Location

The present study was carried out in southwestern Colombia, Department of Nariño, in four representative zones for the environments characterized for potato production in this region. These were chosen by edaphoclimatic characteristics, such as altitude, precipitation, temperature, and prevalent slope (^{Benavides et al., 2021}). The evaluations were carried out in farms located in the municipalities of Guaitarilla (HE1), Providencia (HE2), Pupiales (HE3), and Guachucal (HE4) located in each environment (Tab. 1). The chemical properties of the soils of the site where the evaluation was developed are described in table 2.

Plant material

The potato cultivars used were the ones with the highest planting frequency for this group in the producing area of Nariño (^{Benavides et al., 2021}): ‘Diacol Capiro’, ‘Pastusa Suprema’, ‘ICA Única’, ‘Superior’, ‘Roja Huila’, and ‘Parda Bilingüe’ that have been described by ^{Ñústez (2011)} and ^{Moreno (2000)}.

Experimental design

In each environment, experimental plots were established in a randomized complete block design with three replications. The treatments corresponded to three levels of fertilization (Tab. 3). The first was based on the interpretation of the nutrient contents in the soils and the needs of the potato crop (^{Colombia ICA, 1992}); the second and third levels corresponded to 50 and 75% of the requirements of a crop to produce 40 t ha^-1 (^{García and Pantoja, 1998}). Along with the evaluated levels, secondary and micronutrients were applied in doses per hectare of MgO 30 kg, S 15 kg, B 2.25 kg, Cu 0.6 kg, Mn 3 kg and Zn 4.8 kg. We established a sowing density for the experiment of 33,333 plants/ha. Each experimental plot occupied an area of 45 m² of which 27.6 m² were used for recording yield variables.

Variables evaluated

Each 15-d destructive sampling was carried out to record the dry matter during the productive cycle. The leaf area index (LAI) was estimated from the total leaf area and the area of soil covered by it. With this information, the growth indices were calculated as follow: net assimilation rate (NAR), crop growth rate (CGR), and relative growth rate (RGR) (^{Mohamed et al., 2017}; ^{Hunt, 1978}; ^{Pandey et al., 2017}). At the end of the cycle, the harvest index (HI) was calculated for each cultivar. In addition, we evaluated the yield per experimental unit, number and weight of tubers by size at harvest, according to the Colombian technical standard for potato classification NTC - 341 (^{Colombia ICONTEC, 1996}), where large and very large sizes are those of commercial interest.

Statistical analysis

We used the multivariate method of principal component discriminant analysis (DAPC) for the identification and evaluation of groups by cultivars and environments, using physiological indices and yield variables together. We performed an analysis of variance according to the experimental design and Tukey's test (P<0.05) for each cultivar, for comparison of averages of the yield of each component. The statistical program R version 4.0.4 was used (^{R Core Team, 2021}) and the packages Adegenet (^{Jombart, 2008}) for DAPC and ExpDes (^{Ferreira et al., 2013}) for analysis of variance.

Economic analysis

For economic determination of the effects of nutrition levels, the methodology of partial budget net benefit and marginal analysis was used (^{CIMMYT et al., 1988}; ^{Reyes, 2001}). An adjustment rate was applied to the experimental yields to represent the yield under commercial production conditions. The selling price used was the historical departmental average for each cultivar from 2018 to 2020 (^{Colombia Fedepapa, 2021}). The price of production before harvest was estimated from the difference of the unit costs of marketing and harvesting. The marginal rate of return (MARR) was determined as the ratio of gross field profit to net field profit. The minimum rate of return (MIRR) was established as the sum of the 2020 average of the fixed term deposit rate (FTD) (^{BANREP, 2020}), the expected inflation for Colombia in 2021 (^{BANREP, 2021}), and the expected return on equity reported in 2020 that considers a 3-year average (^{Corficolombiana, 2020}). For the selection of the best treatments, the MARR and MIRR rates were contrasted.

Cultivar responses to productive environments in Nariño

The physiological and yield responses of the cultivars generated groupings showing differences among the environments (Fig. 1A). This suggested that there were different growth, development, and production responses due to the particularities of each environment (Tab. 2). These results underline the differentiation of homogeneous production environments carried out for the department of Nariño (^{Benavides et al., 2021}). The variables that most contributed to the grouping in the DAPC for environments were mainly physiological and secondly productive.

