategy, in addition to involving more diverse parents, was adopted by the the Indian Agricultural Research Institute (IARI) New Delhi breeding program, with breeding and selection being completed in Delhi in the north, and Dharwad in south India (Berger et al. 2006). Cultivar BG 362 developed by this program was found to be high yielding in Australian environments included in clusters 1 and 2 (Berger et al. 2004). Several cultivars developed with a similar approach to this program have also been found to be widely adapted in India (Berger et al. 2006). Although it is yet to be confirmed, wider adaptation from the shuttle breeding approach seems arise from greater photoperiod sensitivity being incorporated through selection at higher latitudes/cooler environments. This probably enables these cultivars to achieve greater source and sink potential, and early flowering, as well as other drought avoidance characteristics through selection at lower latitudes in cluster 6 environments. In contrast, the germplasm originating only from the breeding programs within a homoclime e.g. cluster 6 locations of central and peninsular India, tends to be more specifically adapted to this region, as it seems to encourage development of traits that are relevant to adaptation in that environment (Berger et al. 2006). Such germplasm tends to perform poorly in cooler environments, which is probably associated with a lack of required photoperiod sensitivity. In Australia, the national breeding program for chickpea is located at Tamworth, with a node in Queensland located at 268 Australian Journal of Agricultural Research Warwick, which both incidentally occur in cluster 3. Since these locations do not homoclime with locations in major chickpea growing areas in central Queensland, south-eastern Queensland, and Western Australia, materials bred at these locations may not find adequate adaptation in warmer or short growing environments of central and southern Queensland, as well as in WA. A shuttle breeding approach similar to that was adopted by ICRISAT, which was responsible for the breeding of ICCV 10, and that of BG lines developed by IARI, is likely to lead to development of cultivars that are more widely adapted to other homoclimes. Conclusions The APSIM chickpea model, with revised crop parameters, has been able to improve prediction of time to maturity without adversely affecting prediction of grain yield. This provided us greater confidence in conducting the homoclime analysis of different chickpea growing environments of India and Australia. Identification of homoclimes reported in this study seems to have arisen due to a better integration of interactions between diurnal and seasonal changes in climate, plant and soil attributes achieved through the use of the APSIM chickpea model. This would have been difficult to visualise by simply comparing climatic averages between locations. It is recognised that all locations, especially in India where chickpea is currently being grown, could not be included in this analysis due to a paucity of climatic data. Those locations could be part of the cluster in close proximity to them. Availability of daily climatic data would constrain wider applicability of this approach, but the potential benefits demonstrated in this study should encourage creation of such databases for most environments. The homoclime analysis used in this study, and supported by some examples from previously published studies, suggests this approach may be useful for improving the efficiency of national and international germplasm exchanges, introduction of newly evolved high yielding cultivars in new environments and rationalizing breeding and testing sites. The analysis has generally been able to confirm observations made from the recent extensive field experimentation that analysed G
E interactions in chickpea (Berger et al. 2004, 2006). Further indepth analysis of published or unpublished datasets, or new experiments, would be useful to further validate the conclusions of this study and confirm the value of the methodology used. Use of only two genotypes in the study may be viewed as a limitation of this study. However, since the main objective of the study was to identify environmental similarity rather than evaluate cultivar performance across several environments, use of a wider range of cultivars was considered unnecessary. In the future, when more promising genotypes are parameterized for the APSIM chickpea model, it should become possible to evaluate their performance across different homoclimes using this approach. Acknowledgments The financial support received from the Grains Research and Development Corporation (GRDC) under project DAQ533 for this study is gratefully acknowledged. Authors are also thankful to Andrew Robson for assisting with the GIS work. Thanks to Drs Pooran Gaur and CLL Gowda of ICRISAT for supplying climatic and phenology data of trials conducted in India. Y. Chauhan et al. References Beech DF, Brinsmead RB (1980) Tyson: a chickpea (Cicer arietinum) cultivar for grain production. Journal of Australian Institute of Agricultural Science 46, 127–129. Berger JD, Ali M, Basu PS, Chaudhary BD, Chaturvedi SK et al. (2006) Genotype by environment studies demonstrate the critical role of phenology