Agronomy Journal Grow Your Career With ASA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Miyasaka, S. C.
Right arrow Articles by Kodani, L. S.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Miyasaka, S. C.
Right arrow Articles by Kodani, L. S.
Agricola
Right arrow Articles by Miyasaka, S. C.
Right arrow Articles by Kodani, L. S.
Related Collections
Right arrow Other Tubers
Right arrow Crop Models
Agronomy Journal 95:545-557 (2003)
© 2003 American Society of Agronomy

MODELING

Site and Planting Date Effects on Taro Growth

Comparison with Aroid Model Predictions

Susan C. Miyasaka*,a, Richard M. Ogoshib, Gordon Y. Tsujib and Leslie S. Kodania

a Univ. of Hawaii, Hawaii Branch Stn., 461 W. Lanikaula St., Hilo, HI 96720
b Dep. of Trop. Plant and Soil Sci., Univ. of Hawaii, 1955 East-West Rd., Room 206, Honolulu, HI 96822

* Corresponding author (miyasaka{at}hawaii.edu)

Received for publication February 25, 2002.

   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Taro [Colocasia esculenta (L.) Schott cv. Bun-long] is a tropical root crop with the potential to be grown commercially on former sugarcane (Saccharum officinarum L.) lands in Hawaii. To determine the effects of varying environmental conditions on crop production and to validate an aroid simulation model developed earlier, taro was grown under rainfed conditions at two sites that differed in elevation (90 and 335 m) on Hawaii island and at four planting dates (Winter, Spring, Summer, and Fall) on 27 February, 28 May, 27 August, and 24 November 1992. Biomass harvests were conducted at bimonthly intervals. Using weather, soil, cultivar, and management practices in this field trial, the aroid crop simulation model predicted dry weight of plant components, leaf area index (LAI), and time to harvest maturity defined as leaf stage. Fresh and dry weights of corms increased linearly from 1 to 13 months after planting (MAP), indicating continuous partitioning to the storage organ. The increase in corm fresh weight was significantly greater for Spring + Summer plantings compared with Fall + Winter plantings, primarily due to lower incidence of corm rot. Due to its indeterminant growth, taro can be harvested between 6 and 13 MAP, depending on incidence of pests and soil and weather conditions that could cause early maturation. The aroid model simulated well the maximum LAI; however, it underestimated both potential dry corm yield and length of time to harvest maturity, indicating that further development of this aroid simulation model is needed.

Abbreviations: LAI, leaf area index • MAP, months after planting • RMSE, root mean square error • RMSE(s), systematic root mean square error • RMSE(u), unsystematic root mean square error


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
COMMERCIAL SUGARCANE production completely ceased on the island of Hawaii in 1996, opening up new land areas for diversified crops such as taro. Taro is an herbaceous, perennial, tropical root crop composed of a main plant and suckers (Fig. 1) . It is harvested primarily for its corm, which is a starchy, underground stem (Plucknett et al., 1970). Taro is adapted to moist environments and can be grown under rainfed or irrigated upland (i.e., nonflooded) as well as flooded conditions (Plucknett et al., 1970). It is one of the most important staple crops in the Pacific Islands, and it is grown widely throughout Africa, Asia, the West Indies, and South America (Plucknett et al., 1970). Under upland conditions in Hawaii, Western Samoa, or Fiji, it is a 9- to 11-mo crop (Plucknett and de la Pena, 1971; Reynolds, 1977; Sivan, 1982), and only corms are harvested due to the small size of the cormels.



View larger version (39K):
[in this window]
[in a new window]
  		Fig. 1. Taro plant components.

 
Reynolds (1977) concluded that there were five growth stages of taro: (i) a period of establishment with root formation and leaf production during the first month, (ii) a period of rapid root and shoot development with initiation of corm development during one to four months, (iii) a climax of root and shoot growth with a rapid increase in corm formation during four to six months, (iv) a senescence period of decreasing root and shoot growth with continued increase in corm size during 6 to 9 mo, and (v) a period of decreasing corm weight perhaps due to rot and of initiation of further vegetative growth with increased root and shoot growth during 8 to 10 mo. The initial period of dry matter loss during establishment can be minimized with adequate water and nutrients, as shown by a greenhouse study in which dry matter accumulation in corms occurred within 10 d after planting and increased nearly linearly to 30 d (Jacobs and Clarke, 1993).

Sivan (1982) reduced the number of growth phases by combining the third and fourth phases and eliminating the fifth phase. However, corm rots, caused by Pythium spp., Sclerotium rolfsii Sacc., or Phytophthora colocasiae Rac. (Ooka, 1994; Plucknett et al., 1970), can be a major factor limiting taro yields. Yield losses of corms were measured at 25% under flooded conditions (Parris, 1941) or up to 30% under upland conditions (Miyasaka et al., 2001).

Because taro growth is indeterminant, it is difficult to define crop maturity (Singh et al., 1992). Maturity could be defined as when: (i) average yield is achieved, (ii) maximum yield is achieved, (iii) acceptable quality is achieved, (iv) corm diameter declines near the petiole (i.e., stump diameter), (v) cormel dry weight exceeds corm dry weight, or (vi) a certain phenological leaf stage is reached. Using the first criteria, maturity is achieved when corm weights reach the average taro yield on the island of Hawaii of 14 Mg ha-1 (Hawaii Agric. Stat. Serv., 1994). Second, Sivan (1982) defined maturity as the peak corm yield stage, which occurred at 10 MAP in two cultivars in Fiji. Third, the taro corm could be considered mature when it reaches a dry matter percentage of 20, the minimum acceptable level for good quality (S.C. Miyasaka, unpublished data, 1998). Fourth, farmers consider taro to be mature when corms are observed to become reduced in diameter near the petiole, forming a bottleneck shape. Prasad and Singh (1992) assumed maturity occurred when 50% of plants had corms with this bottleneck shape. Fifth, under upland conditions, corms are the economically important part of plant, whereas cormels are used primarily as vegetative propagating materials. Because corms and cormels are considered to be competing sinks for photosynthates (Goenaga, 1995), the main plant could be considered harvestable when dry matter accumulation in cormels is greater than that in the more commercially important corms. Finally, Prasad and Singh (1992) found that the 28th leaf stage coincided with 50% of plants having corms with a bottleneck shape.

