The constantly changing crop growing conditions necessitate the continuous development of crop varieties that guarantee farmers decent yields (Allard, 1999; Singh, 2006). Such changes which have come in the form of unpredictable rainfall patterns, extreme temperatures (Mba et al., 2012) and loss of soil fertility render previously well established varieties unfit to be grown competitively. The solution to such a problem is development of new varieties that fit the prevailing management and climatic conditions (Lane and Jarvis, 2007). New varieties require testing in growing environments where the crop is mostly grown to study their performance and suitability in specific growing conditions (Farshadfar et al., 2013). Variety testing in multi-location environments is done for at least two seasons in order to establish their stability and adaptation to wide growing conditions. In Zimbabwe, the tobacco growing region is divided into three namely the fast, the slow and the medium growing regions. Before new varieties are officially released for cultivation by farmers, they are tested in each of these areas. The objective of the testing would be to identify the best performing varieties in each of the growing environments for farmer recommendations. Environments differ in their ability to support different genotypes and environments it is refereed to edaphic, climatic and management conditions done by the farmer depending on the growing environment in which the farmer produces a crop (Mohamed, 2013, Muniswamy et al., 2017). Years also differ in their ability to support a crop, the reason why they are tested as rain fed in multi-location trials (Gauch, 1988). So year and location combinations are referred to as the environment in this study. The objective of the study was to identify varieties which do best in each of the tested environments and also those which have had adaptation. This could make it easy for variety agronomists to recommend cultivars for specific growing environments.

Agronomic management of the trial
 
Seedlings for the experiment were raised using the conventional tobacco seedbed approach at all the sites each year. The six new tobacco varieties and a control were grown at three locations, Kutsaga Research Station near Harare, Firhill Farm in Trelawney and Oldonyo Farm in Tengwe (Karoi). Kutsaga Research Station is in the slow, Trelawney in the medium while, Tengwe is in the fast tobacco growing area. The experiment ran for three seasons from 2012-13 to 2014-15. At each location the experiment was set as a randomised complete block design (RCBD) with three replicates. Plants were planted on raised ridges. Each plot consisted of two rows spaced at 120 cm between rows and plants spaced at 56 cm within rows. Planting was done during the third week of October each year using the water planting method which involves planting a seedling in a hole filled with 5 litres of water. Basal fertiliser (N:P:K - 6:23:20) was applied in the ridge before planting and Ammonium Nitrate (AN) applied at a rate of 8 g per planting station 4 weeks after planting (WAP). Topping was done at 8 WAP while reaping started at 8 weeks and continued until all the 18 leaves were taken off the land for curing and graded using the TIMB classification guidelines of 1994 to separate saleable leaf from unsaleable leaf.
 
Data analysis
 
General across site analysis of variance was carried out on saleable yield in Genstat Version 18 and the GGE biplot in the same software was used to show stable varieties across the sites and seasons. 

Combined analysis of variance (ANOVA) for saleable yield
 
The combined analysis of variance for the seven varieties in nine location-year environments showed that all the main effects in the study were highly significantly different (P < 0.05) among each other (Table 1). The genotype main effects showed significant differences indicating differences in yield performance among the evaluated genotypes in different environments across years in which they were environments.  Such differences are expected since each genotype is a collection of a set of genes that respond to the environments where they are subjected differently.  The environment main effect also indicated significant differences (P< 0.05) showing that there were differences in the ability to support genotypes among nine environments evaluated. The main objective of conducting multi-environment trials (MET) is to generate information that identifies varieties that suit different locations and unique or specific cropping situations (Yadah et al., 2014; Gauch and Zobel, 1997). Such information helps agronomists recommend varieties that do best in specific environments. It also helps plant breeders rank varieties based on their performance in different growing conditions and be able to choose the best varieties to for eventual release to farmers for cultivation (Zobel et al., 1988). In MET analysis, it is common practice to treat environments as location - year combinations and go on to identify the best location year combination environment. This is because the across site analysis that considers the environment alone would have shown non-significant genotype by environment interactions. Location - year combinations are then treated as environments and varieties are identified that suit the best location-year combination environment. It is true that location-year combinations are different in scientific terms but the limitation with this approach is that it assumes that season characteristics are fixed and can be repeated, which is not the case.  Unless the analysis is based on covariance using weather data, such recommendations may not useful to either the breeder or the agronomists or farmer since it is very difficult to repeat a year or season that is exactly like what has been experienced as well as its effects on each crop growing enevirironment in the future and combine it with variety that is always fixed. For that reason, this paper does not identify mega-environments and identify “which –won – where” for the reason that the results of this study targets mainly the agronomists who recommend varieties to growers. Stability is what is important under this scenario and that is what is discussed mainly.
 