Figure 1. DAPC biplot with groupings of six potato (Solanum tuberosum subsp. andigena) cultivars in four homogeneous production environments. The dispersion in the first two principal components is plotted, using in (A) the homogeneous environments (HE) as a priori groups and in (B) the cultivars. The biplot synthesizes information on physiological variables and yields components. Inertia ellipses of each homogeneous environment or cultivar are presented around the experimental units. 

The physiological indices that provided a greater contribution to the differentiation of behavior in environments and between cultivars were CRT and NAR. CRT is an index of dry matter production efficiency that allows quantifying the growth rate. Its contribution to the discrimination of environments is influenced by the environment where the crop develops (^{Mora-Aguilar et al., 2006}). Although, the CRT also allows the separation of behaviors between genotypes since it is a measure that integrates the physiological behavior of plants (^{Mora-Aguilar et al., 2006}; ^{Santos et al., 2010}), it was also important in the grouping of cultivars. On the other hand, NAR is an indicator of average photosynthetic efficiency, and depends on genetic factors, so it is useful for estimating the efficiency and physiological differences between potato cultivars (^{Borrego et al., 2000}; ^{Jerez et al., 2016}; ^{Santos et al., 2010}).

For ‘Roja Huila’ and ‘Parda Bilingüe’ cultivars, statistical differences were found between environments (Tab. 4), showing higher yields in environments HE2 and HE4, exceeding 31 and 37 t ha^-1 and with a better proportion of tubers of higher commercial value with respect to the other environments. In the DAPC these two cultivars formed the first separate group (Fig. 1B), given that the NAR and CRT indices were the highest (NAR between 0.194 and 0.249 g dm^-2 d^-1 and CRT between 0.141 and 0.148 g g^-1 d^-1) in the phase of highest growth and their lower yields compared to those evaluated. These cultivars are planted by producers in the department of Nariño (^{Benavides et al., 2021}) and are marketed in low volumes, mainly for regional fresh consumption. ‘Parda Bilingüe’ can be used for frying (^{Tinjacá and Rodríguez, 2015}). According to the results, these cultivars are an option for producers in the HE2 environment.

‘Diacol Capiro’, ‘Pastusa Suprema’ and ‘Superior’ in the DACP formed the second group of cultivars, according to the physiological responses in the environments (between 0.144 - 0.194 g dm^-2 d^-1 for NAR and between 0.103 - 0.116 g g^-1 day^-1 for CRT), in the phase of highest growth; and the yields achieved were higher than those of ‘Roja Huila’ and ‘Parda Bilingüe’ and lower than those of ‘ICA Única’ (Fig. 1B; Tab. 4). However, the yields of these three cultivars responded differently within the productive evaluated environments. ‘Diacol Capiro’ reached yields between 28 and 32 t ha^-1 in environments HE1, HE2 and HE3, like commercial yields ranging between 25 and 30 t ha^-1 (^{Moreno, 2000}). In HE4 it exceeded the potential yield reported by ^{Ñústez (2011)} of 40 t ha^-1. ‘Pastusa Suprema’, in the environments HE3 and HE4 exceeded 54 t ha^-1 and was statistically superior to the yields in the environments HE1 and HE2 that did not reach 46 t ha^-1. However, the results exceeded the yields under optimal production conditions that for ^{Ñústez (2011)} are around 45 t ha^-1. In the ‘Superior’ cultivar no statistical differences were observed in the four environments. Its yield was between 35 t ha^-1 and 44 t ha^-1, higher than the 32 t ha^-1 produced in Nariño in 2018 (^{Colombia Fedepapa, 2021}).

The responses of these cultivars could be attributed to the optimal ranges of adaptation of each cultivar. Thus, for ‘Diacol Capiro’ adequate production is reported at altitudes between 2,000 and 3,200 m a.s.l. (^{Moreno, 2000}; ^{Ñústez, 2011}). For ‘Pastusa Suprema’ adequate production is seen from 2,500 to 3,200 m a.s.l. (^{Ñústez, 2011}). For the ‘Superior’ cultivar a defined range is not reported in the literature; however, in this research its yields were adequate for the four environments. In addition to altitude, the observed differences may be due to the interaction of cultivars with the different edaphoclimatic factors of the environments in the producing area; these are temperature, precipitation, soil pH, and slopes. In the HE1 environment, temperatures range from 14°C to 17°C, precipitation up to 3,000 mm year^-1, moderately acid pH and slightly steep slopes (5-12%). In the HE4 environment the temperature range is 9-12°C, precipitation up to 2,800 mm year^-1, neutral or nearly neutral soils, and steep slopes (50-75%) (^{Benavides et al., 2021}).

