Effect of Harvesting Corn after Frost in Alberta (Canada) on Whole-Plant Yield, Nutritive Value, and Kernel Properties

: This study compares yield, nutritive value, and kernel properties of whole plant corn (WPC) harvested before and after a light frost in short growing season areas. Six corn hybrids grown in two years at three locations within Alberta (Canada) were harvested before or after the ﬁrst frost. Samples of WPC were analyzed for dry matter (DM) content, neutral detergent ﬁber (NDF) concentration, starch concentration, and 48-h in vitro DM and NDF digestibility (DMD and NDFD, respectively). Cob samples were analyzed for DM, and kernels were analyzed for DM, hardness, particle size distribution, density, and stage of maturity. Delaying harvest to after frost increased DM content of WPC at all locations but exceeded the recommended range (32–38%) in the two warmest locations. Whatever the year and hybrid, DM yield was either not affected or decreased after frost. Postfrost harvest increased starch concentration and modiﬁed kernel characteristics only if these were less than expected before frost. Fiber concentration was not affected by harvesting time. Frost had either no impact or increased DMD or NDFD of WPC. We conclude that delaying harvest until after frost in short growing season areas can be beneﬁcial when whole-plant DM content is low before frost.


Introduction
Corn silage (CS) is increasingly used in ruminant diets as a source of digestible starch and fiber [1]. Given that its average inclusion rate in dairy rations can be as much as 40% of dry matter (DM) in North America and Europe [2], maximizing CS quality by optimal cropping management is essential to promote animal performance. In locations with cooler climates and short growing seasons, harvesting time needs to be carefully determined to maximize both the quantity and quality of biomass. Harvesting immature corn results in low DM yield and starch concentration, although the digestibility of fiber is enhanced. Delaying harvest extends the grain filling period and increases DM yield, but fiber digestibility and the potential for long-term preservation [3] may be negatively affected.
In areas with a cool and short growing season, small grains such as barley, wheat, and oats are usually favored for silage production as their development is less heat-dependent than corn [4]. However, small grains usually have lower yields and digestibility compared with corn. The recent development of short-season corn hybrids makes it feasible to grow corn hybrids for silage in northern locations characterized by a short growing season. Moreover, the annual growing season for corn in the Canadian prairie provinces (Alberta, Saskatchewan, and Manitoba) has increased by 3 to 12 days over the past century as a consequence of climate change [5]. Consequently, corn production has continually increased in the Prairies even though this area is still challenged by low temperatures in spring that delay planting and the occurrence of frost early in the autumn. About 50 days are necessary from silking to harvest to reach corn maturity [6]; however, this period can be limited to 30-40 days [7] in cool environments. Early frost damages the plants through tissue breakdown, which increases plant DM content, causes leaching of nutrients (nitrogen and minerals; [8]), and negatively affects fiber digestibility [9]. Severe frost may also induce loss of plant fractions such as leaves and inhibit translocation of sugars to ears and the conversion of sugars to starch in the kernels, which are active processes, thereby preventing kernel filling [10].
In a previous study, we demonstrated that when using new hybrids, CS production is possible in Northern areas even though attainment of whole-plant maturity before frost may be difficult to achieve in some years and locations due to annual variability of corn heat unit (CHU; a temperature-based index used to measure the cumulative heat over the growing season) accumulation [7]. The impact of harvesting the corn crop for silage after the occurrence of frost in areas with a short growing season is unclear. We hypothesize that delaying harvest until after frost will not be necessary with new early-maturing hybrids as sufficient maturity and nutrient concentrations may be attained within the limited growing season. Therefore, the aim of the present work is to compare corn whole-plant characteristics, including yield, nutritive value, and kernel properties of short-season hybrids harvested before and after frost.