A crop model, SUBSTOR-Aroid (Singh et al., 1992), simulates development and growth processes of taro on a daily interval. Four types of input data are needed to run the model: (i) daily weather, (ii) soil properties and initial conditions, (iii) cultivar, and (iv) management practices. Crop development is simulated based on thermal time, or degree day. Successive leaves appear after a specified number of growing degree days, using a minimum temperature of 10°C (Singh et al., 1998). The parameters that define the phyllochron interval are PHINT, TIPINT, and TIPGRAD, where PHINT is the basic phyllochron interval and TIPINT and TIPGRAD modify the PHINT value for the first 11 leaves. The phyllochron interval gradually lengthens from Leaf 1 to 11, stays constant until Leaf 17, and then lengthens again. As discussed earlier, harvest maturity is estimated to occur at the 28th leaf stage (Prasad and Singh, 1992).

Intercepted total global solar radiation, which is a function of leaf area and an extinction coefficient, is converted to dry matter via a first-degree equation (Singh et al., 1998). Dry matter partitioning to roots, stems, leaves, corms, and suckers is a function of developmental phase and quantity of photosynthate produced. Partitioning coefficients change as the crop develops. Crop growth is reduced by stresses of low water, low N, and low or high temperatures (Singh et al., 1998). Water stress develops when evapotranspiration exceeds the uptake of available water in the soil. Available water is simulated from a one-dimensional model of soil where saturated and unsaturated flow is allowed. Nitrogen stress occurs when potential N uptake exceeds that available in the soil. Available N is modified by fertilizer and crop residue application, leaching, denitrification, NH3 volatilization, immobilization to organic matter, and mineralization from organic matter. Temperature stress develops when air temperature is above or below the optimum of 28°C.

With the potential expansion of taro production into former sugarcane lands, there is a need for an increased understanding of how taro growth and yield are impacted by changes in temperature, total solar radiation, and rainfall due to site or season. The objectives of this research were to (i) determine the growth and yield of rainfed, upland-grown taro at two sites differing in elevation and at four planting dates and (ii) validate an aroid simulation model. Hypotheses tested by this research were as follows: (i) The Low elevation site would have greater yields and shorter duration to maturity because of greater total solar radiation and higher temperatures compared with the High elevation site; (ii) the High elevation site would have greater yields due to higher rainfall compared with the Low elevation site; (iii) Spring and Summer plantings would have greater yields and shorter duration to maturity compared with Fall and Winter plantings because of greater total solar radiation and higher temperatures during the early growth phase; and (iv) the crop simulation model, SUBSTOR-Aroid, could adequately predict yield and maturity date of taro.


   MATERIALS AND METHODS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Description of Sites
Taro cultivar Bun-long was grown under rainfed conditions at two sites differing in elevation (90 and 335 m) along the Hamakua Coast of Hawaii (19°53' N, 155°7' E). Both fields were formerly in sugarcane production and selected as representative of the range in elevation of former sugarcane lands along the Hamakua Coast of Hawaii. The soil series at the Low site is a Hilo silty clay loam (hydrous, isohyperthermic Acrudoxic Hydrudands). The soil series at the High site is an Akaka silty clay loam (hydrous, isomesic Acrudoxic Hydrudands). Weather data were collected at both sites using automated dataloggers (CR10X, Campbell Sci., Logan, UT)1 with temperature sensors, tipping-bucket rain gauges, and quantum sensors. Daily maximum and minimum air temperatures, total solar radiation, and rainfall were recorded.

Soil Characteristics
Initially, soil from each site was sampled at two depths (0–15 and 15–30 cm) and analyzed at the University of Hawaii's Agricultural Diagnostic Service Center for pH, total N by a micro-Kjeldahl method (Isaac and Johnson, 1976), organic C by the method of Heanes (1984), available P by a modified Truog method (Ayers and Hagihara, 1952), and exchangeable cations by the ammonium acetate (pH 7) method (Thomas, 1982). Based on soil pH (Table 1), rates of lime were calculated to raise pH to approximately 6.0.


View this table:
[in this window]
[in a new window]
  		Table 1. Soil total N, organic C, dilute sulfuric acid extractable P, and exchangeable cations of unamended soil (on an oven-dried soil basis) at Low and High sites at two depths.

 
After liming but before planting, further soil samples were taken for the Winter, Spring, Summer, and Fall plantings at six depths (0–15, 15–30, 30–45, 45–60, 60–75, and 75–90 cm), and pH and soil moisture were determined. In addition, soil NO3–N and NH4–N were extracted with 2 M KCl and determined using the copperized Cd reduction method and the indophenol blue method, respectively (Keeney and Nelson, 1982).

Soil data for bulk density, soil surface color, drainage, and permeability were estimated from previous results of the Hilo and Akaka series (Soil Conservation Service, 1976). Volumetric soil water content at field capacity (-0.01 MPa) and permanent wilting point (-1.5 MPa) for each horizon was derived from an empirical function of sand content, organic matter content, bulk density, and gravimetric water content at -1.5 MPa (Legowo, 1987).

