Published online 19 October 2005
Published in Agron J 97:1537-1542 (2005)
DOI: 10.2134/agronj2005.0067
© 2005 American Society of Agronomy
677 S. Segoe Rd., Madison, WI 53711 USA
Modeling
Yield Forecasting for Olive Trees
A New Approach in a Historical Series (Umbria, Central Italy)
Marco Fornaciari*,
Fabio Orlandi and
Bruno Romano
Dep. of Plant Biol., Agroenviron. and Anim. Biotechnol., Univ. of Perugia, Borgo XX Giugno 74-06121 Perugia, Italy
* Corresponding author (mfdp{at}unipg.it)
Received for publication March 7, 2005.
 |
ABSTRACT
|
|---|
In recent years, the relationship between flowering and fruit production was studied and evaluated in several wind-pollinated species. In olive (Olea europaea L.), the pollen-monitoring technique was introduced to determine pollen indexes as indicators of flowering, evaluating in some cases the predictive role of the variable. Recently, to investigate the reproductive efficiency, the Pollen Index (calculated during the entire flowering period) was replaced by the pollen emissions during the effective pollination period (EPP). In this study, in particular, an EPP elaborated (EPPe) value was derived from the EPP values and the average values of the meteorological variables. The regression analysis, considering winter chill accumulation, summer sums of minimum temperatures, and the EPPe, confirmed the strong relationship among meteorological variables, pollen emission, and final production in our study areas. This study has shown the need to use pollen data obtained from aerobiological monitoring in harvest-forecasting models in anemophilous plants such as olive. In particular, in the statistical models, pollen provides a synthesis of the historical information of the entire biological-reproductive cycle of the plant while meteorological trends interpret the incidental phenomena.
Abbreviations: CU, chilling units • EPP, effective pollination period • EPPe, effective pollination period elaborated • GDD, growing degree days • GDH, growing degree hours
|
INTRODUCTION
|
|---|
IN THIS STUDY, biological data related to flowering are used to predict the annual production of olive in the Mediterranean area. Agricultural forecasting has received particular attention in recent decades in an attempt to respond to different needs such as deciding crop rotations, responding to product demand, pricing, stock determination, work-related decisions, financial planning, etc. Agricultural forecasting is extremely difficult because the biological cycle processes that lead to the final production are strongly affected by meteorological events and climatic trends.
In September 1988, the European Economic Union (EEC) financed an agricultural pilot project (MARS-STAT PROJECT, 88/503/EEC) to conduct a census of the lands in agricultural use and their estimated yields. The latter were performed using agronomic-type (de Wit, 1965; Bouman et al., 1995) and agroclimatic (Palm and De Bast, 1987; Vossen and Rijks, 1995) models that give a sufficient amount of information but are lacking with respect to crop forecasting. Better results were obtained with forecasting models for herbaceous species that used the smoothing technique (Boken, 2000), which uses the production-related variables from the previous year (Prodt-1). With this method, high levels of variability can be explained. Time trend analyses (Palm, 1995) are another development in crop forecasting techniques, used particularly with long-term data series. This too has performed well for herbaceous species in which the incidence of technical evolution (fertilization, genetic selection) seems to be more strongly correlated.
In the 1990s, the relationship between flowering and fruit production was studied, and pollen was introduced as one of the variables. Pollen was monitored in the atmosphere during the entire flowering period of grape (Vitis vinifera L.) and olive, which are totally or partially anemophilous species. The flowering event can be considered to be a source of historical information regarding the vegetative and reproductive processes that lead to the formation of the flower apparatus and the actual reproductive phases of fruit formation. Some studies have shown how the pollination is related to production (Cour and Van Campo, 1980; Besselat and Cour, 1990; Candau et al., 1998) with the simultaneous use of climatic-type explicative variables. Recent studies have used the so-called annual Pollen Index as an indicator of flowering (Fornaciari et al., 1997, 1998), evaluating in some cases the predictive role of the variable (Minero et al., 1998; Clementi et al., 2001; Fornaciari et al., 2002). In a recent evaluation of reproductive efficiency expressed by the pollen parameter, Orlandi et al. (2005a) replaced the Pollen Index with the EPP to determine the relationship with annual fruit production in olive.