Experimental Design
Corn was grown in two years (2013 and 2015) in three different locations within Alberta: Lacombe (probability of receiving 2000 CHU between 50% and 60%), Lethbridge (probability of receiving 2000 CHU between 70% and 80%), and Vauxhall (probability of receiving 2000 CHU between 80% and 90%) [5]. Soil description, long-term climatic information, planting practices, water supply, and fertilization strategies for these sites have already been described [7]. Briefly, the texture of the soil is characterized as fine sandy loam in Lacombe and silt loam in Lethbridge and Vauxhall.
At each location, six short/medium season CS hybrids were planted in four replicates in a randomized complete block design each year (Table 1). Because of lack of seed availability, one hybrid in Lethbridge (P8622AM) could not be used in both years, and, therefore, a replacement hybrid (P8210HR) with a similar CHU rating was used in 2015. This modification was considered to have a low impact on the experimental design, and no distinction was made in further analytical treatments. Corn was seeded between 6 May and 3 June and harvested before (between 18 and 24 September) or after (between 29 September and 7 October) light frost ( Table 2). The CHU accumulated between seeding and harvesting was calculated for each year and location as the sum of daily CHU, as provided by the Government of Canada [11]. Initiation for summation of CHU at each location and year used the method outlined by Huggins-Rawlins [12]. The final day of CHU accumulation at each location and year corresponded to the harvesting date.

Sampling and Chemical Analyses
Upon harvesting, the kernel reproductive stage was assessed according to standard scaling [13], and sampling was carried out as described earlier [7]. Briefly, five plants per replicate were separated into ear and stover and weighed. The remaining plants were harvested, weighed to determine total wet weight biomass (including the weight of the five plants), and a representative sample was retained. Frozen samples (whole plants and cobs) were shipped to the Lethbridge Research and Development Centre (LeRDC, Agriculture and Agri-Food, Alberta, Canada) after determining kernel stage maturity [14].
Whole plant samples were dried to a constant weight at 55 • C to determine DM content before being ground in a Wiley mill (A.H. Thomas, Philadelphia, PA, USA) through a 4-mm screen. One subsample was further ground through a 1-mm screen, and another subsample was ball-ground (mixer mill MM 400; Retsch Inc., Newtown, PA, USA). All samples from Year 1 were analyzed in duplicate for DM content and concentrations of organic matter (OM), neutral detergent fiber (NDF), acid detergent fiber (ADF), acid detergent lignin (ADL), crude protein (CP), and starch, as well as 48-h in vitro DM digestibility (DMD) and NDF digestibility (NDFD) [7]. All samples that had been chemically analyzed were scanned using near-infrared reflectance spectroscopy (NIRSystems 6500 Monochromator, Foss NIRSystems Inc., Silver Spring, MD, USA). Prediction equations were developed, and the R 2 and R 2 of cross-validation (R 2 CV) were calculated. These were greater than 0.83 for DM, OM, ADF, ADL, starch, and CP. Therefore, for samples from Year 2, these nutrients were predicted using the equations generated from Year 1 samples. Because of lower values for NDF (R 2 CV = 0.75, n = 232), NDFD (R 2 CV = 0.42, n = 235), and DMD (R 2 CV = 0.46, n = 230), the samples from Year 2 for these components were analyzed as described for Year 1.
Silks and husks were removed from frozen cobs before they were broken in half. Ten kernels were removed from the edge of each half cob and dried at 55 • C for 3 days to determine kernel DM content. Cobs were then stored for 7 days in a ventilated room at 26 • C before removing the grains from the rachis with a cob sheller. Grain density was measured based on the test weight method [15]. Kernel hardness was determined using the Stenvert test [16] and according to the assays of Blandino et al. [17] and Ma and Dwyer [18]. Resulting ground subsamples from Year 2 were further used to assess particle size distribution by sieving on a Ro Tap particle separator (model RX-29; W.S. Tyler, Mentor, OH, USA) equipped with two screens (600 and 420 µm) and a bottom pan [17].