Land Preparation
Both sites were deep-plowed, plowed, and tilled before planting. Lime at 4.5 Mg ha-1 of CaCO3 equivalents composed of 20% dolomite and 80% crushed coral (86% CaCO3) was broadcast and tilled in. Phosphorus, as triple superphosphate, was banded into rows at 670 kg P ha-1, based on the recommended rate from a preliminary fertilization trial conducted along the Hamakua Coast of Hawaii island (Sato et al., 1990b). A high P rate was applied because these volcanic ash soils (e.g., Akaka, Hilo, and Kaiwiki series) are known to have a high P-fixing capacity (Hue et al., 2000).

Experimental Treatments
Treatments consisted of four planting dates (Winter, Spring, Summer, and Fall) on 27 February, 28 May, 27 August, and 24 November 1992. The experiment was installed at each site as a randomized complete block design with four treatments (planting dates) and four blocks. Taro was planted from vegetative propagating materials consisting of upper 0.5 cm of corm and lower 24 cm of petiole. Spacing of plants was 0.3 by 0.9 m. Plot size was 56.7 m2 containing 210 plants.

Within 1 wk of planting, a pre-emergent herbicide, oxyfluorfen [2-chloro-1-(3-ethoxy-4-nitrophenoxy)-4-(trifluoromethyl) benzene] (Goal 1.6 E, Rohm and Haas Co., Philadelphia, PA), was sprayed at 0.56 kg a.i. ha-1 to control weeds. Nitrogen, as urea, was broadcast at 0, 1, 2, 3, 4, and 5 MAP for a total rate of 1550 kg N ha-1. Potassium, as muriate of potash, was broadcast at 0, 1, 2, 3, 4, and 5 MAP for total rate of 2010 kg K ha-1. In an earlier taro study on the drier island of Molokai, maximum corm yield occurred at 960 kg N ha-1, with a slight decrease in yield at 1340 kg N ha-1 (Silva et al., 1992). A high N rate was applied in this study because large losses of N fertilizer due to leaching and denitrification were expected at the two Hamakua Coast sites with annual rainfall exceeding 3000 mm.

Taro root aphid, Patchiella reaumuri Kaltenbach, infested a portion of the field at the High site. Because crop losses due to this pest of up to 75% have been reported for upland-grown taro (Sato et al., 1990a), dimethoate [O, O-Dimethyl S-(N-methylcarbamoylmethyl) phosphorodithioate] (Dimethoate 2.67 EC, Platte Chem. Co., Fremont, NE) was sprayed monthly at 3.5 L ha-1 per application to control these pests from February through November 1993.

Biomass harvests were conducted at 1, 3, 5, 7, 9, 11, and 13 MAP. Fresh and dry weights of leaf blades, petioles, and corms were determined on 10 main plants as well as their attached suckers. Roots were sampled from two plants, and fresh and dry weights were determined. Stump diameters were measured at the junction between the main corm and petioles. Leaf areas and root lengths were measured using a digital image analysis system (Decagon Devices, Pullman, WA). Leaf area index was calculated using leaf areas of both main and sucker plants.

Actual corm yield was determined as fresh weight of corms (both healthy ones and diseased ones with rotten portions removed) per plot area. Incidence of corm rot was calculated as the number of corms with rot relative to the total number of corms. Fungal pathogens from corms with rot were isolated and identified courtesy of Dr. Wen-Hsiung Ko, Department of Plant and Environmental Protection Sciences, University of Hawaii.

Potential Corm Yield
Losses due to corm rots could reach 30% under upland conditions (Miyasaka et al., 2001). Because the SUBSTOR-Aroid model (Singh et al., 1992) doesn't include losses due to disease, potential corm yield was calculated as average per plant fresh weight of undiseased corms multiplied by total number of plants per hectare.

Statistical Analysis
Analysis of variance was conducted using SAS programs (SAS Inst., 1982). The experiment was analyzed as a split-split plot design with main plot of site, subplot of planting date, and sub-subplot of harvest date. Single degree-of-freedom contrasts (i.e., preplanned contrasts based on our hypotheses) were estimated for Spring vs. Summer and for Winter vs. Fall. If no significant differences between these planting dates were found, then single degree-of-freedom contrasts were estimated for Spring + Summer linear trends, Spring + Summer quadratic trends, Winter + Fall linear trends, and Winter + Fall quadratic trends. A probability level of <=0.05 was considered to be statistically significant.

Crop Model
The SUBSTOR-Aroid model v. 3.5 (Singh et al., 1998) simulated taro growth and development. Daily maximum and minimum air temperatures, total solar radiation, and rainfall for 1992–1993 were formatted according to model standards (Tsuji et al., 1994). Data on initial conditions for pH, soil N, and water content were formatted for input into the model. A root factor in the model modifies root proliferation in a horizon based on root abundance, pH, and Al content (Table 2).


View this table:
[in this window]
[in a new window]
  		Table 2. Volumetric water content (øv), bulk density, and initial soil conditions at the Low (Hilo series) and High (Akaka series) sites on Hawaii island. Initial conditions are for the Spring planting.

 
Management practices for this study included four planting dates, plant density, planting material weight, zero irrigation applied, and N fertilization rates. Development and growth coefficients needed to simulate the growth and yield of the taro cultivar Bun-long were obtained from Singh et al. (1998).

Development coefficients for cultivar Bun-long of degree days to initiate root formation, first leaf emergence, establishment, rapid vegetative growth, and rapid corm and cormel growth are shown in Table 3. Coefficient A is directly proportional to sucker number and cormel dry weight in the following two equations:

where Y = sucker number per plant; A = coefficient that defines sucker number, unitless (Table 3); B = current number of leaves where Leaf 1 starts at planting; C = number of leaves at end of juvenile phase where Leaf 1 starts at planting and

where Z = amount of carbohydrate partitioned to suckers, g plant-1; D = amount of carbohydrate produced today, g plant-1; F = N stress factor, range 0 (high stress) to 1 (no stress), unitless; and G = soil water deficit factor, range 0 (high stress) to 1 (no stress), unitless.