In general, the meteorological variables used in the forecasting models refer to the particular production year and are elaborated on the basis of various, arbitrarily chosen time periods (weeks, decades, etc.) that give values that are strongly correlated with production. At times, these results are difficult to interpret from a biological–physiological point of view. In addition, olive has a 2-yr reproductive cycle: the buds form in the first year, and the production process is completed in the second year. Olive is therefore subject to climatic effects over a rather long time span.
The aim of this work was to test the validity of using the EPP to predict production over a 20-yr period (1982–2004) using a modified parameter (EPPe) obtained through qualitative/quantitative evaluation with meteorological variables during EPP and, on the other side, the use of climatic variables linked to production, as climatic trends, chilling amounts and growing degree days (GDD)/hour amounts, as well as heat accumulation recorded in the summer period (June to September). The variables that were used explain more fully the influence of all the meteorological trends on the physiological processes of the reproductive cycle and avoid random correlation phenomena.
Given the economic importance of olive in the Mediterranean area due to the organoleptic quality of extra virgin olive oil, the aim of this study was to provide a reliable harvest-forecasting model that could be used in agronomic and economic planning strategies.
|
MATERIALS AND METHODS
|
|---|
The study was performed in the area surrounding Perugia (central Italy) in a territory characterized by olive groves (27230 ha in the entire Region, 18190 ha in the Province of Perugia), where "Umbria Extra Virgin Olive Oil," Denomination of Controlled Origin is produced (Fig. 1)
. The olive varieties cultivated in the area are mainly Leccino and Frantoio, along with Moraiolo and other local cultivars such as Dolce Agogia, San Felice, and Rajo. The olive groves were uniformly planted at the end of the 1980s on the sloping hills (5- by 5-m layout) (Tombesi, 1997; Pannelli et al., 1994).
The yearly olive yield, considered as a dependent variable in the study, expresses the total olive production for the Commune of Perugia (megagrams of milled olive). The data used for the regression analysis were obtained from the data bank of "AgeControl S.p.A." (Agency for the Control of the Italian Crop Production).
The pollen was monitored using a volumetric pollen trap, located at 493 m above sea level, on the roof of the Agricultural Faculty of the University of Perugia, which is capable of capturing olive pollen released from an area within a 30-km radius. Phenological observations in the field were made at the same time as the pollen monitoring to test the significance of the monitoring itself (Edmonds, 1979).
The pollen monitored in the atmosphere has been captured continuously since 1982 and is reported as the number of daily pollen grains/m3 (Fornaciari et al., 2000), during the entire flowering period. Starting with daily data, annual EPPs were constructed (1982–2004) by adding the pollen concentrations of the 4 d preceding the maximum pollen concentration peak in the atmosphere. This period corresponds to maximum flowering, as confirmed by phenological observations in the field. New indicators were then extracted from these. The EPPe was derived from the interaction between the EPP values and the mean values of precipitation and maximum and minimum average temperatures recorded during the EPP. The EPPe value was derived from the direct proportional ratio [EPPe = (EPP x T)/100, where T = temperature (°C)] and the inverse [–EPPe = (EPP x T)/100] between the EPP values and the mean values of the meteorological variables indicated.
Meteorological data were obtained from the Central Ecological Office Weather Station (Italian Ministry of Agriculture) designated 211PG (Fig. 1), which registers a series of meteorological parameters (i.e., solar radiation, atmospheric pressure, evapotranspiration). The apparatus has been located near Perugia since 1978 and collects data that is representative for almost all of the provincial territory. Daily values were elaborated to obtain chilling units (CU), growing degree hours (GDH), and GDD for the year being studied, t, and the preceding year, t – 1, for all the years considered.
Chilling units were calculated using the Utah Method (Richardson et al., 1974). A thermal range of 3 to 9°C was considered optimal, and the maximum chilling value was assigned. Above and below these temperatures, the chilling effect was reduced. The chilling amounts were calculated starting from two different dates (1 December and 1 January) until six final dates (15 January, 1 February, 15 February, 1 March, 15 March, and 1 April), giving 12 different values. The onset dates were determined in other studies taking into account the biological cycle of olive in the investigated area (Orlandi et al., 2002).