Statistical Analyses
Statistical analyses were carried out with R software (Version 3.4.0; [19]). For each location, the fixed effects of year (2013 versus 2015), hybrid, harvest period (before versus after frost), and their interactions (harvest period × year, harvest period × hybrid, and year × hybrid) on specific variables (nutrients, DM yield, cob proportion, and kernel properties) were tested with a linear model (function "lme" of the package nlme) using block (4 blocks of 6 hybrids each) as the random effect. The triple interaction (harvest period × year × hybrid) was not significant in most cases and was, for that reason, not reported. The "emmeans" function (package emmeans) was further applied to each model to calculate LSmeans and to conduct a pairwise comparison with a Tukey adjustment when single effects were significant. Data were considered significant at p < 0.05, and tendencies were discussed at 0.05 ≤ p ≤ 0.10.
To illustrate the effect of the three main tested effects (year, hybrid, and harvest period) on nutritive value (whole plant DM, NDF, ADF, ADL, starch, CP, DMD, and NDFD), wholeplant DM yield, cob proportion (except for Lethbridge and Vauxhall, where cob proportion was measured in 2013 only), and kernel properties (DM content, hardness, density, and stage), principal component analyses were conducted on the data of each location (function "dudi.pca" of the package ade4), and graphical representations of individuals were obtained with the "fviz_pca_ind" function (package factoextra).

Nutritive Value and DM Yield by Location
The effects of year, harvest period, hybrid, and their interactions on nutritive value and DM yield of whole-plant corn grown in Lacombe, Lethbridge, and Vauxhall are presented in Tables 3-5, respectively. For whole-plant DM content, NDF and starch concentrations, DMD, NDFD, or DM yield, boxplots are presented when a significant effect was observed for the interaction of year and harvest period. In Lacombe (Table 3), the effect of year was significant (p < 0.021) for all variables except ADL which averaged 22.1 g/kg (p = 0.40). During the first year (2013), the hybrids were characterized by substantially greater DM yield (+9.2 T/ha), DM content (+7.8 percentage points), fiber concentration (+27 g/kg for NDF), and DMD (+19 g/kg) and lower CP concentration (−3.5 g/kg), starch concentration (−22 g/kg), and NDFD (−15 g/kg). Delaying harvest until after frost increased (p < 0.002) whole-plant DM content (+8.1 percentage points) and starch concentration (+42 g/kg) while decreasing ADF concentration (−9 g/kg) and tending (p = 0.087) to decrease CP concentration (−1.5 g/kg). Other variables were not affected. The effect of frost on nutritive value and yield was similar in both years (no harvest × year interactions) for most variables, except for DM, CP, and starch, for which effects were not consistent between years ( Figure 1). Differences (p < 0.005) were observed among hybrids for all variables except digestibility values (a tendency for NDFD). The effect of hybrid was consistent over years (except for DM content) and harvesting time (except for NDF concentration).  In Lethbridge (Table 4), CP and starch concentrations were similar between years (71.6 and 318 g/kg on average, respectively), whereas all other variables differed (p < 0.001): in 2013, NDFD and DM yield were lower than in 2015 (−39 g/kg and −2.7 T/ha, respectively), whereas DM content and concentrations of NDF, ADF, and ADL, as well as DMD, were greater in 2013 (+11.2 percentage points and +42, +23, +4.1 and +19 g/kg, respectively). Harvesting time did not affect NDF, ADL, and starch concentrations or DM and NDF digestibility of whole-plant corn. However, harvesting after frost increased (p < 0.001) DM content (+13.8 percentage points on average, with a stronger effect in 2013; Figure 2) and CP concentration (+4.8 g/kg), and decreased (p < 0.026) ADF concentration (−8 g/kg) and DM yield (−2.2 T/ha on average, with a stronger effect in 2013; Figure 2). The effect of harvest was consistent across years for NDF, ADF, and starch concentrations as well as for DMD (tendency) and NDFD. The opposite effect of frost was observed between years for CP and ADL concentrations. The effect of hybrid was significant (p < 0.001) for all variables except NDFD, but the effect of hybrid was inconsistent between years (p < 0.047) and harvesting period (p < 0.035) for many of the variables.  In Vauxhall (Table 5), all variables were affected by year (p < 0.015). In 2013, wholeplant corn was characterized by greater DM content (+11.8 percentage points), starch concentration (+82 g/kg), DMD (+74 g/kg), and NDFD (+33 g/kg) and lower CP (−4.9 g/kg), NDF (−24 g/kg), ADF (−27 g/kg), and ADL (−3 g/kg) concentrations and DM yield (-1.1 T/ha). Delaying harvest until after frost increased (p < 0.001) DM content (+6.2 percentage points), DMD (+24 g/kg), and NDFD (+25 g/kg). By contrast, late harvest decreased (p < 0.001) CP (−2.8 g/kg), NDF (−39 g/kg), ADF (−19 g/kg), and ADL (−2.6 g/kg) concentrations, as well as DM yield (−2.5 T/ha). No impact of frost was observed on starch concentration (268 g/kg on average). However, the effect of harvesting time on nutrient composition was highly dependent on the year (Figure 3). Whole-plant corn DM content and CP, starch, and NDFD concentrations were different (p < 0.047) among hybrids, but other variables were not affected (except a tendency for ADL concentration). The effect of hybrid was In Vauxhall (Table 5), all variables were affected by year (p < 0.015). In 2013, wholeplant corn was characterized by greater DM content (+11.8 percentage points), starch concentration (+82 g/kg), DMD (+74 g/kg), and NDFD (+33 g/kg) and lower CP (−4.9 g/kg), NDF (−24 g/kg), ADF (−27 g/kg), and ADL (−3 g/kg) concentrations and DM yield (-1.1 T/ha). Delaying harvest until after frost increased (p < 0.001) DM content (+6.2 percentage points), DMD (+24 g/kg), and NDFD (+25 g/kg). By contrast, late harvest decreased (p < 0.001) CP (−2.8 g/kg), NDF (−39 g/kg), ADF (−19 g/kg), and ADL (−2.6 g/kg) concentrations, as well as DM yield (−2.5 T/ha). No impact of frost was observed on starch concentration (268 g/kg on average). However, the effect of harvesting time on nutrient composition was highly dependent on the year (Figure 3). Whole-plant corn DM content and CP, starch, and NDFD concentrations were different (p < 0.047) among hybrids, but other variables were not affected (except a tendency for ADL concentration). The effect of hybrid was repeatable for most variables over years (except DM content, starch concentration, and DM yield) and harvesting time (except a tendency for starch).

