Measurement of Dry Matter and Starch in Modern Cassava Genotypes during Long Harvest Cycles

Abstract

Starch (ST) is the main determinant of cassava root industrial quality and is routinely quantified using the specific gravity (SG) method, which is simple but lacks precision. This study aimed to estimate the dry matter (DM) and ST content in nine modern cassava genotypes using the SG method and to develop updated empirical equations that relate SG values with DM and ST content. Two field experiments were conducted using a randomized block design with nine genotypes, nine to ten harvest periods, and four replicates. The correlation between root SG and DM was stronger than that between root SG and ST; however, harvest seasonality strongly influenced this relationship. Genotypes varied in their earliness of ST accumulation in the roots, and genotype-specific equations estimated root DM and ST with greater accuracy than equations based on pooled data from all genotypes. Furthermore, the equations from the literature were less accurate in estimating the root DM and ST content than the equations generated in this study. Therefore, these equations can be used by processing industries to improve the predicted cassava root DM and ST content; however, it may be necessary to include the percentage of ST that industries cannot remove from fresh cassava roots during processing.

1. Introduction

Brazil is the fifth-largest producer of cassava (Manihot esculenta Crantz) in the world, after Nigeria, the Democratic Republic of Congo, Thailand, and Ghana [1]. Cassava root is an excellent source of carbohydrates [2,3] and can be processed to generate various products that serve as food sources for more than 800 million people worldwide [4].

Starch (ST) is the main component of cassava root and has several industrial applications, including the production of bioenergy and biodegradable plastics [2,5]. In the processing industry, the price paid to producers of fresh cassava roots varies with the ST content [2,5,6,7]. Therefore, fast and accurate assessment of root ST content is critical for the greatest benefit to both producers and industry, as the former will obtain fair pay for their produce, whereas the latter will obtain the highest possible yield and quality of raw material for what they pay [5,6].

The most accurate method for determining ST content involves laboratory analysis [2]. However, this method requires laboratory facilities [8] as well as drying of the samples in an oven for at least 24 h [3,9], followed by milling, preparation, and chemical analysis of the samples via acid or enzymatic hydrolysis, which is a tedious and time-consuming process when large volumes of samples are involved [3,6,9].

A faster, non-destructive,—but less accurate—method for estimating root ST content is the specific gravity (SG) method, which involves weighing roots in the air and then in water [3,8,9,10,11,12]; SG values are then used to estimate dry matter (DM) and ST content using empirical equations. In Brazil, as in many other countries, including India, Thailand, Ghana, and Colombia, processing industries determine cassava root ST content using this method [3,5,6,11,13]. Furthermore, some studies have sought to generate empirical equations to estimate the ST content in cassava root more accurately using SG data [3,8,10,11,12]. Thus, in some situations, the root ST content was estimated by subtracting a constant value of 4.65 from the root DM content estimated by the SG method [14].

Early attempts to associate SG with ST content established that samples should consist of at least 5 kg of fresh cassava roots [8]; however, recent studies suggest that the sample size may be smaller [12]. In general, empirical equations used to estimate root ST content have been generated using limited datasets, old genotypes that were not planted in Brazil, small weight samples harvested very early (6 months of the cycle), or cycles that did not exceed 12 months [3,8,9,10,11,12]. However, in Brazil, it is common to harvest cassava for long periods at different maturity stages ranging from 7 to 24 months [13]. Harvest seasonality affects the dynamics of ST accumulation in the roots [7], during which the cassava plant undergoes two vegetative cycles during the rainy and hot seasons, separated by a period of physiological rest that occurs during the coldest and driest seasons [15,16]. In general, root ST accumulation peaks in the dry and cold seasons and then tends to decrease with the resumption of the new branch and leaf growth [7,11,12], increasing again in the later stages of the annual growing cycle [7]. Younger roots have smaller ST granules, whereas older roots have larger ST granules [17]. However, at the beginning of the second vegetative cycle, cassava roots present a mix of large ST granules (fully formed in the first cycle) and small ST granules that are synthesized and stored in the roots [17]. It is not known whether this change in ST granule morphology during the growing season interferes with the ST content determination using the SG method. Furthermore, ST accumulation prevalence in the storage roots during the growing season can vary among genotypes [7].

Cassava genotypes IAC 14 and IAC 90 are older and are widely planted in the state of São Paulo, Brazil [18]. Recently, however, the cassava breeding program of the Brazilian Agricultural Research Corporation (Empresa Brasileira de Pesquisa Agropecuária—EMBRAPA) released earlier and higher-yielding cassava genotypes [7], such as BRS CS 01, BRS 419, BRS 420, BRS Ocauçú, and BRS Boitatá. These cassava genotypes have been well accepted by cassava farmers in the south-central region of Brazil. However, farmers have repeatedly reported that the ST content predicted by the equations available in the literature and used by Brazilian processing industries do not adequately predict the ST content of cassava root. Furthermore, some studies have claimed that the SG method overestimates the ST content of cassava root [19,20], which could reduce the ease and practicality of this method. Therefore, updating the equations used to estimate DM and ST content in the storage roots of modern cassava genotypes will produce benefits for cassava producers, who will have a more accurate estimate of root quality, as well as for processing industries that will be able to make a more accurate and faster assessment of the DM and ST content of storage roots.

Thus, we hypothesized that the evaluation of a larger number of cassava genotypes for longer periods would generate better-adjusted equations to estimate the DM and ST content of storage roots using the SG method. Thus, this study aimed to estimate the DM and ST content of cassava storage roots using the SG method, with the following specific objectives: (i) to determine the SG values, DM, and ST content in nine modern cassava genotypes during two vegetative growing cycles, and (ii) to develop updated empirical equations that relate the SG values with DM and ST content.

2. Materials and Methods

2.1. Site, Soil, and Climate

The field experiments were conducted at two sites (Site-year 1—SY1: São Pedro do Turvo, SP (22°33′11″ S, 49°46′57″ W; elevation: 525 m); Site-year 2—SY2: Paraguaçú Paulista, SP, 22°34′ S, 50°30′ W; elevation: 523 m) in a cassava cultivation area of Tereos Starches and Sweeteners Brazil S.A., in São Paulo State, Brazil. According to the classification by Köppen, the climate in the region is Cwa type, that is, subtropical with dry winters (temperatures below 18 °C) and hot summers (temperatures above 22 °C). The rainfall and temperatures recorded during cassava cultivation are shown in the supplementary Figure S1.