The GDH were calculated by using the method of Anderson et al. (1986) while the GDD were calculated using the method proposed by Baskerville and Emin (1969), which uses 12 threshold temperatures from 4 to 15°C.
The GDH and GDD amounts were calculated from daily values starting from two onset dates (1 January and 1 February) until the dates of maximum pollen concentrations in the atmosphere (peak of pollination).
Regarding the meteorological variables, the sums of the monthly and total cumulative values were calculated for the maximum, minimum, and average temperatures in the summer (June–September) to obtain a summer thermal stress indicator and indirectly, a water stress indicator. A linear regression model was constructed using the S-Plus statistical software to apply the normal tests for verifying the robustness of the model.
|
RESULTS AND DISCUSSION
|
|---|
The annual EPPe values, used in the forecasting model, are reported in Table 1. In all the years studied, better results were obtained using the average temperature in the direct proportion ratio with the annual EPP values compared with the values obtained using the maximum and minimum temperatures and precipitation values. This is probably due to the fact that the average temperature value expresses the thermal trend better for a given period by mediating the extreme values that can be recorded with the other temperature variables. In addition, in a summer-flowering species, the precipitation values, rare and/or occasional, are not correlated with the event while it plays a determining role in the phases immediately preceding the phenomenon (Fornaciari et al., 1998). The EPPe values (Table 2) show a certain annual variability that is closely linked to the production variability as can be seen in the correlation levels. Compared with previous studies, using the pollen parameter in the production forecasting models notably increases the explained variance value using the EPPe with respect to the Pollen Index (Ip) (Fornaciari et al., 1997; Minero et al., 1998), the Concentrated Pollen Index (Ipc) (Fornaciari et al., 2002; Galan et al., 2005), or the same EPP (Orlandi et al., 2005b). The high performance values of the pollen parameter confirm, on the one hand, the relationship between the flowering event and the successive phases of fruit formation and, on the other hand, explain the particular elaboration used in this study (EPPe). Using only the pollen concentrations in the period of full flowering and linking those values to the thermal values of the subperiod considered allows the level of wind pollination to be expressed and therefore the level of pollen transport efficiency in the atmosphere. This is to say that with higher average temperature values, the pollen transport capacity increases; this is also due to lower atmosphere pressure (Edmonds, 1979). The increased reproductive efficiency of the pollen is also due to physiological phenomena associated with the level of pollen grain dehydration (Sanzol and Herrero, 2001). The analysis of correlation between production and the meteorological variables (CU, GDD, and GDH) in the years t and t – 1 showed the best results with the cold accumulation December–February t – 1 period while scarcely relevant results were obtained with the two parameters related to heat accumulation in both years. This result confirms the strong relationship between cold and olive production in our study areas (Fornaciari et al., 1997; Orlandi et al., 2005b). This relationship could be determined by the olive cultivars used in Umbria, as well as by the climatic characteristics of central Italy (almost the northern-most boundary for olive production) that are not typical Mediterranean conditions where olive has been historically cultivated (Rallo and Martin, 1991; Orlandi et al., 2005b).
View this table:
[in this window]
[in a new window]
|
Table 1. Effective pollination period elaborated (EPPe) values realized calculating the direct proportionality ratio between effective pollination period (EPP) and meteorological variable averages in the same periods.
|
|
Regarding the potential influence of the thermal values registered during the summer period on production, the sum of the minimum daily temperatures in the period 1 June to 30 September is highly correlated (Table 2) with the quantity of olive harvested. It is the highest correlation value recorded between the variables used in the model (r = 0.73). This relationship can be associated with the physiological phenomena of fruit growth and maturation.
The crop forecasting model can therefore be expressed as:
The regression coefficients and the respective signs are reported in Table 3. The variation explained is high (R2 = 0.8917). The levels of significance for all the variables introduced into the forecasting model are also high (t value < 0.01). A dummy variable (DUM) was introduced into the model that considered the particularly high production in the year 2000, which was difficult to explain from the correlation levels with the variables considered. The anomaly is between the low quantity of pollen released into the atmosphere and the high final production value (the highest among all those considered in the historical series).