The cultivar ‘ICA Única’ differed in the DACP (Fig. 1B), since it reached growth rates of 0.149 g dm^-2 d^-1 of NAR and 0.118 g g^-1 d^-1 of CRT and attained yields superior to other cultivars. This produced more than 49 t ha^-1 in the four environments evaluated, highlighting its behavior in HE3 and HE4 for exceeding 60 t ha^-1 (Tab. 4). One of the characteristics of this variety is the production of large tubers, mainly observed in the environments HE2, HE3 and HE4 that produced yields of large and very large tubers greater than 47.5 t ha^-1. The variation in yield between environments may be due to the adaptation of the variety to the environments HE3 and HE4, where yields were higher (from 30 to 45 t ha^-1) than the commercial ones reached at altitudes between 2,600 to 3,200 m a.s.l. (^{Moreno, 2000}). A commercial limitation of the cultivar is the tendency to greening of the skin (^{Ñústez, 2011}), so the light conditions for transport or storage must be considered. This is one of the reasons why this variety is less sown in Nariño compared with the cultivars of the second group.

Cultivar response to fertilization

The evalutated genotypes did not show an interaction between fertilization levels and environment, apart from ‘ICA Única’ in the number of tubers (Tab. 4) that allow a general discussion of the levels evaluated in all environments. The L3 level significantly increased the total yield in the cultivars ‘Pastusa Suprema’, ‘ICA Única’, ‘Roja Huila’, and ‘Parda Bilingüe’ with experimental values of 56, 63, 29 and 35 t ha^-1.

Levels L2 and L3, with respect to L1, increased the number or yield of tubers of higher commercial value in all cultivars, except for ‘Superior’ that did not respond to the increase in fertilization, although its yield was higher than that reported for Nariño in 2018 of 32.1 t ha^-1 (^{Colombia Fedepapa, 2021}). In ’Diacol Capiro’, the levels increased the number of tubers of higher commercial value from 4 to 6 tubers/plant. In ‘Pastusa Suprema’ the yield of this type of tubers reached 37 t ha^-1 and in ‘Parda Bilingüe’ 24 t ha^-1 with approximately 5 tubers/plant. Finally, ‘ICA Única’ had interaction between environments and levels with higher values in HE1 with level L2 (7.5 tubers/plant) and in environment HE3 with level L1 (7tubers/plant).

Yield components in ‘Diacol Capiro’, ‘Pastusa Suprema’, ‘ICA Única’, ‘Roja Huila’ and ‘Parda Bilingüe’ indicate that L2 and L3 levels outperformed L1. This response suggested that the L1 level did not supply enough nutrients for the cultivars to increase their yield. Meanwhile, L2 and L3 efficiently provided nitrogen (N), phosphorus (P₂O₅) and potassium (K₂O). The highest yields are found when applying the highest dose, 3,000 kg ha^-1 of 10-20-20 fertilizer to ‘Diacol Capiro’when compared to doses of 500, 1500, and 3000 (^{Rios et al.,2010}).

In general terms, soils in the production zone of Nariño belong to the order of Andisols (low availability of phosphorus, generated by the presence of allophane, as well as by the formation of insoluble compounds with iron and aluminum) (^{Bravo et al., 2016}; ^{García and Pantoja, 1998}). There is also reduced mineralization of organic matter (^{Sánchez and Rubiano, 2015}) and limited availability of potassium, due to fixation processes that can immobilize between 20 and 40% of the nutrient (^{Henriquez et al., 1994}). These soils have low pH that can be lower than 5.5 in some productive areas (^{Marcillo et al., 2021}) and can limit the availability of nutrients for the crop. Considering these characteristics in the soils of the producing area, the contributions provided by the L2 and L3 levels decreased the restrictions for a good supply of nutrients to the crop. This generated a positive effect on the production of most of the cultivars evaluated.In addition, the fractionation of nutrients and time of application must be considered according to the phenological state.