2.2. Experimental Design and Treatments

The experiments were laid in a randomized block design with nine genotypes (IAC 14, IAC 90, BRS CS 01, BRS 419, BRS 420, BRS Ocauçú, BRS Boitatá, 1097/13, and 2011 02-43), over nine to ten harvest periods for which SG characteristics, and DM and ST content were evaluated, with four replicates. IAC 14 and IAC 90 are older genotypes that were developed by the breeding program of the Instituto Agronômico de Campinas (IAC), São Paulo, Brazil. BRS CS 01, BRS 419, BRS 420, BRS Ocauçú, BRS Boitatá, 1097/13, and 2011 02-43 are modern genotypes developed by Embrapa’s breeding program in Bahia, Brazil. In SY1, harvests were performed 7 (May 2019), 9 (July 2019), 11 (September 2019), 12 (October 2019), 14 (December 2019), 15 (January 2020), 17 (March 2020), 19 (May 2020), and 21 (July 2020) months after planting (MAP). In SY2, harvest was performed on 5 (March 2020), 7 (May 2020), 9 (July 2020), 10 (August 2020), 12 (October 2020), 13 (November 2020), 14 (December 2020), 15 (January 2021), 17 (March 2021), and 19 (May 2021) MAP. Each plot consisted of four 50 m long rows, spaced 0.90 m apart. On each day, six randomly selected cassava plants were harvested from the two central rows within each plot, excluding the edge plants.

2.3. Soil Tillage, Cassava Planting, and Management

The soils at study sites SY1 and SY2 were low-fertility soils (

Table 1. Soil chemical properties of the experimental areas at a depth of 0-0.2 m before cassava planting (n = 4).

Properties	SY1 (São Pedro do Turvo)	SY2 (Paraguaçú Paulista)
pH (CaCl2)	3.8	4.1
Organic Matter (g dm−3)	10.0	10.0
Presin (mg dm−3)	5.0	1.0
K (mmolc dm−3)	0.82	0.91
Ca (mmolc dm−3)	2.0	3.0
Mg (mmolc dm−3)	1.0	1.0
Al (mmolc dm−3)	11.0	4.0
H+Al (mmolc dm−3)	39.0	32.0
Cation exchange capacity (mmolc dm−3)	43.0	36.0
Base saturation (%)	10	12
S (mg dm−3)	11.0	4.0
B (mg dm−3)	0.14	0.35
Cu (mg dm−3)	0.30	0.20
Fe (mg dm−3)	103.0	19.0
Mn (mg dm−3)	3.5	1.4
Zn (mg dm−3)	0.1	0.1

). Cassava was planted in no-tillage areas using recommendations for the region and an adapted mechanical planter [21]. The experimental area was occupied by pasture at site SY1, while site SY2 was occupied by corn residues and weeds. Branches from the middle third of the stems of approximately one-year-old healthy plants were used for planting. The stems were harvested manually using a machete the day before planting from an area reserved for genotype propagation that had been planted in an area close to the experimental sites. Healthy plant stems with >20 mm diameter and without symptoms of pests, diseases, or nutritional deficiencies, were manually selected.

The stems were planted using a plant spacing of 0.9 × 0.6 m. For planting fertilization, 130 kg ha⁻¹ of N-P₂O₅-K₂O 10-50-00 fertilizer formula was applied at both SY1 and SY2. In turn, 45 kg ha⁻¹ N (urea, 45% N) and 60 kg ha⁻¹ K₂O (potassium chloride, 60% K₂O) were applied as topdressing fertilization at 120 days after planting (DAP) in SY1, and SY2, 40 kg ha⁻¹ N and K₂O were applied at 90 DAP using the fertilizer formula N-P₂O₅-K₂O 20-00-20. During the physiological rest phase, which occurred between 9 and 10 MAP (July–August), plant shoots were pruned at a height of 10 cm. Crop management followed the requirements and technical recommendations for the entire cassava growth cycle.

2.4. Storage Root Harvest and Analysis

Six plants were harvested from each plot at each time point. Storage roots were detached from the plant, homogenized, and a 5 kg subsample per sample of fresh roots was transported from the field to the laboratory (Figure 1). The roots were washed to remove excess soil, weighed in air (5 kg), and placed in water to obtain SG [19]. SG = 5 kg/(5 kg FWW) where FWW is the weight measured on a hydrostatic scale after submerging the roots in water. To determine root DM content, the same samples were crushed, weighed (fresh weight), and oven-dried to a constant weight at 65 °C under forced air circulation. The dried samples were weighed again, and the DM content was calculated using the gravimetric method.

Subsequently, the dry samples were ground in a Wiley mill and subjected to ST content analysis. For analysis, 200 mg of the sample was placed in a 125 mL Erlenmeyer flask and 42 mL of distilled water, 100 μL of the alpha-amylase enzyme, and 1 mL of 2 mol L⁻¹ sodium acetate buffer solution at pH 5.35 were added. The samples were heated at 90 °C and agitated for 2 h. Then, the samples were cooled to 50 °C, and 100 μL of amyloglucosidase enzyme was added, and the samples were maintained at 55 °C and agitated for 2 h. The sugar content was quantified using a glucose oxidase/peroxidase reagent, as described by the AOAC [22]. The sugar content was converted to ST content by multiplying the results by a factor of 0.9 [13], which was then converted into ST content on a fresh weight basis (% FW).

The SG, DM, and ST content of roots from SY1 and SY2 were grouped according to plant age, and the data were plotted in graphs with SG values as the independent variables and DM or ST content as the dependent variables. Using this dataset, we generated genotype-specific regression equations to estimate DM and ST content separately for each genotype and a general equation based on pooled data from all genotypes.