Cob Proportion and Kernel Properties by Location
The effects of year, harvest period, hybrid, and their interactions on cob proportion and kernel properties of whole-plant corn grown in Lacombe, Lethbridge, and Vauxhall are presented in Tables 6-8, respectively. The interactions between year and harvest period for kernel DM content at each location are shown in boxplots (Figure 4).
In Lacombe (Table 6), the year did not affect the cob proportion or the kernel hardness. Kernel DM content and stage were lower (p < 0.001) in 2015 (−1.8 percentage points and −0.43 points), whereas kernel density was greater (p < 0.001) that year (+7.7 points). Harvesting before frost increased (p < 0.001) cob proportion (+4.0 percentage points in late harvest) and altered kernel properties except for the proportion of large particles. Late harvest increased kernel DM content (+5.2 percentage points), hardness (+3.4 points), small particle proportion (+1.5 percentage points), density (+1.7 points), and stage (+0.41 points). Only the proportion of medium particles decreased with late harvest (−1.1 percentage points). The effect of harvest was similar between years for cob proportion, kernel density, and stage (tendency) but different (p < 0.007) for kernel DM content and hardness. Hybrid affected (p < 0.001) all cob-related variables, and this effect was repeatable over years for most variables except for kernel DM content, density, and stage (p < 0.038) and over harvesting periods except for kernel hardness (tendency), stage, and proportions of large and small particles (p < 0.030).