The linear regression plot and that of the residuals are reported in Fig. 2
. Almost all of the residuals fall within the 5000 Mg of waste value. The real and fitted values of yearly olive yields are also shown; the efficiency of the analysis performed in the reconstruction of the values is evident (Fig. 3)
. The regression model gives an average margin of error, equal to 17.2%, that considers the entire historical series and eliminates the extreme values (the best and worst). It is interesting to observe that if the production values for the years in which particularly adverse weather conditions were registered (freezes of 1986, 1991, and 2003) are eliminated, the average margin of error decreases to 13.6%.
|
CONCLUSIONS
|
|---|
This study has shown the need to use pollen data obtained from aerobiological monitoring in harvest-forecasting models in anemophilous plants such as olive. This variable is strongly correlated with the final fruit production; the pollen data provide a synthesis of the historical information of the entire biological–reproductive cycle of the plant. The method used has demonstrated that of all the pollen released into the atmosphere by the plant, only a certain amount of pollen within a restricted time period has real reproductive efficiency linked to fruit formation. The relationship with the climatic variables becomes evident at this critical moment and then during the growth and maturation phases. Besides determining the condition for better pollen transport, these variables also strongly influence the phenomena associated with fruit physiology, including the noted catastrophic fruit drop.
Further study is needed on the freezing phenomena (winter and spring freezes) that are typical of the central Italian areas and strongly influence the capacity of pollen forecasting. This meteorological event causes the most damage with respect to the flowering buds, which greatly reduces the quantity of pollen registered in those years. The plant, however, in the same years, compensated for the scarce flowering by greatly reducing the level of fruit fall, thereby producing a harvest quantity that was not directly influenced by the meteorological event (Morettini, 1972).
It is important to consider the predictive capacity of the model to make an accurate forecast in the presence of a long historical series, long in advance. This could help indicate the need for intervention on the current crop (i.e., emergency irrigation and fertilization) or for the stored product in the case of a forecast for an upcoming abundant production.
|
REFERENCES
|
|---|
- Anderson, J.L., E.A. Richardson, and C.D. Kesner. 1986. Validation of chill unit and flower bud phenology models for "Montmorency" sour cherry. Acta Hortic. 184:71–78.
- Baskerville, G.L., and P. Emin. 1969. Rapid estimation of heat accumulation from maximum and minimum temperatures. Ecology 50:514–517.
- Besselat, B., and P. Cour. 1990. La prévision de la production viticole à l'aide de la technique de dosage pollinique de l'atmosphère. Bulletin de l'O.I.V. 63:721–740.
- Boken, K.V. 2000. Forecasting spring wheat yield using time series analysis: A case study for the Canadian Prairies. Agron. J. 92:1047–1053.[Abstract/Free Full Text]
- Bouman, B., H. Van Keulen, and H.H. Van Laar. 1995. The "school of de Wit" crop growth simulation models. Agrometeorological models: Theory and applications in the MARS Project. Report EUR 16008 EN. Office for Official Publ. of the E.U., Luxembourg.
- Candau, P., F. Minero, J. Morales, and C. Tomas. 1998. Forecasting olive (Olea europaea L.) crop production by monitoring airborne pollen. Aerobiologia 14:185–190.
- Clementi, M., S. Clementi, M. Fornaciari, F. Orlandi, and B. Romano. 2001. The golpe procedure for predicting olive crop production from climatic parameters. J. Chemom. 15:1–8.
- Cour, P., and M. Van Campo. 1980. Previsions de récolte à partir de l'analyse du contenu pollinique de l'atmosphère. C.R. Acad. Sci., Ser. D 290:1043–1046.
- de Wit, C.T. 1965. Photosynthesis of leaf canopies. Agric. Res. Rep. 663. Pudoc, Wageningen, the Netherlands.
- Edmonds, R.L. 1979. Aerobiology: The ecological systems approach. Dowden, Hutchinson & Ross, Inc., Stroudsberg, PA.