Predicted DM and ST content was calculated using individual equations for each genotype and equations combining all genotypes. Predicted DM content was calculated using the following equations from previous studies [8,10,11,12]:

(a)

Empirical equations to obtain DM content

Predicted DM content (%) = 15.75 + 0.0564 × ((weight in water × 3)/5))

Predicted DM content (%) = −124.9 + (142.3 × SG)

Predicted DM content (%) = −142 + (158.3 × SG)

Predicted DM content (%) = −180.22 + (192.89 × SG)

(b)

Empirical equations to obtain ST content

Predicted ST content was calculated using the following equations from previous studies [8,11,12]. For the equation of Kawano et al. [11], the predicted root ST content was obtained by subtracting the constant value of 4.65 from the predicted DM content [14], as follows:

Predicted ST content (%) = predicted DM content − 4.65

Predicted ST content (%) = −147 + (159.1 × SG)

Predicted ST content (%) = −170.3 + (179.36 × SG)

2.5. Statistical Analysis

The SG, DM, and ST content datasets were separated by genotype and harvest time (plant age) and subjected to ANOVA using the SISVAR statistical software package [23]. Means for SG, DM, and ST content were compared using the LSD test at p < 0.05. Using the Pearson test (r) at p < 0.05, the predicted DM or ST content was correlated with the measured DM or ST content. Correlation coefficients were obtained for all collected data using STATISTICA version 6 [24], regardless of the genotype or harvest time.

3. Results

3.1. Effect of Genotype on SG and Measured Root DM and ST Content as Related to Plant Age

Analysis of the dataset showed that SG, root DM, and ST content varied widely among genotypes and harvest periods (

Table 2. Specific gravity of storage root samples from nine modern cassava genotypes harvested at different times after planting (data from SY1 and SY2 were pooled together; n = 8).

MAP (Month)	IAC 14	IAC 90	BRS CS01	BRS 419	BRS 420	BRS Ocauçú	BRS Boitatá	1097/13	2011 02-43	Mean
5 (March)	1.115 abcd	1.120 ab	1.119 ab	1.109 cde	1.124 a	1.104 e	1.112 bcde	1.115 abc	1.106 de	1.114
7 (May)	1.141 abc	1.138 a	1.146 a	1.130 c	1.137 abc	1.146 ab	1.137 bc	1.144 a	1.131 abc	1.139
9 (July)	1.139 ab	1.135 ab	1.146 ab	1.133 b	1.135 ab	1.147 ab	1.143 ab	1.146 a	1.126 ab	1.139
10 (August)	1.142 ab	1.135 bc	1.143 ab	1.126 d	1.136 ab	1.138 ab	1.127 cd	1.144 a	1.119 d	1.134
11 (September)	1.119 ef	1.115 f	1.136 bc	1.130 cd	1.126 de	1.146 a	1.139 ab	1.133 bcd	1.119 ef	1.129
12 (October)	1.120 ab	1.110 abcd	1.111 d	1.102 cd	1.111 abcd	1.118 bcd	1.117 abc	1.121 a	1.102 cd	1.112
13 (November)	1.104 ab	1.101 ab	1.096 bc	1.084 c	1.099 b	1.100 ab	1.101 ab	1.112 a	1.097 bc	1.099
14 (December)	1.120 ab	1.113 ab	1.112 bc	1.106 bc	1.117 ab	1.120 c	1.123 ab	1.116 a	1.114 ab	1.116
15 (January)	1.130 ab	1.132 ab	1.124 b	1.121 b	1.128 b	1.136 ab	1.137 b	1.133 a	1.121 b	1.129
17 (March)	1.143 a	1.139 ab	1.136 b	1.128 ab	1.136 b	1.146 ab	1.143 b	1.145 a	1.128 ab	1.138
19 (May)	1.145 ab	1.143 bc	1.140 bc	1.140 bc	1.138 bc	1.144 c	1.148 c	1.158 a	1.129 bc	1.143
21 (July)	1.159 a	1.166 a	1.156 ab	1.141 c	1.148 bc	1.159 a	1.164 a	1.165 a	1.138 c	1.155
Mean	1.132	1.128	1.133	1.123	1.128	1.135	1.134	1.138	1.120

Values within rows followed by the same lowercase letters are not significantly different by the LSD test (p < 0.05).

and

Table 3. Dry matter (DM) and ST contents of storage root samples from nine modern cassava genotypes harvested at different times after planting (data from SY1 and SY2 were pooled together; n = 8).