Cob Proportion and Kernel Properties by Location
The effects of year, harvest period, hybrid, and their interactions on cob proportion and kernel properties of whole-plant corn grown in Lacombe, Lethbridge, and Vauxhall are presented in Tables 6-8, respectively. The interactions between year and harvest period for kernel DM content at each location are shown in boxplots (Figure 4). In Lacombe (Table 6), the year did not affect the cob proportion or the kernel hardness. Kernel DM content and stage were lower (p < 0.001) in 2015 (−1.8 percentage points and −0.43 points), whereas kernel density was greater (p < 0.001) that year (+7.7 points). Harvesting before frost increased (p < 0.001) cob proportion (+4.0 percentage points in late harvest) and altered kernel properties except for the proportion of large particles. Late harvest increased kernel DM content (+5.2 percentage points), hardness (+3.4 points), small particle proportion (+1.5 percentage points), density (+1.7 points), and stage (+0.41 points). Only the proportion of medium particles decreased with late harvest (−1.1 percentage points). The effect of harvest was similar between years for cob proportion, kernel density, and stage (tendency) but different (p < 0.007) for kernel DM content and hardness. Hybrid affected (p < 0.001) all cob-related variables, and this effect was repeatable over years for most variables except for kernel DM content, density, and stage (p < 0.038) and over harvesting periods except for kernel hardness (tendency), stage, and proportions of large and small particles (p < 0.030).  In Lethbridge (Table 7), the year affected (p < 0.004) all variables: in 2013, cob DM content was greater than in 2015 (+4.1 percentage points), whereas kernel hardness, density, and stage were less in 2013 (−3.5, −2.0, and −1.02 points, respectively). Late harvest increased (p < 0.029) cob and kernel DM content (+5.7 and +9.7 percentage points, respectively), proportion of small particles (+1.4 percentage points), kernel density (+1.8 points), and stage (+0.46 points). Large particle proportion decreased (p = 0.006) when harvested after frost (−1.6 percentage points), whereas harvest did not affect kernel hardness or proportion of medium particles. The effect of frost was different between years for kernel DM content, hardness (tendency), and density. All variables were affected by hybrid (p < 0.029). This effect was repeatable over harvesting periods for most variables, except for kernel DM content and density, and over years, except for kernel DM content and hardness. In Vauxhall (Table 8), kernel characteristics were different between years (p < 0.002), with lower DM content (−5.0 percentage points) and greater hardness (+9.7 points), density (+1.2 points), and stage (+0.64 points) in 2015. Delaying harvest to after frost increased (p < 0.029) kernel DM content (+4.0 percentage points), proportion of large particles (+1.2 percentage points), and stage (+0.21 points) and decreased proportion of small particles (−1.1 percentage points). Harvesting time did not affect cob DM content, kernel hardness, proportion of medium particle size, or density. The effect of harvesting time was similar between years only for kernel hardness and density. Hybrid affected (p < 0.011) all variables except kernel density (tendency). The observed effects were consistent over years, except for kernel hardness, and over harvest period, except for proportion of medium particles and kernel stage.  At all locations, strong positive correlations were observed among fiber fractions (NDF, ADF, and ADL), and these fiber fractions were negatively correlated with starch concentration. In Lacombe and Lethbridge, DMD and NDFD had a lower contribution to the overall variability compared to Vauxhall, but parameters were positively correlated at all locations. Digestibility of DM and NDF were negatively correlated with NDF at all locations except in Lethbridge, where no significant correlation was observed between NDFD and NDF. In Lacombe, DMD was also positively correlated with whole-plant DM content and yield, whereas NDFD was negatively correlated with DM yield. In Lethbridge, DMD was positively correlated with starch concentration and DM yield and negatively correlated with kernel density; NDFD was positively correlated with kernel hardness and stage of maturity. In Vauxhall, both DMD and NDFD were positively correlated with starch concentration and negatively correlated with kernel hardness, density, and stage; DMD was also positively correlated with whole-plant DM. Figures 6-8 illustrate the effect of harvesting time, year, and CHU rating of the hybrid, respectively, on whole-plant corn nutritive value and kernel properties at each location. Samples did not clearly cluster according to harvesting period, though some groupings with starch concentration and negatively correlated with kernel hardness, density, and stage; DMD was also positively correlated with whole-plant DM.  year, and CHU rating of the hybrid, respectively, on whole-plant corn nutritive value and kernel properties at each location. Samples did not clearly cluster according to harvesting period, though some groupings can be distinguished in the Lacombe and Vauxhall samples. Clustering was very consistent for year and hybrid (CHU rating of the hybrid less or greater than 2400) effects.   year, and CHU rating of the hybrid, respectively, on whole-plant corn nutritive value and kernel properties at each location. Samples did not clearly cluster according to harvesting period, though some groupings can be distinguished in the Lacombe and Vauxhall samples. Clustering was very consistent for year and hybrid (CHU rating of the hybrid less or greater than 2400) effects.