View this table:
[in this window]
[in a new window]
  		Table 3. Development and growth coefficients for simulating cultivar Bun-long growth and yield.

 
The maximum leaf area coefficient defines the maximum potential area for individual leaves relative to the leaf area of a theoretical plant. The theoretical reference plant has a first leaf with an area of 17 cm2, and 15th leaf area (the largest leaf) is 3200 cm2. Individual leaf areas for specific cultivars are multiples of the reference plant leaves. For example, when the leaf area coefficient is 1.5, the first and 15th leaf areas are 39 (1.5 x 17) and 7200 (1.5 x 3200) cm2. The minimum dry weight–partitioning fraction to the corm refers to the minimum fraction of biomass relative to the whole plant that is allocated to the corm.

Model performance was evaluated according to the criteria of Willmott (1981). Using a linearly regressed curve of simulated yield on observed yield, the root mean square error (RMSE), systematic RMSE [RMSE(s)], and unsystematic RMSE [RMSE(u)] were calculated as follows:

where Pi = simulated value, Oi = observed value, and i = a + , where a = y-intercept of linear regression and b = slope of linear regression. The RMSE is a measure of the total deviation of simulated yield from observed yield. The RMSE(s) measures the deviation of the simulated yield from the observed yield attributed to systematic bias when the linear regression equation has a y-intercept other than 0 and slope other than 1. The RMSE(u) is the random deviation about the regression line, or the random component of RMSE. The square of RMSE is equal to the sum of the square of RMSE(s) and the square of RMSE(u).

An index of agreement (d) was calculated as follows (Willmott, 1981):

where i = Pi - , i = Oi - , and = Oi/n. The index of agreement is a measure of how well the observed deviations about the observed mean value match the predicted deviations about the same observed mean value. The value of the index of agreement ranges from 0 to 1, where 1 indicates that the observed and simulated values are highly related.

Sensitivity analysis was conducted on both the partitioning coefficient to the corm and corm growth duration that are specified in the model for aroid species (not the coefficients specified for cultivar in Table 3). These parameters were increased by 10% and the simulation model run for both sites and all planting dates.


   RESULTS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Environmental Conditions
Before soil amendment, the Low site had higher soil pH, extractable P, and exchangeable Ca and lower total N and organic C compared with the High site (Table 1). At planting, the High site had higher volumetric water content relative to the Low site (Table 2).

At the Low site, maximum and minimum temperatures averaged 1.4 and 2.4°C greater than those at the High site, respectively (Fig. 2) . The Low site received an average of 2.48 MJ m-2 d-1 more solar radiation than the High site. During January to July 1992, total rainfall at the High site was 1073 mm greater than the Low site. Data were regressed linearly to determine the relationship between rainfall at the Low site and the High site. After July 1992, the rain gauge misfunctioned, and rainfall at the High site was estimated based on this linear regression equation.



View larger version (25K):
[in this window]
[in a new window]
  		Fig. 2. Daily maximum (unfilled symbols) and minimum (filled symbols) air temperatures and daily total solar radiation averaged monthly during 1992 to 1993 at two sites differing in elevation on the island of Hawaii. Total monthly rainfall in 1992 to 1993 at two sites on the island of Hawaii; however, after July 1992, the rainfall gauge misfunctioned at the High site.

 
Fresh Weight of Corms
Potential fresh weight corm yield (extrapolated only from healthy corms) increased significantly (P = 0.0001) between 1 and 13 MAP. There were no significant effects of site or planting date, so potential corm yields were averaged across treatments and regressed against MAP (Fig. 3) . Averaged potential fresh weight corm yields increased linearly between 1 and 13 MAP. Similar results were found for potential corm dry weights (data not shown).



View larger version (16K):
[in this window]
[in a new window]
  		Fig. 3. Potential and actual fresh weight corm yields in Mg ha-1 as affected by months after planting (MAP). For potential corm yields (PCY): PCY = -2.0 + 3.1 x MAP, r2 = 0.79. For actual corm yields (ACY) in Spring + Summer plantings: ACY = 0.3 + 2.1 x MAP, r2 = 0.69. For actual corm yields (ACY) in Fall + Winter plantings: ACY = 0.32 + 1.6 x MAP, r2 = 0.52. Error bars are standard errors of mean.

 
Actual fresh weight corm yield (calculated from both healthy corms and diseased corms with rotten portions removed) increased significantly (P = 0.0001) between 1 and 13 MAP. Planting date significantly (P = 0.005) affected actual corm yields, in which Spring + Summer plantings had a significantly greater (P = 0.0006) actual corm yield averaged across sites and MAP than that of Fall + Winter plantings. There was no significant effect of site on actual corm yields; however, there were significant interactions in site x planting date (P = 0.04) and site x planting date x MAP (P = 0.01). Site influenced the effect of planting date on actual yield, with Winter plantings having the lowest yields at the Low site and Fall plantings having the lowest yields at the High site (data not shown). Because single degree-of-freedom contrasts showed no significant difference in slopes over time between Fall and Winter plantings or Spring and Summer plantings, the actual yields of Fall and Winter plantings were averaged, those of Spring and Summer plantings were averaged, and then each was regressed against MAP (Fig. 3). Part of the site x planting date x MAP interaction was due to the significantly greater (P = 0.01) linear increase in yield over time of Spring + Summer plantings relative to that of Fall + Winter plantings.