Individual application of each element generates yield increases with the highest doses, with up to 228 kg ha^-1 of nitrogen (^{Pinochet et al., 2018}),with 500 kg ha^-1 of P₂O₅ (^{Fernandes and Soratto, 2016}) or with doses up to 300 kg ha^-1 of K₂O (^{Zelelew et al., 2016}). Evaluations of compound applications also indicate that high doses improve potato yields (^{Rios et al., 2010}), however, this is true up to a certain limit, as too high amounts could have a negative effect on commercial tuber yields (^{Pinochet et al., 2018}).

Additionally, the proportion of each nutrient applied with reference to total fertilization should be considered, because it can generate a negative nutritional effect on the crop. The proportion of nitrogen supplied by L3 was less than 20% of the total fertilization, while for ^{Pinochet et al. (2018)} it was higher than 43% in the presence of 200 kg ha^-1 of P₂O₅ and 100 kg ha^-1 of K₂O, reporting yield decreases with doses higher than 228 kg ha^-1 of N. With L3 no such effects were observed when applying 262 kg ha^-1 of N, possibly due to the balance of proportions of this formulation.

Economic analysis

In the dominance analysis, the L1 level was used as a basis for the identification of fertilization alternatives. In the cultivars ‘Diacol Capiro’, ‘ICA Única’, ‘Pastusa Suprema’ and ‘Superior’ higher economic benefits and lower variable costs could be achieved with L2 for ’Parda Bilingüe’ with L3 and for ’Roja Huila’ with L1 (Tab. 5). From this information, in the partial budget analysis, favorable changes were found, with a MARR higher than MIRR (22.3%) in the fertilization level for ’Diacol Capiro’, ‘ICA Única’ and ’Superior’, while ‘Pastusa Suprema’ had a positive response, but it was not higher than MIRR. or ‘Parda Bilingüe’ a negative MARR was shown.

The L3 fertilization level had the highest yields in the cultivars ‘ICA Única’, ‘Pastusa Suprema’, ‘Parda Bilingüe’ and ‘Roja Huila’ while in ‘Superior’ no agronomic differences were found. In the case of ‘Diacol Capiro’, only the effect on the number of tubers of higher commercial value was seen. This contrasts with the results of the economic analysis, where L2 is indicated as the alternative that would generate higher income to the producers in ‘Diacol Capiro’, ‘Superior’ and ‘ICA Única’, while for the cultivars ‘Roja Huila’ and ‘Parda Bilingüe’ the level L1 is indicated since their income would be higher and their costs lower than the proposed alternatives. This depends on the relationship between the additional yields and the production costs to generate those yields, therefore, producing higher tuber yields does not imply higher economic gains for the farmer (^{Morales-Hernández et al., 2015}). The selling price is a highly influential variable in the analysis and decision making, hence the difference in economic results is also related to the price contrast between cultivars. In the period 2018-2020, ‘Diacol Capiro’ had average selling prices of COP $663/kg, while ‘Roja Huila’ had COP $455/kg (^{Colombia Fedepapa, 2021}).

The failure to find economic response in ‘Roja Huila’ and ‘Parda Bilingüe’ could also be related to other technical factors not considered, such as the availability of good quality seed, linked to the presence of negative phytosanitary factors that affect crop yield and prevent the expression of its productive potential (^{Wassihun et al., 2019}; ^{Avendaño and González, 2015}; ^{Esprella et al., 2012}).

In the case of ‘Pastusa Suprema’, the MARR was not higher than the MIRR, that is, the remuneration that the farmer would receive for the investment when using the L2 level would not be higher than the options of the fixed income financial market. However, with L2 there was an increase in profit of 18.5% with the MARR. This rate may be acceptable for producers with marginal technological levels, whose income is below that rate, placing the L2 level as a fertilization alternative for that specific case.

Finally, it should be noted that the amounts of commercial fertilizer (packages per hectare) applied by farmers are like those used in L2; however, the contribution of nutrients is different. L2 provides 180 - 400 - 358 kg ha^-1 of N, P₂O₅ and K₂O, while the producer's fertilization provides 292 - 297 - 354 kg ha^-1 (^{Monsalve et al., 2020}).