MAP (Month)	IAC 14	IAC 90	BRS CS01	BRS 419	BRS 420	BRS Ocauçú	BRS Boitatá	1097 -13	2011 02-43	Mean
	DM content (% FW)
5 (March)	30.9 bc	32.8 b	32.1 b	30.3 bc	36.8 a	32.5 b	30.7 bc	33.3 b	28.3 c	32.0
7 (May)	38.7 a	37.5 a	40.5 a	38.4 a	38.4 a	40.1 a	40.5 a	38.8 a	36.0 a	38.8
9 (July)	39.5 ab	39.1 b	41.2 ab	38.7 b	39.2 b	41.4 a	41.1 ab	39.5 ab	39.7 ab	39.5
10 (August)	40.6 ab	38.0 cd	38.4 bcd	37.5 cd	38.7 bc	39.5 bc	37.6 cd	42.6 a	36.2 d	38.8
11 (September)	34.0 c	34.6 c	39.8 ab	37.1 abc	36.0 abc	40.4 a	40.2 a	35.1 bc	36.3 abc	37.1
12 (October)	35.9 ab	35.8 ab	36.6 ab	34.9 b	34.1 b	37.3 ab	36.2 ab	36.3 a	33.5 b	35.6
13 (November)	33.8 a	32.4 ab	30.7 abc	29.0 bc	28.4 c	33.2 a	31.7 abc	33.9 a	29.1 bc	31.4
14 (December)	33.1 ab	34.0 a	33.8 ab	34.2 a	33.6 ab	32.8 ab	30.3 b	34.0 a	31.7 ab	33.7
15 (January)	37.1 a	37.5 a	37.4 a	38.4 a	38.0 a	37.3 a	39.3 a	38.2 a	34.0 a	37.5
17 (March)	38.9 a	39.1 a	40.2 a	39.6 a	39.1 a	40.3 a	40.7 a	41.1 a	36.1 a	39.5
19 (May)	40.1 a	40.3 a	41.5 a	40.0 a	39.9 a	41.4 a	41.4 a	43.2 a	37.7 a	40.6
21 (July)	43.1 a	43.5 a	44.3 a	38.8 a	41.8 a	44.5 a	44.1 a	45.5 a	40.1 a	42.9
Mean	37.4	36.8	38.3	36.6	37.0	38.7	38.5	38.6	34.8	37.4
	ST content (% FW)
5 (March)	31.3 ab	32.5 ab	32.3 ab	29.9 bc	34.9 a	32.5 ab	30.1 bc	33.0 ab	27.3 c	31.5
7 (May)	34.9 a	34.5 a	36.9 a	34.4 a	34.7 a	35.1 a	36.4 a	34.6 a	33.2 a	35.0
9 (July)	36.0 ab	33.4 b	37.3 ab	34.1 b	34.4 b	34.5 ab	37.0 a	33.7 b	32.8 b	34.8
10 (August)	38.1 a	36.6 ab	38.0 a	31.3 c	37.5 a	37.2 a	36.5 ab	39.1 a	34.1 bc	36.5
11 (September)	31.3 ab	29.4 ab	31.7 ab	29.6 ab	30.5 ab	31.6 ab	34.0 a	28.4 b	30.0 ab	30.7
12 (October)	30.8 ab	31.2 ab	31.9 ab	29.3 bc	28.0 c	32.1 ab	30.2 ab	32.5 a	29.5 abc	30.3
13 (November)	31.0 ab	33.2 a	29.9 abc	24.7 d	24.2 d	30.1 abc	28.9 bc	30.3 ab	26.4 cd	28.7
14 (December)	32.8 ab	31.1 ab	32.3 ab	30.4 ab	28.3 b	31.0 ab	33.1 a	31.9 ab	27.7 b	30.9
15 (January)	35.0 a	36.6 a	35.5 a	33.4 ab	30.5 b	34.8 a	35.4 a	35.3 a	30.8 ab	34.2
17 (March)	37.3 ab	38.2 ab	37.6 ab	36.3 abc	32.3 c	38.6 ab	39.5 a	38.5 ab	35.1 bc	36.5
19 (May)	38.1 ab	38.8 ab	37.9 ab	36.7 bc	33.3 c	38.1 ab	40.7 a	37.2 abc	32.7 c	36.9
21 (July)	33.4 c	38.7 ab	34.8 abc	34.8 abc	34.1 bc	36.1 abc	36.8 abc	39.9 a	32.6 c	35.7
Mean	34.5	34.8	35.1	32.4	31.7	34.2	35.1	34.8	30.8

Values within rows followed by the same lowercase letters are not significantly different by the LSD test (p < 0.05).

). However, the correspondence between SG and DM or ST content in the studied genotypes was not consistent across harvests. For example, genotype IAC 14 and 1097/13 showed high SG at all harvest times, except at 11 MAP (Table 2). However, DM content in both genotypes was high at 7–10 MAP and 12–21 MAP, whereas ST content in IAC 14 was high at 5–19 MAP, whereas in the 1097/13 genotype, ST content was high at 5–7, 10, and 12–21 MAP (Table 3). In turn, genotypes BRS CS01 and BRS 420 showed a high value for SG at 5–10 MAP; BRS Ocauçú showed a high value for SG at 7–11 MAP, and BRS Boitatá showed a high value for SG at 11–14 MAP (Table 2). Furthermore, between 15–21 MAP, DM did not differ among genotypes, although root DM content was high in genotypes BRS CS 01, BRS Ocauçú, and BRS Boitatá at 7–9 and 11–14 MAP, whereas in IAC 90, DM content was high, particularly at 12–14 MAP (Table 3). Regarding ST content, genotypes BRS CS01 and BRS Ocauçú showed high ST content across harvests, except for IAC 14 and IAC 90 which showed low ST content at 21 and 9 MAP, respectively. Genotype 1097/13 showed low ST content at 9 and 11 MAP, whereas BRS Boitatá showed low root ST content at 5 and 13 MAP.

3.2. Estimation of DM and ST Content of Cassava Roots Based on SG

Simple linear regression equations were plotted between SG and DM content and between SG and ST content separately for each genotype or all genotypes in the dataset combined (

Table 4. Coefficients of determination and empirical equations relating SG with DM and ST contents in storage roots of nine modern cassava genotypes (data grouping 324 samples of SY1 and 360 samples of SY2).

Genotype	SG × DM Content		SG × ST Content
	Regression Equation	R2	Regression Equation	R2
IAC 14	y = −192.0112 + 202.7012 *** x	0.69	y = −109.5272 + 127.0557 ** x	0.37
IAC 90	y = −153.1717 + 168.5676 *** x	0.72	y = −117.4801 + 134.6916 *** x	0.48
BRS CS01	y = −176.9792 + 190.0264 ** x	0.75	y = −97.3650 + 117.0525 * x	0.42
BRS 419	y = −194.5820 + 205.7919 *** x	0.62	y = −151.2487 + 163.2803 *** x	0.64
BRS 420	y = −228.2490 + 235.0584 *** x	0.71	y = −216.7411 + 220.5523 *** x	0.56
BRS Ocauçú	y = −179.0768 + 192.3948 *** x	0.82	y = −102.9561 + 121.5908 *** x	0.35
BRS Boitatá	y = −237.0694 + 243.2902 *** x	0.70	y = −152.0970 + 166.1325 *** x	0.36
1097/13	y = −208.8248 + 217.7174 *** x	0.82	y = −146.2954 + 159.4273 ** x	0.58
2011 02-43	y = −238.1149 + 244.0085 *** x	0.77	y = −177.9014 + 186.8521 *** x	0.60
All genotypes	y = −191.2806 + 202.4617 *** x	0.92	y = −140.9525 + 154.4288 *** x	0.77

*, **, and *** are significant at 5%, 1%, and 0.1%, respectively.

). The coefficient of determination (R²) for the equations generated to estimate root DM content ranged from 0.62 to 0.92, and all values were significant (p < 0.050). Moreover, the equations adjusted to estimate root ST contents showed lower R² values (0.35 to 0.77) than those obtained for estimating DM contents, despite which, all equations showed a significant (p < 0.050) adjustment (Table 4). Finally, the equations obtained when all genotypes were pooled to estimate root DM and ST content resulted in higher R² values than any genotype-specific equation.

