-
In this study, a total of 50 genotypes of hybrid bermudagrass (C. dactylon × C. transvaalensis) were used in the greenhouse, out of which 48 were planted in the field experiment. The hybrids were developed at Oklahoma State University by mating various genotypes of common (tetraploid) bermudagrass with African (diploid) bermudagrass.
Greenhouse study
-
The plants were grown in a greenhouse at the US Arid-Land Agricultural Research Center in Maricopa, Arizona (USA). Stolon cuttings from the genotypes were taken and planted in 27.95 cm × 55.88 cm watertight plastic trays with 36 square cell pack insert filled with optimal porosity growth mix (coarse sphagnum peat moss (80%−90%), perlite (Berger, Saint-Modeste, Canada). The planting of the cuttings was conducted mid April 2021. The greenhouse's growing temperature was set to 32/27 °C (day/night) for the experiment with the natural summer season day light duration. The plants were kept in a well-watered state by watering every other day with one drip emitter per flat. After establishment, the plants were equally divided into three, 12 plugs each and arranged in replications. Then the experiment was laid out in a randomized complete block design with three replications. To simulate the same age and canopy height, all the plants were mowed to approximately 7.5 cm height using Makita cordless grass shears a week before data collection.
Crop Circle RapidScan CS-45 (Holland Scientific, Lincoln, NE, USA), a height-independent active crop canopy sensor, was used to collect the canopy reflectance data at bands of 670, 730, and 780 nm. Each flat was scanned for about two seconds while the device was held steady at a height of 0.7 m above the canopy. The data were collected twice from all the three replicates the same day and the average was considered for analysis. From the reflectance data recorded at 670, 730, and 780 nm bands, six SVI known to estimate photosynthetic area and chlorophyll content were calculated (Table 1).
Table 1. Published formulae for different spectral vegetation indices used in the study.
Spectral reflectance indices Formula References Normalized difference vegetation index (NDVI) (R780 − R670)/
(R780 + R670)[46] Normalized difference red edge index (NDRE) (R780 − R730)/
(R780 + R730)[47] The chlorophyll index using red edge (CIRE) (R780/R730) −1 [48] Normalized difference vegetation index-Red-Red edge (NDRRE) (R730 − R670)/
(R730 + R670)[49] The MERIS terrestrial chlorophyll index (MTCI) (R780 − R730)/
(R730 − R670)[50] Canopy chlorophyll content index (CCCI) NDRE/NDVIR [51] Field study
-
The field experiment was established with 48 genotypes in a randomized complete block design with two replications. Field plots were planted late-August 2021 at Maricopa Agricultural Center near Maricopa, Arizona, USA (33.079 °N, 111.977 °W). Four plugs were planted at a square of 0.5 m in the middle of 1.5 m × 1.5 m plot. The canopy reflectance data were collected following the method described above from the center of each plot twice in mid-July 2022. The establishment rate was visually estimated every month and recorded in percentages until at least 90% of the 1.5 m × 1.5 m square plot area is covered. Winter color data were recorded mid December 2021 and mid January 2022 using 1−9 scale (1 = brown, 9 = fully green).
Statistical data analysis
-
The data were tested for normal distribution and the absolute values of skewness and kurtosis was less than 1.0, suggesting underlying assumption in parametric testing is fulfilled. Analysis of variance was conducted using the aov() function in R [52]. It was thought that the genotype had a random effect. For mean separation, the least significant difference (LSD) at p ≤ 0.05 probability was utilized.
Scatter plot, frequency distribution, and correlation analysis among the indices was conducted using the PerformanceAnalytics package in R[52]. The range of indices among the accessions was depicted using box plots.
To identify superior genotypes among the 50 bermudagrass genotypes evaluated based on their performance for all the indices, a multi-trait selection index analysis was performed using the multi-trait genotype-ideotype distance index (MGIDI)[53]. To illustrate the correlations between the six indices, principal component analysis was computed using genotype means.