Discussion
Despite the development of new corn hybrids for silage production in short growing season areas such as the Canadian Prairie Provinces, the question of optimum harvesting time relative to frost has not yet been resolved. In this work, we investigated the effect of a light frost on yield, nutritive value, and kernel properties of whole-plant corn grown for silage.

Frost Increased DM Content and Risk for Lower Yield
Before frost, whole-plant DM content in Lacombe (26.5%) was less than the typical range (32-38% DM) of corn grown in adapted areas, but this target was achieved in Vauxhall (36.7%) and slightly exceeded in Lethbridge (41.2%). The lower DM content observed and expected in Lacombe was attributed to delayed maturity caused by the lower mean

Discussion
Despite the development of new corn hybrids for silage production in short growing season areas such as the Canadian Prairie Provinces, the question of optimum harvesting time relative to frost has not yet been resolved. In this work, we investigated the effect of a light frost on yield, nutritive value, and kernel properties of whole-plant corn grown for silage.

Frost Increased DM Content and Risk for Lower Yield
Before frost, whole-plant DM content in Lacombe (26.5%) was less than the typical range (32-38% DM) of corn grown in adapted areas, but this target was achieved in Vauxhall (36.7%) and slightly exceeded in Lethbridge (41.2%). The lower DM content observed and expected in Lacombe was attributed to delayed maturity caused by the lower mean ambient temperature, higher accumulated precipitation, and lower accumulated CHU between seeding and harvest compared with the two other locations [7]. Delaying harvest until after frost greatly increased DM content of whole-plant corn at the three locations (+6.1, +13.8, and +6.2 percentage points on average in Lacombe, Lethbridge, and Vauxhall, respectively) due to plant tissue breakdown followed by desiccation. However, only at Lacombe was frost advantageous in bringing whole-plant DM content into the recommended range for ensilage (32 to 38% DM). Frost caused whole-plant DM content in the two other locations, especially Lethbridge, to dry excessively beyond the ideal range. Material that is too dry at harvest is difficult to pack densely and leads to weak fermentation during the ensiling process. The resulting silage has high pH and is at risk of microbial or fungal development [20,21]. Therefore, delaying harvest until after frost to increase whole-plant corn DM content in poorly adapted locations such as Lacombe (probability of receiving 2000 CHU <60%; [5]) may be successful, but this practice is not advantageous in warmer areas like Lethbridge in most years. However, it must be noted that the magnitude of the frost effect on increasing DM content depends on the year and the maturity already reached prior to frost. Indeed, the impact of harvesting time is stronger if whole-plant DM content is already high before frost, with this threshold DM value being dependent upon location.
In 2013 and 2015, DM yield of whole-plant corn for silage averaged 14.4 T/ha in Canada (13.6 T/ha in the Prairie provinces), assuming 35% DM content at harvest [22]. The overall DM yield observed in the present study was close to the national average, although slightly lower in Lacombe (10.1 T/ha). Whatever the year and hybrid, wholeplant DM yield was not affected by frost in Lacombe. In contrast, DM yield decreased with late harvest in Lethbridge and Vauxhall, confirming previous work conducted in Québec (Canada), where delaying harvest after frost decreased yield by 38 kg of DM/ha per day after the first frost [23]. In the present study, the extent of yield loss depended on the year in both locations but was not affected by the DM yield before frost. The weak interaction between hybrid and harvest period indicates fairly similar effects of harvesting after the first frost among hybrids, confirming previous work [23].