Genotype-specific equations and a general equation based on the combined data from all pooled genotypes were used to estimate root DM and ST content, and the results were compared with those obtained using the laboratory method (

Table 5. Measured vs. predicted values for DM content in storage root samples of nine modern cassava genotypes at different harvesting times (data from SY1 and SY2 were pooled together; n = 8).

Genotype	(DM %FW) a		MAP
			5	7	9	10	11	12	13	14	15	17	19	21
IAC 14	Measured		30.9 b	38.2 a	39.2 a	40.6 a	34.0 a	35.9 a	33.8 a	33.2 b	37.4 a	39.0 b	40.3 a	43.1 a
	Predicted	IN	34.1 a	38.8 a	38.5 a	39.6 a	34.0 a	34.8 a	31.9 b	34.7 a	37.2 a	39.8 ab	40.1 a	42.3 c
		CM	34.7 a	39.2 a	38.9 a	39.8 a	36.3 a	35.3 a	32.6 b	35.1 a	37.7 a	40.2 a	40.6 a	42.7 b
IAC 90	Measured		32.8 b	39.1 a	39.1 b	38.0 a	34.6 a	36.4 a	32.4 b	34.0 a	37.8 a	39.3 a	40.8 a	43.5 a
	Predicted	IN	35.6 a	39.4 a	39.2 b	37.6 a	34.9 a	34.0 b	33.2 a	34.7 a	37.5 a	39.1 a	39.4 a	43.0 a
		CM	35.6 a	40.0 a	39.8 a	38.6 a	35.7 a	33.6 b	32.0 b	34.4 a	37.8 a	39.6 a	40.0 a	43.8 a
BRS CS01	Measured		32.1 b	39.4 a	40.1 a	38.4 a	39.8 a	35.7 a	30.7 a	32.9 a	36.5 a	39.2 a	39.8 a	44.3 a
	Predicted	IN	35.1 a	40.0 a	40.1 a	39.7 a	40.3 a	33.3 b	30.8 a	33.6 a	35.7 a	37.5 a	39.6 a	43.4 b
		CM	35.4 a	39.9 a	40.0 a	40.0 a	39.0 a	32.8 b	31.1 a	33.0 a	35.3 a	37.2 a	39.5 a	42.2 c
BRS 419	Measured		30.3 c	37.9 a	38.5 a	37.5 ab	37.1 a	34.6 a	29.0 a	34.2 a	38.2 a	40.4 a	41.0 a	38.8 a
	Predicted	IN	34.9 a	38.3 a	38.4 a	38.2 a	37.9 a	33.1 b	30.2 a	34.1 a	36.7 a	39.3 a	40.7 a	39.8 a
		CM	33.4 b	37.8 a	37.9 a	36.8 b	38.1 a	32.7 b	28.7 a	33.7 a	36.3 a	38.8 a	40.2 a	39.9 a
BRS 420	Measured		36.8 a	38.2 a	38.8 a	38.7 a	36.0 b	34.4 a	28.4 b	31.9 b	37.6 a	38.3 a	40.1 a	41.8 a
	Predicted	IN	36.5 a	39.3 a	38.9 a	39.4 a	36.8 ab	32.9 c	30.5 ab	33.5 a	36.3 a	37.7 a	39.2 a	40.4 a
		CM	36.4 a	39.2 a	38.9 a	38.7 a	37.5 a	33.7 b	31.7 a	34.2 a	36.6 a	37.8 a	39.1 a	40.9 a
BRS Ocauçú	Measured		32.5 a	40.6 a	41.5 a	39.5 a	40.4 a	35.9 a	33.2 a	32.8 a	35.5 b	39.3 b	39.6 a	44.5 a
	Predicted	IN	33.3 a	40.3 a	40.6 ab	39.8 a	40.7 a	34.4 b	32.6 a	33.3 a	38.0 a	40.4 a	39.0 a	42.9 a
		CM	32.6 a	39.6 a	39.9 b	39.1 a	40.6 a	33.4 c	31.8 a	32.2 a	37.1 a	39.7 ab	38.2 a	42.8 a
BRS Boitatá	Measured		30.7 b	39.3 a	39.3 a	37.6 a	40.2 a	34.7 a	31.7 a	30.3 b	37.9 a	39.4 a	39.9 a	44.1 a
	Predicted	IN	33.3 a	38.4 b	40.0 a	37.6 a	40.4 a	34.3 a	30.2 a	34.2 a	36.7 a	38.3 a	38.9 a	43.8 a
		CM	34.1 a	38.0 b	39.3 a	37.0 a	39.4 a	34.5 a	32.0 a	34.4 a	36.6 a	37.9 a	38.4 a	43.4 a
1097/13	Measured		33.3 b	38.8 a	40.5 a	42.6 a	35.1 a	37.3 a	33.9 a	34.0 b	37.7 b	40.9 a	42.7 a	45.5 a
	Predicted	IN	33.6 ab	40.6 a	40.7 a	40.4 b	37.9 a	35.3 b	32.8 a	35.6 a	39.1 a	40.7 a	42.6 a	44.1 a
		CM	34.7 a	40.6 a	40.8 a	40.3 b	38.5 a	35.7 b	34.1 a	36.0 a	39.3 a	40.7 a	42.5 a	43.6 a
2011 02-43	Measured		28.3 c	39.2 a	39.8 a	36.2 a	36.3 a	34.1 a	29.1 b	33.4 a	37.0 a	38.4 a	38.8 a	40.1 a
	Predicted	IN	32.0 b	39.3 a	39.1 a	34.7 a	34.9 a	31.7 c	29.9 ab	34.2 a	36.9 a	39.1 a	39.2 a	37.9 a
		CM	33.0 a	38.9 a	38.7 a	35.4 a	36.3 a	32.6 b	31.1 a	34.7 a	36.9 a	38.7 a	38.8 a	39.4 a

Values within columns followed by the same lowercase letters in the columns, for the same genotype, are not significantly different by the LSD test (p < 0.05). ^a IN: predicted values using the genotype-specific equation; CM: predicted values using the equation derived after combining the data obtained for all the genotypes.

and

Table 6. Measured vs. predicted values for ST content in storage root samples of nine modern cassava genotypes at different harvesting times (data from SY1 and SY2 were pooled together; n = 8).