To select superior genotypes based on multiple spectral indices, each index was analyzed following the mixed-effect model:
, where y is an n[ = ∑rj = 1(gr)] × 1 vector of response variable, i.e. the response of the ith genotype in the jth block; b is an 1 × r vector of unknown and unobservable fixed effects of block b; u is an m[ = 1 × g] vector of unknown and unobservable random effects of genotype u; X is an n × r design matrix of 0 s and 1 s relating y to b; Z is an n × m design matrix of 0 s and 1 s relating y to u; and e is an n × 1 vector of random errors[53].$ y=Xb+Zu+e $ Estimation of the multi-trait genotype-ideotype distance index (MGIDI) was calculated using
2]0.5, where MGIDIi is the multi-trait genotype–ideotype distance index for the ith genotype; γij is the score of the ith genotype in the jth factor; and γj is the jth score of the ideotype.$MGIDIi=[{\sum }_{j\; =\; 1}^{f}(\gamma ij-\gamma j)$ -
Each of the indices' data distribution and density heatmap are shown using a modified box and whisker plot (Fig. 1). As a result, the MTCI has a wide range, ranging from 0.7 to 1.2 with a mean value of 0.9. The average chlorophyll index (CIRE) was 0.5, with a range of 0.29 to 0.75. NDVI and NDRRE had intermediate ranges, while NDRE and CCCI exhibited narrower ranges, distributed quite near to the mean than the others.
Figure 1.
Density heatmap of the range of indices used to evaluate 50 turf type bermudagrass genotypes.
The analysis of variance revealed significant differences among the genotypes for all the estimated indices from the reflectance data taken at 670, 730, and 780 nm, despite variances in the ranges of indices (Table 2). These variations among genotypes call for additional investigation to determine the role of various SVI in identifying superior genotypes among turf bermudagrass experimental hybrids.
Table 2. Analysis of variance for hybrid bermudagrass spectral reflectance indices calculated using published formulae.
Source DF Mean squares NDVI NDRE CIRE NDRRE MTCI CCCI Replication 2 0.0023 0.0003 0.0024 0.0035* 0.0341* 0.0025* Genotype 49 0.0107*** 0.0011*** 0.0111*** 0.0105*** 0.0208*** 0.0018*** Error 98 0.0009 0.0002 0.0017 0.0010 0.0091 0.0006 DF = Degree of freedom; NDVI = Normalized difference vegetation index; NDRE = Normalized difference red edge index; CIRE = Chlorophyll index using red edge; NDRRE = Normalized difference vegetation index red-red edge; MTCI = MERIS terrestrial chlorophyll index; and CCCI = Canopy chlorophyll content index; *, **, *** = significant at p = 0.05, 0.01, or 0.001, respectively. Data from the performance analysis of 50 genotypes of hybrid bermudagrass were displayed as scatter plots, frequency distributions, and relationships between the six indices (Fig. 2). A histogram with a data distribution curve traced diagonally showed that the data were distributed normally or nearly so. The scatter plots and regression lines used to trace them highlight the correlation between the indices. A linear regression model can account for most of the relationship. MTCI and CCCI, however, displayed a non-linear relationship with the other indices.
Figure 2.
Scatter plots, frequency distributions, and correlations among the six spectral reflectance indices used to evaluate 50 hybrid bermudagrass genotypes for genetic variation based on canopy reflectance at different wavelengths.
The correlation coefficients among the indices indicated that NDVI was highly correlated with NDRRE (r = 0.98), CIRE (r = 0.76), and NDRE (r = 0.76). NDRE was perfectly correlated with CIRE (r = 1.00). It was also highly positively correlated with NDRRE (r = 0.62) and MTCI (r = 0.47). CIRE was highly correlated with NDRRE (r = 0.62). Similarly, MTCI and CCCI were highly correlated (r = 0.95) but were weakly or negatively correlated with the rest of the indices used in this study. As such CCCI was negatively correlated with NDRRE (r = −0.64) and NDVI (r = −048).
Multi-indices selection
-
Principal component analysis using MGIDI revealed two principal components accounting for a total of 99.8% (PC1 = 67.3% and PC2 = 32.5%) of the variations in spectral reflectance among the genotypes.