Frost had only Minimal Effects on Dry Matter Digestibility
In vitro DMD values achieved in this study (627 to 701 g/kg DM) were in the range reported in the literature (570 to 700 g/kg; [24]). Frost did not affect DMD of whole-plant corn in Lacombe and Lethbridge, whatever the year or the hybrid, supporting results obtained in 1976 in Eastern Canada [8]. However, higher DMD occurred in Vauxhall (+2.4 percentage points) when harvesting after frost, this effect being independent of hybrid but dependent on the year. These results contradict previous findings that DMD declined with progressive maturity [24]. For instance, DMD of whole-plant corn harvested after the first frost decreased by 8% in another Canadian study conducted in 1977 [25]. As the drydown of whole-plant DMD is partially driven by corn stover digestibility, the development of new corn hybrids that maintain stover quality at advanced stages of maturity may concomitantly help to preserve whole-plant DMD after frost [24]. Interestingly, previous work showed that in vitro DMD [23] or apparent DMD measured using lactating dairy cows [26] of whole-plant corn did not change during the first month after the first frost despite multiple frost occurrences.

Starch Concentration Increased in Response to Frost in the Coolest Areas of Alberta's Prairies
Whole-plant starch concentrations observed in Lacombe (between 194 and 236 g/kg DM) were below the range usually observed in areas well suited for corn production (between 270 and 370 g/kg DM). In contrast, the higher starch concentrations (up to 316 g/kg DM) observed in the two other locations were more closely in line with this range. The lower starch concentration, combined with lower DM content, of corn grown in Lacombe indicates these plants were more immature at the time of first frost. This difference in maturity accounts for the different influence of frost on starch concentration for the three locations, as postfrost harvesting only positively affected starch concentration in Lacombe. Therefore, delaying harvest until after frost as a means of increasing starch concentration of whole-plant corn may be a useful strategy in the coldest areas of the prairie Provinces [5], where starch concentration and DM content are in the lowest ranges. Additionally, late harvesting can also sometimes be advantageous in warmer areas, as observed in 2015 for Vauxhall, for which starch concentrations were low before frost. However, given the variability of frost in general versus maturity at harvest, as indicated by this study, harvesting after frost to enhance kernel development and starch concentration may be a risky management strategy that is hard to preplan, given the downside of over-dry silage.
Kernel characteristics were also modified with harvesting time. Being positively correlated with whole-plant DM content, kernel DM content increased after frost at all locations. One may assume that drier kernels would lead to increased kernel hardness, density, and development stage, as well as a higher proportion of small particles. However, with the exception of Lacombe, these correlations were not very prominent and possibly explained by the large differences in kernel DM content across locations, with Lethbridge and Vauxhall presenting contents up to 30% greater than in Lacombe. In other words, kernel characteristics may depend on kernel DM content until a threshold DM level is achieved, above which hardness and particle size become less predictable.
Previous research showed that increasing DM percentage of corn kernels linearly decreased starch digestibility, with the degree of vitreousness explaining 86% of the variation in ruminal starch digestibility [24]. Numerically, starch digestibility can reach 80% at 40% kernel DM, but digestibility drops to 60% when kernel DM increases to 65%. In the present study, using a kernel processor at harvest would be strongly recommended for whole-plant corn harvested after frost to ensure silage quality in terms of starch accessibility. Hybrid selection should also be considered carefully as kernel DM contents of the various hybrids differed by as much as 15%.