Incidence of Corm Rot
Corm rots were found to be caused by P. colocasiae and by secondary infections following injury by pests such as ginger maggot (Eumerus figurans Walker) (W.S. Ko, personal communication, 1993; University of Hawaii Agricultural Diagnostic Service Center, personal communication, 1993). Percentage of corm rot increased significantly between 1 and 13 MAP (Table 4). Planting date significantly affected incidence of corm rot. Single degree-of-freedom contrasts showed that Fall + Winter plantings had a significantly (P = 0.004) greater incidence of corm rot averaged over site and time relative to Spring + Summer plantings.


View this table:
[in this window]
[in a new window]
  		Table 4. Incidence of corm rot as affected by site, planting date, and month after planting (MAP)

 
There was no significant effect of site on incidence of corm rot; however, significant interactions of site x planting date and planting date x MAP were found (Table 4). Site influenced the effect of planting date on incidence of corm rot, with the Winter planting at the Low site having the greatest incidence of corm rot, whereas the Fall planting at the High site had the greatest incidence of corm rot. Single degree-of-freedom contrasts showed no significant difference in linear increase over time of percentage corm rots between Fall and Winter plantings. However, a significant difference was found between Spring and Summer plantings (P = 0.002) due to a surprisingly large incidence of corm rots at 1 MAP in Summer plantings.

Dry Weights of Main and Sucker Plants
To show the pattern of dry matter allocation within the main plant, dry weights of main-plant components were averaged across site and planting date and plotted against harvest dates (Fig. 4) . Dry weights of leaf blades and petioles reached a maximum at 5 MAP and then declined. Dry weight of roots reached a maximum at 7 MAP and then declined. Dry weight of corms increased nearly linearly over time between 1 and 13 MAP, and no maximum value was achieved. Total dry weight of main plants increased dramatically between 1 and 5 MAP and then increased more slowly between 5 and 13 MAP but did not reach a maximum.



View larger version (21K):
[in this window]
[in a new window]
  		Fig. 4. Dry weights of various plant components from the main plant and suckers of taro averaged across site and planting date. Error bars are standard errors of mean.

 
To show the pattern of dry matter allocation within sucker plants, dry weights of sucker plant components were averaged across sites and planting dates and plotted against harvest dates (Fig. 4). Dry weights of leaf blades and petioles increased to a maximum at 11 MAP and then declined. Dry weight of cormels increased exponentially from 3 to 13 MAP. Dry weight of cormels exceeded that of corms at 9 MAP. Total dry weight of sucker plants increased in a similar pattern to that of cormels.

Leaf Area Index and Stump Diameter
Leaf area index of both main plant and suckers (P = 0.0001) and corm diameter near the petiole (i.e., stump diameter) increased significantly (P = 0.0001) to a maximum with increasing MAP, and then they declined. There was no significant site effect. Planting date significantly affected LAI (P = 0.004) and stump diameters (P = 0.01); however, significant planting date x MAP interactions were found for LAI (P = 0.0001) and stump diameter (P = 0.0001).

Since single degree-of-freedom contrasts showed no significant differences in linear or quadratic increases in LAI or stump diameter between Fall and Winter plantings or between Spring and Summer plantings, LAI and stump diameters were averaged for Fall and Winter plantings and then for Spring and Summer plantings, and then each was regressed against MAP (Fig. 5) . The quadratic changes in LAI (P = 0.0001) and stump diameter (P = 0.0001) as MAP increased were significantly different for Fall + Winter plantings compared with Spring + Summer plantings. Between 1 and 3 MAP, LAI and stump diameter increased more slowly for Fall + Winter plantings compared with Spring + Summer plantings, perhaps due to colder temperatures (Fig. 2) and lower total solar radiation (Fig. 2) in winter months. Then between 3 and 7 MAP, LAI and stump diameter of the Fall + Winter plantings increased dramatically during the spring or summer months. In contrast, LAI and stump diameter of the Spring + Summer plantings leveled off or even decreased slightly between 3 and 7 MAP during the fall or winter months.



View larger version (17K):
[in this window]
[in a new window]
  		Fig. 5. Leaf area index (LAI) of both main plant and suckers, and stump diameter of main plant in cm as affected by months after planting (MAP). For Spring + Summer plantings: LAI = 0.039 + 0.612 x MAP - 0.030 x MAP2, r2 = 0.38; stump diameter (STDIA) = 3.13 + 0.89 x MAP - 0.053 x MAP2, r2 = 0.50. For Fall + Winter plantings: LAI = -1.01 + 1.10 x MAP - 0.070 x MAP2, r2 = 0.57; STDIA = 1.06 + 1.49 x MAP - 0.089 x MAP2, r2 = 0.65. Error bars are standard errors of mean.

 
Percentage Dry Matter in Corms
Dry matter percentage in the corm is one measure of quality. Percentage dry matter in corms tended to increase significantly from 1 MAP until 5 to 13 MAP (Table 5). A significant effect of site was found, in which corms grown at the Low site had a significantly higher percentage dry matter averaged across planting date and MAP compared with those grown at the High site, perhaps due to the lower rainfall (Fig. 2) and reduced soil moisture at the Low site (Table 2). A significant effect of planting date was found, in which corms from Spring + Summer plantings had a significantly higher percentage dry matter than those from Fall + Winter plantings (P = 0.001). Significant interactions of site x MAP and planting date x MAP were found.


View this table:
[in this window]
[in a new window]
  		Table 5. Percentage dry matter of corms as affected by site, planting date, and months after planting (MAP)

 
Comparison of Predicted and Actual Yields
The SUBSTOR-Aroid crop model generally did not simulate well the observed potential dry corm yields (Fig. 6A) . The range in observed values was less than that of simulated values. Dry corm yield from the simulation ranged from 5556 to 10 622 kg ha-1, whereas the values from the actual experiment ranged from 7643 to 9254 kg ha-1. The model did not mimic well the greater potential yields observed when the latter stage of corm growth occurred during the warmer, higher-solar-radiation period of summer compared with the cooler, lower-solar-radiation period of winter. The RMSE and d values were low (Table 6), indicating that observed and simulated corm yields were not closely related. Because RMSE(s) is larger than RMSE(u), a systematic bias in the model plays a major role in preventing a better match between observed dry corm yield and simulated yield.



