Most likely, each producer and technician in the cherry industry has his or her own hypothesis or, rather, at this point, his or her conclusions for the current season; that is, regarding the productive conditions per se of each area/variety, and which have influenced the productive potential of the 2023-24 season.
From this point of view, it is important to carry out the studies objectively. To do this, it is necessary to obtain a large amount of information in order to refine the variables that have the greatest weight in the productivity of cherry cultivation and thus be able to carry out rational management that allows reducing those variables that have the greatest influence. In this context, the Avium technical team has held a series of high-level technical meetings, where the most relevant factors in the current season's production decline have been discussed and analyzed in detail.
The principal component analysis (PCA) technique was used to identify patterns of behavior of various agroclimatic variables, in addition to classical statistical methods to identify correlations between variables. Principal component analysis is a tool that enables the representation of a large set of variables through only two principal components, which explain a percentage of the total variability of the data and allow the identification of patterns of behavior.
Before reviewing the results, it is important to understand the basics of PCA graphical interpretation. Figure 1 below illustrates two vectors (arrows) named A and B. Each arrow represents a variable, and the length of the arrow indicates its weight or importance in the total data set. In other words, as the length of the arrow increases, so does its relevance in the analysis.
Figure 1. Graphical interpretation of vector length (weight) in a PCA. Source: Avium 2024.
From the angle of the vectors point of view, this indicates the relationship between two or more variables. In figure 2.1, both variables B and C are related positively and negatively respectively with variable A. However, in figure 2.2, the right angle of 90° between the arrows representing A and B indicates the absence of a relationship between these two variables.
Figure 2. Graphical interpretation of the angle between variables in a PCA. Source: Avium 2024.
In an analysis of such complexity, a parsimonious way to approach it is through the identification of groups of variables with similar characteristics, added to their interpretation from a technical point of view.
1. Case studies.
First, the focus of the analysis was to identify trends and/or patterns in different agroclimatic zones of central Chile (early, middle and late) during the 2023-2024 season, based on agroclimatic and phenological variables, among others, and their impact on productivity in Santina and Lapins cultivars. Second, a specific analysis was carried out for the Santina variety on Colt rootstock in each particular agroclimatic zone for several seasons, also identifying variables that conditioned productivity with respect to the characteristics of each zone.
1.1 General analysis of productive areas.
The analysis was carried out on the Santina and Lapins varieties in productive plots from the planting year prior to 2019, which were separated into three different zones: early, medium and late, according to agroclimatic characteristics.
The factors or variables that make up the ACP are:
-Projected and/or actual yield in cases of harvests finished on December 10 (kg/ha), where it was normalized according to the following parameters: -Low: 0-5,000 kilos/hectare -Medium: 5,001-10,000 kilos/hectare -High: <10,001 kilos/hectare.
-Agroclimatic zone (early, mid and late).
-Precipitation in August (mm).
-Precipitation in September (mm).
-Rainfall during flowering (mm fallen during the flowering phenological period).
-Soil texture (sandy loam, loam, clay loam).
- Base chill hours 7.2°C (accumulated from May 1 to the date of application of dormancy breakers or, failing that, until July 31).
-Baseline chill hours of 7.2°C accumulated individually in May, June and July.
-Accumulated degree days (GDA) Base 10° C during September.
-Accumulated degree days (CDD) Base 4.5° C during September.
-Bee flight in flower (number of hours with temperature over 15° C and radiation over 300W/m2 in the flowering phenological period).
The PCA only allows analysis of numerical variables, however, categorical variables such as soil texture, agroclimatic zone, among others, were available. To include them, the “one-hot” coding method was used, where new variables were created with binary values from 0 to 1, representing each category of the original variable. In the first version of this technical article we explored the differences in the behavior of the early zone compared to the middle and late zones. The updated results of the second version of the analysis reinforce this conclusion.
Figure 3.1 presents a PCA that explains 68.4% of the variability and uses the variables mentioned above. Broadly speaking, two main groups can be identified, highlighted in red and green. The red group is associated with low kilos/hectare, while the green group includes medium and high kilos/hectare. The low kilos/hectare group (red) is explained by rainfall in August, clay loam soils (greater moisture retention), and a greater accumulation of degree days (GD) base 4.5° C in September (low temperatures in spring). On the other hand, the medium and high kilos/hectare group (green) is mainly associated with a greater accumulation of base chill hours 7.2° C, loam and sandy loam soils, and a greater accumulation of base degree days 10° C in September.