Variety performance across environments
 
Table 1 ANOVA showed that genotype main effects were significantly different (p < 0.05). Although the genotypes gave an above average performance in terms of cured leaf yield, there were some varieties that outperformed others across the locations and seasons evaluated. The genotype T76 outperformed all the other genotypes across all the years and seasons tested (Table 2).  At all sites T75 comes second followed by T74 and ETH03 is always coming last. The consistence of position among the varieties in terms of rankings across all the locations tested show that there are no crossover interactions for saleable yield between the locations and the genotypes.
 
Mean environment performance across years
 
The environment in which a genotype is grown affects the performance of that genotype. Some genotypes interact positively with environments while others interact negatively resulting in a poor performance of the variety. By environment it is referred to the whole set of growing conditions that may affect the growth and development of a genotype. This covers aspects like soil properties, weather pattern, pest and disease regime as well as other farmer practices that contribute towards the success or failure in performance of a variety. As shown on Table 3, environments differed in their ability to support the genotypes evaluated. Trelawney gave the highest performance of 3070kg/ha followed by Tengwe with 2398 kg/ha and lastly came Kutsaga with 2384 kg/ha. Trelawney gave the highest yields because of consistent and good field management by the farmer.  Tengwe usually gives high yields (TRB Annual Report, 2012) but mid season droughts affected the location consecutively resulting in low yields. Kutsaga gets low yields mainly because of variable field crop management from one season to another and too much cured leaf handling losses.
 
Yield stability of varieties across environments
 
The combination biplot shown on Fig 1 show the presence of genotype × environment interaction in flue cured tobacco as was reported by Shava et al. (2019) and the relative stability of all the varieties tested in all the environments.  The biplot is typified concentric rings and the variety that occupies the closest position to the centre of the set of rings is the most stable variety. From the plot, T76 is the most stable varieties among all of them evaluated.  T76 is a variety that has high resistance to alternaria leaf spot as well as angular leaf spot race 1. It has a high capacity to give leaves that have a high filling value. Because T76 has one of its parent’s bred for wide cultivation in the early 60’s (TRB Annual Report, 1964) it can be presumed that its broad adaptation is as a result of the contribution of this line AW3R.  Going outwards, it can be seen that T75 and ETH06 follow.  From the results, T74 is the least stable variety of them all followed by K RK26. The lack of stability of the variety T74 is understandable since it is a late maturing variety developed mainly for the fast growing areas where high temperatures favour its quick growth. The variety that has behaved unusually is K RK26 which is often characterised as have broad adaptation yet the results show otherwise.

From the results of this study, it can be concluded that locations, Kutsaga, Trelawney and Tengwe differ in their ability to support tobacco genotypes. Trelawney is the best environment for tobacco farming and Kutsaga and Tengwe are comparable in their ability to support tobacco farming.  There is indeed significant genotype by environment interactions for saleable yield in tobacco when the environment refers to location-year combinations.  From the study, T76 is the most stable variety and T74 is the least stable in terms of saleable yield. Unlike what is believed in the tobacco industry, K RK26 is not the most stable variety currently on the market. The most stable variety is T76 followed by T75 and ETH06.