View larger version (19K):
[in this window]
[in a new window]
  		Fig. 6. Comparison of observed potential and simulated dry weight corm yields of taro cultivar Bun-long grown at two sites and four planting dates on Hawaii island based on (A) simulated time to harvest maturity using 28th leaf stage and (B) actual days to harvest at 11 months after planting. Error bars indicate one standard deviation of four replicates.

 

View this table:
[in this window]
[in a new window]
  		Table 6. Evaluation of model performance for potential dry corm yield, leaf area index (LAI), and days to maturity when the model is allowed to estimate time to maturity, and evaluation of model performance for dry corm yield when harvest date is predetermined.{dagger}

 
The SUBSTOR-Aroid model generally underestimated the length of time needed to mature by 50 d (Fig. 7) . Like the dry corm yield simulations, the RMSE(s) was larger than the RMSE(u) (Table 6), indicating a systematic bias in the model.



View larger version (24K):
[in this window]
[in a new window]
  		Fig. 7. Comparison of observed and simulated days to maturity for taro cultivar Bun-long grown at two sites and four planting dates on Hawaii island. There are no error bars because biomass harvests were conducted within 1 d.

 
In contrast to difficulties in simulating dry corm yields and duration to maturity, the SUBSTOR-Aroid model simulated LAI fairly well (Fig. 8 ; Table 6). With the exceptions of the Fall planting at the High site and the Winter planting at the Low site, observed and simulated values of LAI at the 11th leaf stage were similar. The RMSE(s) was greater than RMSE(u), indicating that most of the error was systematic, similar to the results for corm dry weight and time to maturity (Table 6).



View larger version (20K):
[in this window]
[in a new window]
  		Fig. 8. Comparison of observed leaf area index (LAI) and simulated LAI at the 11th leaf stage for taro cultivar Bun-long grown at two elevations and four planting dates.

 
During sensitivity analysis of aroid parameters, an increase in the partitioning coefficient did not increase corm dry weight, probably because this coefficient was an upper limit of partitioning to the corm that was never exceeded during the simulations. In addition, increasing corm growth duration by 10% increased average corm dry weight by only 5%, indicating that corm dry weight was not particularly sensitive to changes in this model parameter.

To investigate whether improving the accuracy of the simulated time to maturity would improve the dry corm yield estimation, the model was rerun under the same conditions as initially run except that the simulation was stopped on the actual harvest date of 11 MAP. With this modification to harvest maturity, the model generally overestimated the dry corm yield (Fig. 6B). However, although length of time to mature was eliminated as a source of bias, the dry corm yield still was not estimated well by the model, based on high values of RSME and low values of d (Table 6). The value for d did increase slightly relative to the initial run, indicating that model predictions improved to a small degree.


   DISCUSSION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
Yield of Taro Affected by Planting Date
Potential yields of taro (based on fresh weight of healthy corms) were not significantly affected by planting date (Fig. 3). This result was not consistent with our hypothesis that Spring + Summer plantings would have greater yields and an earlier time to maturity because of greater average temperatures and total solar radiation (Fig. 2) during the period of rapid root and shoot development compared with those of Fall + Winter plantings.

In contrast, Spring + Summer plantings had significantly greater actual fresh weight yield of corms (based on both healthy corms and diseased corms with rotten portions removed) (Fig. 3) and lower incidence of corm rots (Table 4) compared with Fall + Winter plantings. Not surprisingly, the site and planting date treatments with the lowest yields also had the highest percentage corm rots. These results indicate that the primary reason for reduced actual corm yields for Fall + Winter plantings was the greater average incidence of corm rots in these plantings (Table 4) and not lower temperature or total solar radiation during the early growth phase.

The reason for the greater incidence of corm rots in the Fall + Winter plantings (Table 4) is unknown. Perhaps, warm and wet conditions during the corm maturation phase (>5 MAP) were more conducive to insect development or microbial growth, increasing disease incidence. Interestingly, in another field trial, taro had a very low incidence of corm rots between 7 and 11 MAP when rainfall was insufficient to replace 100% of evapotranspiration (Miyasaka et al., 2001). Potential corm yield was 36% greater than averaged actual yield at 13 MAP (Fig. 3), indicating that pests causing corm rots are a major factor limiting taro yields under upland culture.

Growth of Taro Affected by Site
Despite differences in total solar radiation and temperature (Fig. 2), the two sites showed no significant difference in potential or actual fresh weight corm yields (Fig. 3) although site could influence the effect of planting date on yield. Such yield data indicates that taro can be grown across the range of elevations of former sugarcane lands along the Hamakua Coast of Hawaii island. These results are contrary to our two hypotheses that either the Low site would have greater yields and shorter duration to maturity due to warmer temperatures and greater total solar radiation or that the High site would have greater yields due to higher rainfall. Perhaps, one reason for the lack of growth response of the main plant to greater total solar radiation at the Low site is due to light saturation of photosynthesis at low irradiance levels. Sato et al. (1978) showed that photosynthetic rates of mature taro leaf blades reached a maximum at 20 to 30% of full sunlight. Another possible reason for the lack of growth differences due to varying total solar radiation is the adaptation of taro to shade conditions. Taro grown at 30% of full sunlight had increased stomatal and chlorophyll density, presumably increasing photosynthetic efficiency at low light levels (Johnston and Onwueme, 1998; Onwueme and Johnston, 2000).