Genotype	(ST %FW) a		MAP
			5	7	9	10	11	12	13	14	15	17	19	21
IAC 14	Measured		31.3 a	34.8 a	35.6 a	38.1 a	31.3 a	30.8 a	31.0 b	32.8 a	35.2 a	37.3 a	38.0 a	33.4 b
	Predicted	IN	34.1 a	35.1 a	34.9 a	38.7 a	31.3 a	32.6 a	32.3 a	32.6 a	34.2 a	35.8 b	36.0 b	33.3 b
		CM	33.0 a	34.9 a	34.6 a	37.5 a	29.5 a	31.8 a	31.1 b	31.8 a	33.7 a	35.6 b	35.9 b	36.4 a
IAC 90	Measured		32.5 b	35.2 b	34.5 a	36.6 a	29.4 a	31.2 a	33.2 b	31.5 a	36.5 a	38.2 a	38.7 a	38.7 a
	Predicted	IN	36.3 a	36.4 a	36.2 a	37.0 a	28.9 a	32.1 a	35.3 a	32.6 a	34.9 ab	36.1 ab	36.4 b	39.0 a
		CM	33.8 b	35.5 ab	35.3 a	36.5 a	28.8 a	30.6 a	30.6 c	31.2 a	33.8 b	35.1 b	35.4 c	37.6 a
BRS CS01	Measured		32.3 c	36.5 a	36.8 a	38.0 ab	31.7 a	31.9 ab	29.9 b	31.6 ab	35.4 a	37.6 a	37.9 a	34.8
	Predicted	IN	36.1 a	36.3 a	36.4 a	39.8 a	32.2 a	32.2 a	32.6 a	32.3 a	33.6 b	34.8 a	36.1 a	34.1
		CM	33.6 b	35.4 a	35.5 a	37.7 b	32.4 a	29.9 b	29.8 b	30.1 b	31.9 c	33.4 a	35.1 a	35.9
BRS 419	Measured		29.9 b	34.0 a	33.6 a	31.3 b	29.6 a	29.1 a	24.7 a	30.6 a	32.8 a	36.2 a	36.7 a	34.8 a
	Predicted	IN	32.0 a	33.5 a	33.6 a	34.9 a	31.9 a	29.4 a	27.9 a	30.2 a	32.3 a	34.3 a	35.4 a	33.5 a
		CM	31.9 a	33.8 a	33.9 a	34.9 a	31.4 a	29.9 a	27.7 a	30.6 a	32.6 a	34.5 a	35.6 a	33.4 a
BRS 420	Measured		34.9 a	34.2 a	34.0 a	37.5 a	30.5 a	27.4 b	24.2 b	28.4 a	30.1 b	32.3 a	33.8 a	34.1 a
	Predicted	IN	32.3 a	34.3 a	34.0 a	35.3 a	29.7 a	28.3 b	26.2 b	28.9 a	31.5 ab	32.8 a	34.2 a	34.1 a
		CM	34.6 a	34.8 a	34.6 a	36.5 a	30.7 a	30.6 a	30.3 a	31.0 a	32.9 a	33.8 a	34.7 a	34.5 a
BRS Ocauçú	Measured		32.5 a	35.0 a	36.5 a	37.2 a	31.6 a	32.1 a	30.1 a	30.1 a	36.2 a	38.6 a	38.1 a	36.1 a
	Predicted	IN	30.7 a	35.7 a	35.9 a	37.1 a	33.7 a	32.0 a	29.9 a	31.2 a	34.2 a	35.8 a	34.9 a	35.3 a
		CM	31.1 a	35.1 a	35.3 a	36.9 a	34.2 a	30.4 b	30.4 a	29.5 a	33.3 a	35.2 a	34.1 a	36.5 a
BRS Boitatá	Measured		30.1 b	36.4 a	38.6 a	36.5 a	34.0 a	31.0 a	28.9 b	34.0 a	36.8 a	39.5 a	40.7 a	36.8 a
	Predicted	IN	32.1 a	36.0 a	37.1 a	36.6 a	33.6 a	33.2 a	29.0 ab	33.1 a	34.9 a	35.9 ab	36.4 b	37.7 a
		CM	32.4 a	33.9 b	34.9 a	35.1 b	32.9 a	31.3 a	30.6 a	31.2 a	32.8 a	33.8 b	34.2 b	37.2 a
1097/13	Measured		33.0 a	35.3 a	34.2 b	39.1 a	28.4 a	32.5 a	30.3 b	32.0 a	35.9 a	38.4 a	37.2 a	39.9 a
	Predicted	IN	33.0 a	36.3 a	36.5 a	38.4 a	32.0 a	32.4 a	32.4 a	32.7 a	35.3 a	36.4 b	37.8 a	38.8 b
		CM	33.0 a	35.9 a	36.1 a	38.0 a	31.8 a	32.2 a	32.5 a	32.4 a	34.9 a	36.0 b	37.4 a	37.5 c
2011 02-43	Measured		27.3 c	36.6 a	34.3 a	34.1 a	30.0 a	29.5 a	26.4 c	29.1 b	33.1 a	35.0 a	33.9 a	32.6 a
	Predicted	IN	29.7 b	34.6 b	34.3 a	32.2 a	28.8 a	28.7 a	27.7 b	30.6 ab	32.7 a	34.4 a	34.4 a	32.6 a
		CM	31.5 a	34.6 b	34.5 a	33.6 a	29.4 a	29.8 a	29.8 a	31.4 a	33.1 a	34.5 a	34.5 a	32.8 a

Values within columns followed by the same lowercase letters in the columns, for the same genotype, are not significantly different by the LSD test (p < 0.05). ^a IN: predicted values using the genotype-specific equation; CM: predicted values using the equation derived after combining the data obtained for all the genotypes.