Figure 3.2 shows a graph with the weight or importance of each variable, on the graph and each individual component; the larger the size of the blue circle, the greater its importance and, therefore, the length of the arrow in the graph.
Figure 3. Principal component analysis updated as of December 10, 2023. Source: Avium, 2024.
1.1.1 Bee flight behavior (VA).
A crucial factor in the productive equation is fruit set, which requires a successful pollination and fertilization process. Certain environmental conditions are needed during the effective pollination period (EPP), which are related to the match between temperatures above 15°C and radiation above 300 W/m2An indicator was constructed quantifying the number of hours with these conditions, to compare the current season with previous ones.
Figure 4 shows the weekly rate of bee flight hours for the last four seasons, where the 2023-24 season (black line) is well below the 2020-23 average (solid blue line), at least during the first three weeks of September.
Figure 4. Bee flight hours (T>15° C and Rad>300 W/m2) weekly for the last four seasons. Teno Station. Source: Avium 2024.
2) weekly for the last four seasons.
Teno Station.
1.1.2 Precipitation Behavior (PP).
Precipitation during the 2023-24 season was abundant, both in terms of rainfall (mm) and frequency. However, the timing of precipitation becomes more relevant when it coincides with the initial phenological states to the point of, on the one hand, affecting soils with a greater capacity to retain moisture, keeping them at saturation levels for days and/or weeks, and on the other hand, negatively impacting climatic conditions during the PEP.
Figure 5 shows weekly rainfall from August 3 to October 20 in the 2023-24 season (black line), in contrast to the 2020-23 average (solid blue line) and other seasons.
Figure 5. Weekly rainfall (mm) during the last 4 seasons. Teno Station. Source: Avium 2024.
1.2 Analysis of zones between seasons.
It is important to recognize the variables at a general level that could affect crop production in the different varieties, however, being able to do a study on the Santina variety on Colt rootstock this season is undoubtedly necessary, understanding that this was the one that had a greater impact on the lower productivity of the 2023-24 season. When doing this analysis, it must be parameterized according to the different agroclimatic and productive seasons of the particular area, in this way the factors or variables that have the greatest weight in the productive potential can be identified. For the study of each variable, it was carried out with the principal component dimensionality reduction (PCA) technique, explained above, and where the factors or variables that make up the study are the following:
-Yield (kilos/hectare).
-Agroclimatic zone (early, mid and mid/late).
-Precipitation in August, September, October and the accumulated amount between them (mm).
-Base chill hours 7.2°C, 2023 season (accumulated from May 1 to July 31).
-Base cold hours of 7.2°C accumulated individually in May, June and July and the accumulation between them.
-Accumulated degree days (GDA) Base 10°C accumulated from August to October and the monthly individual.
-Accumulated degree days (GDA) Base 4.5°C accumulated from August to October and the monthly individual.
-Bee flight in flower (number of hours with temperature over 15°C and radiation over 300W/m2 in the flowering phenological period).
1.2.1. Early zone.
The characteristics of this area are geared towards obtaining productions that can be commercially positioned as one of the prime fruits of the central zone, where the edaphoclimatic characteristics of the analyzed orchard are identified with clayey textured soils, with accumulations of cold to the limit and with climatic conditions of thermal stress, which could affect an adequate post-harvest in the Santina/Colt combination; precisely these turned out to be one of the most affected in terms of production.
Principal component analysis (PCA) is aimed at performing a descriptive analysis in order to infer associations and trends between variables and/or observations. Figure 6 presents a PCA of the early zone of different climatic variables (chill hours, degree days, etc.) and productive index components (AdYC and production), where different conclusions can be inferred. Firstly, it can be identified that there is an explanation of the 100% of the data (CP1 + CP2), where the greatest variability of the data is explained by the CP1 (principal component 1) with the 62%. It is in this sense that the variables that are related to productivity can be observed (green color) and in the opposite sense, those that would have a lower relationship or relation to this productive item (red color).
Figure 6. Principal component analysis (PCA) of the last three seasons in cv. Santina on Colt rootstock in an early area. Source: Avium 2024.