In addition, taro cultivar Bun-long appeared to respond to increased total solar radiation and temperature by increasing biomass allocation to the commercially unimportant sucker plant. The two sites did not differ in dry matter of any plant component, except for the greater leaf blade and cormel dry weights of sucker plants grown at the Low site relative to the High site (data not shown).

The lack of response of fresh weight corm yields to sites varying in elevation are in contrast to those of Prasad and Singh (1992), who observed an increase in corm dry weight of cultivar Bun-long as elevation decreased from 640 to 282 m. However, maximum and minimum air temperatures were higher for the Hawaii island sites in our study compared with the Maui island sites of Prasad and Singh (1992), indicating that temperatures at sites in our study were not low enough to retard taro growth.

Continuous Partitioning in Taro
According to Loomis and Rapoport (1977), there are two types of partitioning in root and tuber crops: (i) continuous partitioning in which the storage organ growth begins early in seedling stage and continues throughout the vegetative period of plant growth [e.g., sugarbeet (Beta vulgaris L.)] and (ii) phasic partitioning in which early vegetative growth is characterized by shoot and fibrous root development with storage organ growth beginning later [e.g., potato (Solanum tuberosum L.) or cassava (Manihot esculenta Crantz)]. It is clear that taro exhibits continuous partitioning with near-linear increases in fresh or dry weights of corms between 1 and 13 MAP (Fig. 3 and 4). These results are similar to those found for taro grown under upland conditions in Hawaii (Singh et al., 1992), Fiji (Sivan, 1982), Puerto Rico (Goenaga, 1995), and Western Samoa (Reynolds, 1977) but are contradictory to those reported by Ching (1970) for upland-grown taro in Hawaii. Ching (1970) reported that taro exhibited phasic partitioning, but this erroneous conclusion was probably due to lack of sampling until 5 MAP.

Growth Phases of Taro
In our field study, maximum dry weight of petioles, leaf blades, and roots occurred between 5 and 7 MAP (Fig. 4), similar to the third growth stage postulated by Reynolds (1977). However, contrary to Reynolds (1977), no decline in dry matter production of corms was observed between 1 and 13 MAP.

Dry matter accumulation of corms appeared to be exponential in Western Samoa (Reynolds, 1977) and Fiji (Sivan, 1982). However, in our field study conducted in Hawaii, both fresh and dry weights of corms increased nearly linearly between 1 and 13 MAP (Fig. 3 and 4). Similar results were found at another location in Hawaii (Singh et al., 1992) and in Puerto Rico (Goenaga, 1995). This observed difference in dry matter accumulation of corms could be due to differences in cultivars, weather conditions, fertilization rates, and soil moisture. In Hawaii (Singh et al., 1992) and Puerto Rico (Goenaga, 1995), the near-linear dry matter accumulation of corms occurred in plants grown under irrigated conditions with ample nutrients. Similarly, in our field study, high levels of nutrients were supplied, and rainfall was adequate throughout most of the 2 yr of this study (Fig. 2).

Growth of sucker plants in cultivar Bun-long commenced by 3 MAP and increased almost exponentially over time (Fig. 4). This cultivar is known to be a high-tillering cultivar with a greater proportion of assimilates funneled to the cormels (Singh et al., 1992). Typically, however, upland-grown cormels are not sold commercially due to their small size; thus, the large proportion of photosynthate partitioned to cormels is not economically productive.

Leaf Area Index
Leaf area index of Spring + Summer plantings reached a maximum of 3.8 at 11 MAP while LAI of Fall + Winter plantings reached a maximum of 3.4 at 5 to 7 MAP (Fig. 6). In Puerto Rico, maximum LAI of both main plant and suckers of two cultivars was 2.2 at 4 MAP (Goenaga, 1995). In Fiji, maximum LAI of the main plant alone ranged from 2.0 to 3.0 at 4 to 5 MAP, depending on cultivar (Sivan, 1982). The differences in LAI observed between our study and those of other investigators could be due to cultivar or to the lower N and K fertilization rates in Puerto Rico and Fiji compared with those in Hawaii.

Definition of Crop Maturity
Using the definition of crop maturity as achieving the average yield of 14 Mg ha-1 (Hawaii Agric. Stat. Serv., 1994), the taro plants matured at 6.4 mo for the Spring + Summer plantings and at 8.3 mo for the Fall + Winter plantings (Fig. 3). Alternatively, based on the definition of crop maturity as reaching maximum yield, the Fall + Winter plantings matured at 11 MAP (Fig. 3); however, the Spring + Summer plantings didn't reached maturity at 13 MAP. Perhaps, the yield of taro corms for the Spring + Summer plantings didn't decline even at 13 MAP because they were not N, temperature, or water stressed due to high N rates, warmer temperatures during the senescence period, and sufficient rainfall. In contrast, during a cropping cycle with insufficient rainfall from 8 to 10 MAP, taro in unmulched plots declined in yield between 7 and 11 MAP (Miyasaka et al., 2001).

Using the definition of crop maturity as achieving acceptable corm quality of 20% dry matter (S.C. Miyasaka, unpublished data, 1998), plants in all treatments except the Spring planting at the High site matured at 5 MAP (Table 5). Alternatively, based on the definition of maturity as a reduction in corm diameter near the petiole (i.e., stump diameter), plants in all treatments matured at 9 MAP (Fig. 6).

Based on the definition of maturity as cormel dry weight exceeding corm dry weight, taro plants matured at 9 MAP in this field study (Fig. 4). In contrast, Singh et al. (1992) simulated corm dry weight of cultivar Bun-long to equal dry weight of cormels at approximately 11 MAP. Perhaps, the high N fertilization rate in our study magnified the competing sink behavior of cormels. Alternatively, using the definition of maturity as reaching the 28th leaf stage, plants matured at approximately 11 MAP (data not shown).