). The results showed that the predicted root DM content in genotypes IAC 14, IAC 90, BRS CS01, and BRS Boitatá at 5 MAP using either the genotype-specific or the general equation was higher than the measured DM content (Table 5). Additionally, the DM content predicted by both equations in the BRS Boitatá genotype at 7 MAP and in the 1097/13 genotype at 10 MAP was lower than the corresponding measured DM content. Furthermore, using only the equation for all combined genotypes, the predicted DM content in the BRS Ocauçú genotype at 9 MAP and the BRS 419 genotype at 10 MAP was lower than the measured DM content. The DM content predicted by the same equation at 9 MAP for the IAC 90 genotype and 11 MAP for the BRS 420 genotype was higher than the measured DM content. In general, at 12 MAP, the DM content predicted by both equations was lower than the measured DM content for all genotypes, except for IAC 14 and BRS Boitatá. At 13 MAP, the predicted DM content was higher than the measured DM content for genotypes 2011 02-43 and BRS 420 only when the equation for all combined genotypes was used. In contrast, for IAC 90, the predicted DM was higher than the measured DM only when a genotype-specific equation was used. Between 14–21 MAP, the variation between predicted and measured DM content was smaller, with a significant difference only for genotypes IAC 14 at 14, 17, and 21 MAP; BRS CS01 at 21 MAP; BRS 420 at 14 MAP; 1097/13 at 14 and 15 MAP; BRS Ocauçú at 15 and 17 MAP; and BRS Boitatá at 14 MAP.

The ST content predicted at 5 MAP by both equations for genotypes BRS CS01, BRS 419, 2011 02-43, and BRS Boitatá, and by the genotype-specific equation obtained for IAC 90 was higher than the measured ST content (Table 6). Smaller differences were recorded between the predicted and measured ST content at 7 MAP, except for genotypes 2011 02-43 and BRS Boitatá, where the predicted values were lower than the measured values. At 9 MAP, only in genotype 1097/13 was the ST content predicted by both equations greater than the measured ST content, which was also observed in genotype BRS 419 at 10 MAP. Conversely, at 10 MAP, the ST content predicted for BRS Boitatá, using an equation based on the combined data of all genotypes, was lower than the measured ST content. In turn, at 12 MAP, the predicted ST content using the same equation for BRS 420 was higher than the measured ST content, whereas the opposite was recorded for BRS Ocauçú. Furthermore, at 13 MAP, the ST content predicted by the genotype-specific equations was higher than the measured ST content for IAC 14, IAC 90, and BRS CS01, whereas the ST content predicted by both the general and specific equations for 1097/13 and 2011 02-43 was higher than the measured ST content. Finally, in the 15–21 MAP period, there was greater variation between the predicted and measured ST content in genotypes IAC 14 and IAC 90, whereas they were similar for genotypes BRS 419, BRS 420, 2011 02-43, and BRS Ocauçú over the same period.

Measured DM and ST content was correlated with predicted DM and ST content using equations from this study and previous studies [8,10,11,12]. The correlations between the predicted and measured DM and ST content using genotype-specific equations obtained in this study were stronger (r = 0.85 and 0.71, respectively) than those observed when the equations from previous studies were used (Figure 2 and Figure 3).

However, the use of equations based on the combined data from all the genotypes evaluated in this study provided significant correlations between the predicted and measured DM and ST content, with a coefficient of determination equal to that of the equations generated in the previous studies. Despite the positive and significant correlation between the predicted and measured DM content rendered by the previous equations, the predicted DM was lower than the measured DM when the values were lower than 30.5, 31.9, 33.9, and 38.2, for equations of Grossman and Freitas [10], Wholey and Booth [8], Kawano et al. [11], and Maraphum et al. [12], respectively (Figure 2). However, predicted DM was greater than measured DM when the values were above 31.2, 32.8, 35.7, and 45.4, using the equations proposed by Grossman and Freitas [10], Wholey and Booth [8], Kawano et al. [11], and Maraphum et al. [12], respectively. The ST content predicted by the equations in these studies also correlated positively and significantly with the measured ST content (Figure 3). However, ST content predicted by the equations of Grossman and Freitas [10], Wholey and Booth [8], Kawano et al. [11], and Maraphum et al. [12] showed values lower than the ST measured ranging from 2.1–4.5, 1.1–1.6, 1.9–2.0, and 0.8–3.2 units, respectively. However, both the DM and ST content predicted by the equations obtained in this study were closer to the measured values. Only the ST content predicted by the general equation showed values lower than the measured ST content ranging from 0.3–0.5 units (Figure 3).

4. Discussion

In situations where it is not possible to dry the roots or perform laboratory analysis to determine ST content, SG can help to estimate the DM and ST content of cassava storage roots [3,8,11]. In this study, a dataset of 684 samples collected from 2 study sites, 9 genotypes, and 9 to 10 harvest time points (9 from SY1 and 10 from SY2) was compared. The values for SG, DM, and ST content showed wide variations among genotypes and harvest times, which is useful for creating conversion tables for SG values into the corresponding ST content [8]. Consistent with this, Rangel et al. [7] observed strong variations in ST content in the roots of six cassava genotypes as a function of plant age in a long-cycle crop (23.5 MAP) in the south-central region of Brazil.

In general, analysis of this dataset revealed an increase in SG, DM, and ST during the first vegetative cycle between 5–9 MAP (March–July), followed by a reduction after the physiological rest phase from 10–13 MAP (August–November), and a subsequent increase from 13–21 MAP (November–July of the following year). This pattern of variation showed that storage root maturity peaked at approximately 9 MAP (July), and from this period until 13 MAP, stored energy as carbohydrates in the roots were used for the regrowth of plant foliage, similar to the results obtained by Rangel et al. [7]. Although some studies indicate that SG values peak between 8–14 MAP [11,25], this study showed that the increase in SG, DM, and ST content of storage roots during the first vegetative cycle of cassava occurred during autumn and the beginning of winter, that is, during the transition from the rainy and hot seasons to the dry and cold seasons. If cassava is not harvested at maturity, during the transition from winter to spring (beginning of rain and increasing ambient temperatures), SG values and DM and ST content of storage roots decrease because the stored ST in the roots is used as plants resume growth by producing new branches and leaves after a long dry season [7,11,12].