Figure 7 shows the degree of correlation that exists between the climatic variables with respect to the productive potential in cv. Santina on Colt rootstock. The degree of correlation is given with respect to the definition made in Figure 2, where it can be distinguished that there are positive and negative relationships with respect to the items studied. To interpret this graph, it can be observed that those variables that have a positive correlation (positive values) with production (that is, that increasing or decreasing the amount of an agroclimatic variable would have an impact on productivity (bars over the value 0) and also those that have a negative impact on productivity as negative correlations, where when increasing a variable x the productivity item decreases or vice versa), when an agroclimatic variable decreases, the productivity factor increases.
In this type of analysis, it is very important to point out that a correlation coefficient can never be interpreted as a cause and effect relationship, in which two variables covary, that is, they tend to be high or low simultaneously, does not mean that they are the cause of the other. In other words, for example, the fact that a high correlation between greater bud fertility (AdYC) in an orchard and productive yield (kilos/hectare) cannot be interpreted as meaning that productivity is due to AdYC, although it can be proposed as a hypothesis that in some way allows explanations and trends to be given to this great multifactoriality of data that come together in the production of this crop.
Figure 7. Correlation graph of productivity and different variables in cv. Santina on Colt rootstock in an early area in the last three seasons. Source: Avium, 2024.
When analyzing the graph, it can be seen that the variables that would have a positive impact on productivity are: base degree days 10ºC and 4.5ºC (GD10 and GD4.5) for the month of October, the accumulation of cold hours (HF) in general, bee flight (VA) at the beginning and middle of September and fertility expressed in bud fertility analysis (AdYC). In contrast, the variables that are observed to be negatively correlated with productivity are related to GD10 and GD 4.5 for the months of August and September, rainfall in general from August to October, the post-harvest stress index (IE) and VA in the month of October mainly.
In the case of the GD in October, in this area it is assumed that there is a better condition regarding stopping a better productivity, this being an item that should be studied in greater depth in this area. However, in general the HF show a positive tendency to obtain better results in the productive potential, being interesting that the month of July, at the end of the cold accumulation period, has a greater correlation with respect to the months of May and June. Regarding precipitation (PP), a negative relationship is observed with the productive potential, where this variable is an indirect measure, since it is related to the degree of accumulation/saturation of water in the soil of clay characteristics with a high capacity for moisture retention, this condition being negative in the month of August in the amount of mm of rain, prior to the awakening of the plants.
The EI is a variable that is becoming increasingly important in terms of the productive impact of the next season, since it plays an important role in carbon and nitrogen reserves for the beginning of flowering in spring. The variable of bee flight hours (VA) is reflected in Figure 8, where the VA conditions for each season and its degree of correlation with the productive index for each week of the month of September and part of the month of October can be observed.
From the graph it can be concluded that those seasons that had the best VA conditions and active flowering states occurred in the third week of September (09/21/23) and were the ones that had a better productive result (Pearson correlation of 0.38); this in contrast to Tp. 2023-24 where an evident delay was observed in the initial phenological states with full flower (PF) dates in the first week of October (10/06/23) and which shows a negative correlation index between production/VA of -0.72. In practical terms it can be concluded that in this productive area, in this variety, it turns out that the perfect combination of VA and full flower phenological states are associated with the third week of September and that they would have an impact on a better productive result considering only these variables.
Figure 8. Bee flight hours (T>15° C and Rad>300 W/m2) weekly during the last three seasons in the early zone. Source: Avium, 2024.
1.2.2. Middle Zone.
In the orchard analyzed in this area, productions ranging from 16,000 to 24,000 kilos per hectare can be observed in recent seasons, however, productions during 2023-24 were close to 9,000 kilos/hectare. Although this productivity was not projected in this productive unit, this would be a very good production for a large number of field orchards considering the climatic effects on the rootstock combination studied. The soil has a clay loam texture with a high infiltration capacity, with accumulations of 620 to 846 HF in recent seasons, except for 2023-24 which obtained around 590 HF.
In Figure 9, when positioning ourselves in the 2023-24 season, it can be observed that the variables that have the greatest weight, that is, influence on the season, are those related to the EI of March and April, followed by HF of May together with the GD4.5 in order of importance. It should be noted that the kilos/hectare variable is diametrically opposed to the 2023-24 observation, which indicates that those variables that are in the same projection with productivity have great relevance with this item such as VA, HF of July and the AdYC index.
Figure 9. Principal component analysis (PCA) of the last six seasons in cv. Santina on Colt rootstock in a middle zone. Source: Avium, 2024.