- Fornaciari, M., L. Pieroni, P. Ciuchi, and B. Romano. 1997. A statistical model of airborne pollen grains and some meteorological parameters (Olea europaea L.). Agric. Mediterranea 127:134–137.
- Fornaciari, M., L. Pieroni, P. Ciuchi, and B. Romano. 1998. A regression model for the start of the pollen season in Olea europaea L. Grana 37:110–113.
- Fornaciari, M., C. Galan, A. Mediavilla, E. Dominquez, and B. Romano. 2000. Aeropalynological and phenological study in two different Mediterranean olive areas: Cordoba (Spain) and Perugia (Italy). Plant Biosyst. 134(2):199–204.
- Fornaciari, M., L. Pieroni, F. Orlandi, and B. Romano. 2002. A new approach to consider the pollen variable in forecasting yield models. Econ. Bot. 56(1):66–72.
- Galan, C., L. Vazquez, G.M. Heminia, and E. Dominguez. 2005. Forecasting olive (Olea europaea) crop yield based on pollen emission. Field Crops Res. 86(1):43–51.
- Minero, F.J.G., P. Candau, J. Morales, and C. Tomas. 1998. Forecasting olive crop production based on ten consecutive years of monitoring airborne pollen in Andalusia (southern Spain). Agric. Ecosyst. Environ. 69:201–215.
- Morettini, A. 1972. Olivicoltura. REDA, Rome.
- Orlandi, F., M. Fornaciari, and B. Romano. 2002. The use of phenological data to calculate chilling units in Olea europaea L. in relation to the onset of reproduction. Int. J. Biometeorol. 46:2–8.[Medline]
- Orlandi, F., B. Romano, and M. Fornaciari. 2005a. Effective pollination period in olive (Olea europaea L.): A pollen monitoring application. Scientia Hortic. 105:313–318.
- Orlandi, F., L. Ruga, B. Romano, and M. Fornaciari. 2005b. Olive flowering as an indicator of local climatic changes. Theor. Appl. Clim. 81(3–4):169–176.
- Palm, R. 1995. Regression methods including the Eurostat Agromet model and time trends. Joint Research Centre of the E.U. Publication EUR 16008 EN. Office for Official Publ. of the E.U., Luxembourg.
- Palm, R., and A. De Bast. 1987. Principes de construction des modèles agrométéreologiques. Ann. Gembloux 93:39–54.
- Pannelli, G., D. Volpe, P. Preziosi, F. Famiani, S. Lavee, and I. Klein. 1994. Comparison of the vegetative and reproductive characteristics of traditional olive cultivars and selected low vigorous accessions in Central Italy. Acta Hortic. 356:123–126.
- Rallo, L., and G.C. Martin. 1991. The role of chilling in releasing olive floral buds from dormancy. J. Am. Soc. Hortic. Sci. 116:1058–1062.[Abstract/Free Full Text]
- Richardson, E.A., S.D. Seeley, and D.R. Walker. 1974. A model for estimating the completion of rest for Redhaven and Elberta peach trees. HortScience 82:302–306.
- Sanzol, J., and M. Herrero. 2001. The "effective pollination period" in fruit trees. Scientia Hortic. 90:1–17.
- Tombesi, A. 1997. Olive orchard installation, soil arrangement, planting density and training. Proc. Int. Seminar on Olive Growing, Chania, Crete, Greece. 18–24 May 1997. Inst. of Olive Tree and Subtrop. Plants of Chania, Chania, Crete, Greece.
- Vossen, P., and D. Rijks. 1995. Early crop yield assessment of the EU countries: The system implemented by the Joint Research Centre. Report EUR 16318 EN. Office for Official Publ. of the E.U., Luxembourg.
This article has been cited by other articles:
|
|
|
|
 
C. Galan, H. Garcia-Mozo, L. Vazquez, L. Ruiz, C. Diaz de la Guardia, and E. Dominguez-Vilches
Modeling Olive Crop Yield in Andalusia, Spain
Agron. J.,
January 11, 2008;
100(1):
98 - 104.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