Comparing the above-estimated times to crop maturity (e.g., 5–13 MAP), it is not possible to define an exact duration to crop maturity for taro. Under upland conditions in Hawaii, Western Samoa, or Fiji, taro is often harvested between 9 and 11 mo (Plucknett and de la Pena, 1971; Reynolds, 1977; Sivan, 1982). There is an economic advantage to this indeterminant nature of taro in that farmers can wait for optimal market prices before harvesting. The decision to harvest will depend on the market, incidence of pests, and soil and weather conditions that can cause declines in corm yield. In our study, the high fertilization rates with ample moisture probably lengthened the time before corm yields reached a maximum and then declined. A future experiment will examine the effect of high N fertilization rates on crop maturity.

Shortcomings of the SUBSTOR-Aroid Model
The SUBSTOR-Aroid model did not simulate the maturity date well, but this problem can be corrected. The error analysis indicated that the bias is predominantly additive, i.e., the model underpredicts maturity date by 50 d (Fig. 8). Because maturity in the observed data and the simulation was estimated by the appearance of the 28th leaf, the cause of the constant bias may be due to an incorrect parameterization of the coefficients that define the time between the appearance of successive leaves (e.g., PHINT, TIPINT, and TIPGRAD). Recalibrating these coefficients should reduce this bias.

Also, there may be additional factors aside from temperature stress that slow the actual leaf appearance rate, such as water deficit, low irradiance, and nutrient stress, which are not included in the model. For example, increasing drought stress on cowpea [Vigna unguiculata (L.) Walp] slowed the leaf appearance rate (Jamadagni et al., 1995). Also, soybean [Glycine max (L.) Merr.] leaf appearance rate decreased with decreasing photosynthetic photon flux density and decreasing N concentration in irrigation water (Snyder and Bunce, 1983). In our study, if high N fertilization increased the leaf appearance rate of taro similar to its effect on soybean, then the model would have overpredicted maturity date, whereas the opposite effect was observed.

The SUBSTOR-Aroid model contains a systematic bias that prevents successful prediction of potential dry weight yield of corms (Table 6). One possible explanation for the lower-than-expected yield from the model (Fig. 7A) is that the simulation is sink limited. The LAI comparison suggests that the model simulates well the source of photosynthate. However, the model didn't simulate well the greater corm yield observed in planting dates with the latter growth stage occurring in the summer months, indicating that the partitioning of dry weight to the corm may be too low during the fourth growth phase of Reynolds (1977).

In addition, the model appears to be more sensitive to some parameters than the real system because the range in predicted potential dry weight of corms was greater than that observed (Fig. 7A). One possible reason is that taro is capable of maintaining photosynthetic rates at a low irradiance level (Johnston and Onwueme, 1998; Onwueme and Johnston, 2000). Other mechanisms that ameliorate yield reduction due to environmental stress have been observed often in other crops. For example, under drought stress, wheat (Triticum aestivum L.) plants remobilized stem carbohydrate reserves to the grain, accounting for up to 12% of the final grain weight (Rawson and Evans, 1971). Pigeonpea (Cajanus cajan L.) plants under drought stress maintained yield through leaf osmotic adjustment that sustained radiation use efficiency and leaf duration (Subbarao et al., 2000). Perhaps the taro plant too has mechanisms that enable it to maintain a relatively constant flow of carbohydrates to the corm despite environmental stress. A better understanding of taro physiological responses to environmental stress is needed for modification of this aroid model to better simulate upland taro production.

We attempted to remove the detrimental effects of corm rots by measuring only undiseased corms for potential yields; however, it is possible that this estimation was insufficient to correct for yield reduction due to pests. The SUBSTOR-Aroid model doesn't include reductions in yield due to pest problems. Based on estimated reductions in yield of approximately 36% due to corm rots (Fig. 3), future model development needs to incorporate interactions between plant growth and incidence of corm disease.


   ACKNOWLEDGMENTS
 
The authors thank Dr. Charles E. McCulloch of the Department of Epidemiology and Biostatistics, University of California–San Francisco, for assistance in statistical analysis. We acknowledge the invaluable work of Dr. Upendra Singh of the International Fertilizer Development Center and Hemant Prasad, former graduate student, who developed the original SUBSTOR-Aroid model. Also, we thank Dr. Singh and Dr. Paul Wilkens, both of the International Fertilizer Development Center, who updated the SUBSTOR-Aroid model and allowed us to test its most recent version. In addition, we thank the following individuals: George Hirowatari, Jr., and the late Shigeru Kansako for generous donations of taro propagating materials; Mauna Kea Agribusiness for making leased land available free of charge; Dennis Ida, Farm Manager of Waiakea Experiment Station, for the scientific illustration of a taro plant; Dennis T. Matsuyama, Research Associate of Beaumont Research Station, for collection of weather data and assistance in biomass measurements; and Lane Matshushita, Jon Katada, Angel Magno, and other staff members at the Waiakea Experiment Station for their help in planting and harvesting.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Contrib. from the College of Trop. Agric. and Human Resour., Journal Ser. no. 4629.

1 Mention of a trade name, proprietary product, or vendor does not constitute an endorsement or warranty by the University of Hawaii, nor does it imply its approval to the exclusion of other products or vendors that may also be suitable. Back


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Miyasaka, S. C.
Right arrow Articles by Kodani, L. S.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Miyasaka, S. C.
Right arrow Articles by Kodani, L. S.
Agricola
Right arrow Articles by Miyasaka, S. C.
Right arrow Articles by Kodani, L. S.
Related Collections
Right arrow Other Tubers
Right arrow Crop Models