Factor analysis using indices as selection differentials selected OSU2102, OSU2120, OSU2053, OSU2108, OSU2039, OSU2123, OSU2075, OSU2015, OSU2118, and OSU2124 as superior over the others (Fig. 3). On the other hand, OSU2106, OSU2109, and OSU2037 were very close to the threshold line. Tifway, which was included as a check was very close to these three. The other cultivar used as a check, TifTuf was not close to the selected lines and had average MGIDI. According to the selection index analysis, OSU2115 and OSU2114 were the least in MGID index and forms the base of the index.
Figure 3.
Genotype rankings for multi-trait genotype-ideotype distance index (MGIDI). The selected genotypes based on MGIDI index are shown in red.
Factors analysis to the MGIDI indicated that genotypes OSU2102 and OSU2039 have strong contribution in factor 1 (FA1) (Fig. 4). OSU2053, OSU2075, OSU2123, OSU2015, OSU2118, and OSU2124 have strong contribution to factor 2 (FA2) (MTCI and CCCI). The smallest contribution of FA1 on OSU2053 and OSU2075 suggests that these genotypes have high measure in SVI related to chlorophyll content but weak in SVI related to photosynthetic area. On the other hand, OSU2102 has the smallest contribution of FA2 indicating its good photosynthetic area but weak in chlorophyll content. Genotypes OSU2039 and OSU2015 were selected based on the modest contribution of both factors (FA1 and FA2) implying their good photosynthetic area and chlorophyll content.
Figure 4.
The strength and weakness view of the selected genotypes with the proportion of the contributing factors to the computed multi-trait genotype-ideotype distance index (MGIDI). The dashed line represents the average of the two factors (FA1 and FA2).
Regression analysis of indices for establishment and winter color
-
Simple linear regression was conducted to test the predictive ability of two vegetation indices (NDVI and CIRE) that are related to photosynthetic area and chlorophyll content for spreading (establishment rate) and two others (MTCI and CCCI) that are related to chlorophyll content for winter color retention (Fig. 5). There was significant (p < 0.10) positive correlation of NDVI and CIRE with establishment rate. Similarly, MTCI and CCCI showed significant (p <0.05) correlation with winter color. The extent of spreading regressed with two spectral indices revealed the important relationship of photosynthetic area with shoot growth rate. The likely importance of chlorophyll content for fall and winter color retention as measured by CCCI, MTCI, and CIRE was also high.
-
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.
-
About this article
Cite this article
Serba DD, Wu Y, Hejl RW, Williams CF, Bronson KF. 2023. Spectral reflectance estimated genetic variation in hybrid turf bermudagrass. Grass Research 3:22 doi: 10.48130/GR-2023-0022
Spectral reflectance estimated genetic variation in hybrid turf bermudagrass
- Received: 09 May 2023
- Accepted: 02 November 2023
- Published online: 20 November 2023
Abstract: High throughput phenotyping (HTP) utilizing both remote and proximal sensing technologies has emerged as a vital tool for evaluating the biophysical characteristics of turfgrass. This study was conducted to assess the genetic diversity of hybrid turf bermudagrass using spectral reflectance indices and use of HTP for germplasm enhancement. A total of 50 accessions of the hybrid bermudagrass (Cynodon dactylon × C. transvaalensis) were grown in the greenhouse in three replications. The spectral data were gathered using a height independent active crop canopy sensor, 'RapidScan CS-45', which measures canopy reflectance at the wavelengths of 670 nm, 730 nm, and 780 nm. The reflectance data were used to derive three indices related to canopy photosynthetic area and other three related to chlorophyll content. All vegetation indices showed significant genotype-to-genotype variation. Ten superior genotypes were identified using the multi-trait genotype-ideotype distance index (MGIDI) as a selection differential. On 48 of the genotypes that were established in the field in two replications, establishment rate and winter color data were also gathered. The results of a linear regression analysis demonstrated the importance of spectral vegetation indices (SVI) for the turfgrass quick establishment (percentage area coverage) and winter color retention. This study brings attention to the potential use of the proximal sensing in turfgrass germplasm enhancement for establishment speed, aesthetic value, and mild-winter color retention.