Fiber Concentration and Digestibility Were not Negatively Affected by Frost
Over the three locations, NDF concentration of whole-plant corn ranged between 485 and 572 g/kg, which is close to the usual range (between 405 and 635 g/kg DM; [24]). Except in Vauxhall, where frost reduced fiber concentration, delayed harvest did not affect NDF or ADL concentrations of corn harvested for silage, whatever the year. This finding supports a previous study in which delaying harvest of whole-plant corn from 34% to 41% DM did not influence fiber (NDF, ADF, and ADL) concentrations [20]. Similarly, earlier studies showed no impact of frost on cellulose [8] or fiber [26] concentrations in silage corn. However, other studies reported that delaying harvest to after frost increased fiber concentration of whole-plant corn [9] or that fiber concentration continued increasing after frost (+0.68 g/kg DM per day after the first frost, [23]; +65 g/kg DM after the fifth frost, [26]). Interestingly, ADF concentration of whole-plant corn was affected by frost in our study, unexpectedly with lower instead of higher concentrations at late harvest. Confirming a previous study in Québec [23], hybrid affected fiber concentration of whole-plant corn in Lacombe and Lethbridge, but without a clear relationship with the CHU rating of the hybrid. The impact of harvesting time was consistent among hybrids at all locations.
Fiber digestibility is an important indicator of forage quality, affecting animal performance. In a meta-analysis including different forages harvested for silage (including whole-plant corn), a one-unit increase in NDF digestibility measured in vitro or in situ was associated with a 0.17 kg increase in DM intake and a 0.25 kg increase in 4% fat-corrected milk [27]. Similarly, decreasing NDF digestibility of corn silage resulted in lower average daily gain and gain-to-feed ratios in growing cattle [28].
Frost did not affect the NDFD concentration of whole-plant corn in Lacombe and Lethbridge, whatever the year or hybrid, whereas greater NDFD was observed in Vauxhall after frost, with a stronger impact in 2015. Mertens [29] indicated that NDF concentration and its digestibility are the primary factors contributing to variation in DMD. This relationship is corroborated by our results, especially at Vauxhall, where DMD was increased by frost and NDFD was strongly correlated with DMD.
Previously published studies generally found no effect of corn maturity on NDFD. In vitro, NDFD was similar between whole-plant corn harvested for silage at 34% or 41% DM [20]. Additionally, apparent total tract NDF digestibility measured on nonlactating Holstein cows was not different between whole-plant corn harvested at an early (<300 g/kg of DM) or late (>400 g/kg of DM) stage of maturity [30]. No specific effect of frost was reported on apparent NDF digestibility of corn silage fed to sheep [9]. In vitro fiber digestibility was also not affected by multiple frost occurrences [23,26]. However, increasing the maturity of corn silage fed to lactating dairy cows reduced the apparent ADF digestibility of the diet [31].
Fiber digestibility is mostly related to the lignification process [29,32]. In accordance with expectations, the component pattern plots clearly demonstrate the negative correlation between ADL concentration and NDFD of whole-plant corn at all locations. ADL concentrations were not affected by frost in Lacombe and Lethbridge, where frost did not affect NDFD either. By contrast, late harvest resulted in lower ADL concentrations in Vauxhall, which may partly explain the higher NDFD after frost at this location.

Conclusions
This study confirmed our hypothesis that delaying the harvest of whole-plant corn for silage until after frost should not be systematically practiced in short-season growing locations. New adapted hybrids can attain acceptable maturity for ensilaging prior to first frost, with high yield and starch concentration. However, in very short-season regions known for less-than-optimal heat unit accumulation, harvesting corn after frost may allow it to reach the targeted DM and nutrient concentrations without large detrimental impacts on yield and digestibility. Overall, the optimal harvesting time will vary yearly and by location and is a function of plant maturity (DM content) prior to the occurrence of frost. Reproducing this study at different locations may give further insights into possible interactions between the effect of harvesting time and environmental conditions, as the latter might affect the resistance of the plants to frost.
Author Contributions: V.S.B. and K.A.B. conceived, designed, conducted the experiment, provided guidance in the laboratory analysis of samples, and helped write the final paper; J.G. analyzed the data, prepared the tables and figures, and wrote the first draft of the manuscript. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.