The results of our study revealed that during the second vegetative cycle, root SG values and DM and ST content began to increase only at the end of spring (late November), when plants had already grown new foliage. Therefore, the ST content of the cassava roots depends on the harvest period [7,11,26,27]. Furthermore, the duration of high ST content in cassava roots may vary among genotypes, as reported by Rangel et al. [7]. For example, for early harvests, such as 9 MAP, the genotype BRS Boitatá accumulated more ST in its roots than genotypes IAC 90, BRS 419, BRS 420, 1097/13, or 2011 02-43. In contrast, for late harvests (i.e., 21 MAP), the ST content of genotype 1097/13 was higher than that of genotypes IAC 14, BRS 420, or 2011 02-43. However, when the average of all nine to ten harvests was considered, genotypes BRS CS01, 1097/13, BRS Ocauçú, and BRS Boitatá showed DM and ST content greater 38% and 34.8%, respectively (Table 3). Despite the variations between harvest times, the genotypes of the IAC group showed similar DM and ST content in their storage roots to that in most of the genotypes of the BRS group, mainly in the later harvests of the second vegetative cycle.

The advantage of the SG method for estimating ST content in cassava storage roots lies in its ease, simplicity, and speed, whereas the main disadvantage is its low level of precision that characterizes the method [3,8,9,10,11,12]. From an industrial perspective, laboratory analysis of root ST content, despite being more accurate, delays processing flow because it is tedious and time-consuming for large volumes of samples [3,6,9].

For genotypes IAC 14 and 1097/13, the number of harvests for which the predicted DM did not coincide with the measured DM which was greater than that for the other genotypes. Regarding the ST content, which is the most important component of cassava roots, genotypes IAC 90 and BRS Boitatá showed a greater number of harvests, which as predicted did not coincide with the measured ST content. Generally, there were two periods during which the predicted DM values differed more frequently than the measured values. Thus, one of the limitations of this study is that the SG method showed a tendency toward overestimation of DM content at 5 MAP for seven of the nine genotypes studied using both equations. In contrast, at 12 MAP, the SG method tended to underestimate the DM content in the six genotypes. Considering field performance, it is unusual for harvesting to be conducted very early (5 MAP) or during the plant regrowth period, when the DM content decreases as reserves are used for regrowth of the plant canopy. There was less disagreement between predicted and measured DM and ST content between 7–11 and 14–21 MAP. Considering the ST content (Table 6), the highest frequencies of differences between the measured and predicted values using the two equations occurred at 5 and 13 MAP. However, for both periods, the limitation of the SG method was that the ST content tended to be overestimated. Therefore, despite the equations showing a good estimation of ST content of the storage roots throughout the harvest, the equations showed limitations in estimating ST content in cassava roots harvested very early (5 MAP) or during the plant shoot regrowth phase (13 MAP). Very young roots generally have smaller starch granules, and roots that remain for more than one vegetative cycle in the field have a mix of fully formed granules (formed during the first cycle) and small granules formed due to the resumption of ST accumulation in the roots during the second vegetative cycle [17]. Under these conditions, SG values were not closely related to ST content of the storage roots, indicating that the empirical equations from our study could not be used to predict ST content in all situations.

In addition, obtaining either an overestimation or underestimation of DM and ST content by the SG method using the equation based on pooled data from all genotypes was more common across harvests than when using genotype-specific equations. Thus, the use of the latter to estimate the DM and ST content of cassava storage roots appears to be a more accurate approach than using an equation based on a dataset comprising several genotypes.

When comparing the equations used for estimating DM and ST content in this study with those used in other studies [8,10,11,12], the genotype-specific equations in this study showed stronger correlations between SG and DM content and between SG and ST content (Figure 2 and Figure 3). In general, the predicted DM values obtained using the equations derived in this study were very close to the measured values. The equations by Grossman and Freitas [10], Wholey and Booth [8], and Kawano et al. [11] provided close estimates of the DM content of cassava roots within a concentration range of 30.5–31.2%, 31.9–32.8%, and 33.9–35.7%, respectively. For roots with a DM content below these ranges, the equations overestimated the DM content, and for values above these ranges, an underestimation of the measured DM content was commonly observed. The equation proposed by Maraphum et al. [12] was the closest to the measured DM content values.

The same applies to the ST content of cassava roots, and the equations of previous studies did not provide the same accuracy as those used in this study. In general, among the equations from previous studies, the equation by Grossman and Freitas [10] was the least suitable for accurately predicting the ST content of cassava roots, whereas the equation proposed by Wholey and Booth [8] was the best for predicting the measured root ST content, despite a variation of 1.1–1.6% between the predicted and measured values. In this study, when using the equation based on the dataset combining all genotypes, the variation between the predicted and measured ST content ranged between 0.3–0.5%, whereas the variation was only 0.1–0.2% when the genotype-specific equations were used.

Considering the estimation of ST content from predicted DM content, the results obtained in this study allow us to propose using constant values of 3.43 and 3.84 with predicted DM content obtained by the genotype-specific or the general equation, respectively, instead of the constant of 4.65 suggested by Conceição [14].

Finally, updating the empirical equations proposed herein for the estimation of the DM and ST content of cassava roots may contribute to greater transparency in the relationship between cassava farmers and cassava processing industries. Therefore, it is desirable for the cassava processing industry to adopt the equations proposed in this study for validation and process improvement.

5. Conclusions

The correlation between SG and DM content in the cassava roots was stronger than that between SG and ST content. Harvest seasonality had a significant influence on the existing correlation between SG and DM or ST content of the roots, and there was a difference among genotypes in the earliness of ST accumulation in the roots.

Genotype-specific equations were better for estimating the DM and ST content of cassava roots than the equations based on pooled data from all genotypes tested. The SG method tended to overestimate the DM content for harvests at 5 MAP and underestimate it at 12 MAP, whereas for ST, there was a tendency for ST content to be overestimated for harvests at 5 and 13 MAP.

Our study showed that the genotype-specific equations generated in this study can be used with greater precision by processing industries to estimate the DM and ST content of cassava roots. The predicted DM content of the roots can also be used as an indicator of the ST content; however, for conversion, constants 3.43 and 3.84 were more suitable than the conventionally used constant value of 4.65. Finally, the empirical equations proposed in this study allow for a better prediction of the root DM and ST content of modern cassava genotypes grown during long harvest cycles and may contribute to greater transparency in the relationship between cassava farmers and cassava processing industries. However, future studies should include the percentage of ST that industries cannot remove from fresh cassava roots during processing in the generated equations.