in adaptation of chickpea (Cicer arietinum L.) to high and low yielding environments of India. Field Crops Research 98, 230–244. doi: 10.1016/j.fcr.2006.02.007 Berger JD, Turner NC (2007) The ecology of chickpea: evolution, distribution, stresses and adaptation from an agro-climatic perspective. In ‘Chickpea breeding and management’. (Eds S Yadav, R Redden, W Chen, B Sharma) pp. 47–71. (CAB International: Wallingford, UK) Berger JD, Turner NC, Siddique KHM, Knights EJ, Brinsmead RB, Mock I, Edmondson C, Khan TN (2004) Genotype by environment studies across Australia reveal the importance of phenology for chickpea (Cicer arietinum L.) improvement. Australian Journal of Agricultural Research 55, 1071–1084. doi: 10.1071/AR04104 Booth TH, Nix HA, Hutchinson MF (1987) Grid matching: a new method for homoclime analysis. Agricultural and Forest Meteorology 39, 241–255. doi: 10.1016/0168-1923(87)90041-4 Cattell RB (1966) The scree test for the number of factors. Multivariate Behavioral Research 1, 245–276. doi: 10.1207/s15327906mbr0102_10 Chapman SC, Cooper M, Hammer GL (2002) Using crop simulation to generate genotype by environment interactions effects for sorghum in water-limited environments. Australian Journal of Agricultural Research 53, 379–389. doi: 10.1071/AR01070 Chapman SC, Cooper M, Podlich D, Hammer GL (2003) Evaluating plant breeding strategies by simulating gene action and dryland environment effects. Agronomy Journal 95, 99–113. Clarke HJ, Siddique KHM (2004) Response of chickpea genotypes to low temperature stress during reproductive development. Field Crops Research 90, 323–334. doi: 10.1016/j.fcr.2004.04.001 Holt JS, Boose AB (2000) Potential for spread of Abutilon theopharsti in California. Weed Science 48, 43–52. doi: 10.1614/0043-1745(2000)048 [0043:PFSOAT]2.0.CO;2 Johansen C, Singh DN, Krishnamurthy L, Saxena NP, Sethi SC (1994) Genotypic variation in moisture response of chickpea grown under linesource sprinklers in a semi-arid tropical environment. Field Crops Research 37, 103–112. doi: 10.1016/0378-4290(94)90038-8 Malhotra RS, Singh KB (1991) Classification of chickpea growing environments to control genotype
environment interaction. Euphytica 58, 5–12. doi: 10.1007/BF00035334 McCown RL, Hammer GL, Hargreaves JNG, Holzworth DP, Freebairn DM (1996) APSIM: a novel software system for model development, model testing, and simulation in agricultural systems research. Agricultural Systems 50, 255–271. doi: 10.1016/0308-521X(94)00055-V Regan KL, Siddique KHM, Brandon NJ, Seymour M, Loss SP (2006) Response of chickpea (Cicer arietinum L.) varieties to time of sowing in Mediterranean-type environments of south-western Australia. Australian Journal of Experimental Agriculture 46, 395–404. doi: 10.1071/EA05091 Roberts EH, Hadley P, Summerfield RJ (1985) Effects of temperature and photoperiod on flowering in chickpea (Cicer arietinum L.). Annals of Botany 55, 881–892. Robertson MJ, Carberry PS, Huth NI, Turpin JE, Probert ME, Poulton PL, Bell M, Wright GC, Yeates SJ, Brinsmead RB (2002) Simulation of growth and development of diverse legume species in APSIM. Australian Journal of Agricultural Research 53, 429–446. doi: 10.1071/AR01106 Russell JS, Moore AW (1976) Classification of climate by pattern analysis with Australiasian and Southern Africa data as an example. Agricultural Systems 16, 46–69. Identifying chickpea homoclimes Australian Journal of Agricultural Research Russell JS, Webb HR (1976) Climatic ranges of grasses and legumes used in pastures: results of a survey conducted at the 11th International Grassland Congress. Journal of the Australian Institute of Agricultural Science 42, 156–166. Saxena NP (2003) Management of drought – a holistic approach. In ‘Management of agricultural drought – agronomic and genetic options’. (Ed. NP Saxena) pp. 103–122. (Oxford & IBH Publishing: New Delhi) Saxena NP, Saxena MC, Johansen C, Virmani SM, Harris H (1996) ‘Adaptation of chickpea in the West Asia and North Africa Region’. (ICRISAT: Hyderabad, India and ICARDA: Aleppo, Syria) SILO (2005) Enhanced meteorological data [Online]. Available at: www. nrm.qld.gov.au/silo (verified 20 Dec 2005) Singh P (1991) Influence of water-deficits on phenology, growth and dry matter allocation in chickpea (Cicer arietinum). Field Crops Research 28, 1–15. doi: 10.1016/0378-4290(91)90070-C Smart R (2003) Portuguese homoclimes in Australia. Australian & New Zealand Wine Industry Journal 18, 48–50. 269 Srinivasan A, Johansen C, Saxena NP (1998) Cold tolerance during early reproductive growth of chickpea (Cicer arietinum L.): characterization of stress and genetic variation in pod set. Field Crops Research 57, 181–193. doi: 10.1016/S0378-4290(97)00118-4 Thomas , Fukai S (1995) Growth and yield response of barley and chickpea to water stress under three environments in southeast Queensland. I. Light interception, crop growth and grain yield. Australian Journal of Agricultural Research 46, 17–33. doi: 10.1071/AR9950017 Zhang H, Pala M, Oweis T, Harris H (2000) Water use and water-use efficiency of chickpea and lentil in a Mediterranean environment. Australian Journal of Agricultural Research 51, 295–304. doi: 10.1071/AR99059 Manuscript received 25 May 2007, accepted 8 November 2007 http://www.publish.csiro.au/journals/ajar