Plant material selection and collection Preliminary, three major jackfruit growing areas, namely Gazipur, Narsingdi and Bandarban districts/regions were selected. Then members of the research team physically visited major jackfruit gardens and talked to the farmers about the good accessions based on their experiences. At the same time, observations on 36 qualitative characters viz., propagating material, tree vigor, trunk surface, crown shape/habit, tree growth habit, tree nature, branching density, branching pattern, leaflet/leaf blade shape, leaflet apex, leaf base shape, leaf color, fruiting season, extent of fruit drop, fruit bearing habit, fruit bearing, fruit bearing position, fruit shape, stalk attachment to fruit, fruit rind color, fruit surface, shape of spine, spine density, flake shape, flake texture, pulp taste, pulp flavor, pulp juiciness, pulp (fresh flake) color, vivipary, shape of seed, seed surface pattern, seed coat color, adherence of seed coat to kernel, disease infestation, pest infestation for each accession was recorded. Afterwards, based on morphological characterizations using the descriptors provided by the International Plant Genetic Resource Institute (IPGRI, 2000), 30, 25 and 10 accessions of jackfruit from Gazipur, Narsingdi and Bandarban districts, respectively, were collected during 2015-2016. At the time of collection, geographical location of the accession were recorded (Table 1). From these collected accessions, finally, 28 accessions from three districts were selected based on a number of criteria, like fruiting seasons, fruit colour, fruit shape, fruit number, fruit weight, flake shape, flake texture and brix value. Selected accessions were planted at the agroforestry research field of the Bangabandhu Sheikh Mujibur Rahman Agricultural University (24° 09´ N; 90° 26´ E) in the month of August, 2016 for conservation purpose. The size of the pit was 1 m × 1 m × 1 m (L × W × H). Each accession was replicated three times by maintaining 4 m × 4 m distance from the plant to plant in order to grow lower storey crops as an agroforestry system. The soils of the pit were prepared by mixing cowdung, sand and soil at the ratio of 1:0.5:2 (in weight basis). Furthermore, nitrogen, triple superphosphate and muriate of potash (250 g in each case) were applied to the soils in each pit and left for fifteen days. The plants after planting were maintained through intercultural operations like fertilizer and pesticide applications, manuring, irrigation and weeding as and when necessary. The observations on survival and growth characteristics of the plants were recorded periodically. Jackfruit shows a considerable range of variation in morpho-agronomic traits, therefore, to understand the extent of genetic diversity for morphological characters and to select superior types of jackfruit, the traits such as growth habit, canopy structure, leaf size, fruit shape, size, colour, fruit bearing (age and seasonality), maturity and other parameters were studied. Statistical analysis Analysis of genetic variation of the accessions was performed with the program SPSS (version 23.0) following the procedure described by Al-Hadi et al. (2017). Pearsons's coefficient was utilized to estimate the degree of correlation among the different characters of the trees. Coefficients higher than 0.5 were considered as linear associations representing natural variation patterns and to gain broad impressions, the extent of correlation was classified as strongly correlated (r = >0.5) and weakly to moderately correlated (r = 0.5). Heatmap was generated after normalizing the mean values by using MeV version 4.9.0 (http://mev.tm4.org/). A heatmap is a graphical representation of data where the individual values contained in a matrix are represented as colors. The heatmap and hierarchical clustering were performed to understand the accessions-variable relationship in the 28 jackfruit accessions. Quantitative values of 26 morpho-physiological features of these accessions were normalized for clustering. This was followed by performing the hierarchical clustering using the Spearman's rank correlation algorithm.