The IE is one of the variables that again has a high weighting with respect to the impact on lower productivity, where accentuated negative correlations are shown in the months of February, March and April. This can be identified in Figure 10, where the HF in the month of July would also show a good accumulation performance compared to the months of May and June; likewise, the GD10 and GD4.5 in the month of September would have a greater positive impact compared to the month of October.
Figure 10. Correlation graph of productivity and different variables in cv. Santina on Colt rootstock in a medium area in the last six seasons. Source: Avium, 2024.
Bee flight activity (VA) has its best correlation with productivity between week 3 and 4 of September (Figure 11), as long as the phenological states coincide in this period. The productive decline that originated in the past 2023-24 season is probably closely related to this variable, since the flowering phenological states were at the beginning of October, with positive correlation indices, but lower than those of the third and second weeks of September.
Figure 11. Bee flight hours (T>15° C and Rad>300 W/m2) per week during the last six seasons in the middle zone. Source: Avium, 2024.
1.2.3. Late Zone.
The orchard analyzed in this opportunity is characterized by having soils with a loamy-silty texture, with HF accumulation floors of 680 in the last four seasons and in general with good VA conditions for the pollination and fertilization processes. Like the orchards analyzed in the early and middle zone, this zone also presents the same trend in HF, where the month of July is associated with greater productivity, also the VA in the last weeks of September and the GD10 and GD4.5 in October (Figure 12).
Figure 12. Principal component analysis (PCA) of the last four seasons in cv. Santina on Colt rootstock in a late zone. Source: Avium, 2024.
The correlation analysis follows the same dynamics as seen previously, where the EI and the precipitations (PP) are those that would have the greatest downward impact on productivity. Thus, last season, HF accumulation was observed in the month of July with a greater impact on productivity, as in the previous analyses. GD10 and GD4.5 also show an interesting response in productivity (Figure 13).
Figure 13. Correlation graph of productivity and different variables in cv. Santina on Colt rootstock in a late area in the last 4 seasons. Source: Avium, 2024.
Figure 14 shows the bee flight hour curves for each season, with the 2023-24 season being similar to the 2022-23 season, with lower bee flight hours compared to the 2020-21 and 2021-22 seasons. The difference may be that the phenological states of Tp. 2022-23 were between the third and fourth week of September, with Pearson correlation indices of 0.76, while in 2023-24 the flowering states were the first week of October with correlations of -0.9 in terms of the VA/production ratio.
Figure 14. Bee flight hours (T>15° C and Rad>300 W/m2) per week during the last four seasons in the middle zone. Source: Avium, 2024.
2. Conclusions.
At this point, some initial conclusions regarding the most relevant factors in the productive decline of the 2023-24 season are gaining weight. It is important to mention that there is no preferential order for these.
2.1.Precipitation.
The timing of precipitation over the total accumulated for a season has a real impact on productivity. If we analyze the timing of these events, leaving out flooding cases from the analysis, the rainfall in August is the one that has the greatest impact on productivity, followed by rainfall in September. This factor is indirectly related to production, since it brings about phenomena such as poor spring conditions and saturated soils. On the other hand, within the analysis, negative correlations were intrinsically seen between precipitation in contrast to the accumulation of winter cold and degree days, highlighting once again the impact of the timing of precipitation.
2.2.Soil texture.
Soil texture is a factor that is repeated from the previous analysis; this makes a lot of sense when we relate it to lower temperatures in spring and higher rainfall, since the combination of these, plus a soil with smaller particle size or, failing that, a greater capacity to retain moisture, was fatal in many cases.
2.3. Base chill hours 7.2° C.
The accumulation of chilling hours below a threshold of 7.2°C is another relevant factor, since in a season like the current one “at the limit”, many of the cases of low accumulation do not reach the minimum requirements of some varieties. It is important to highlight that in Santina on Colt rootstock, a greater weighting was identified in the accumulated chilling hours during the month of July, compared to May and June.
2.4. Bee flight.
The flight of the bee, as mentioned above, results from hours with temperatures above 15°C and radiation above 300 W/m2, in general, has a particular behavior. Good environmental conditions promote a successful process of pollination and fertilization, however, in the current season, a large part of the blooms occurred with environmental conditions below the average of those regarding this indicator of bee flight. On the other hand, during this season the behavior was not linear, apparently its behavior is asymptotic. The heterogeneity in the budding conditioned the duration of phenological states, in particular flowering, impacting on the final yield.
