Yield and Survival of Shrub Willow Clones over Two Rotations Reveal Important Patterns About Selection for an Evapotranspiration Cover on a Former Industrial Site

Abstract

Shrub willow (Salix spp.) is a promising candidate for evapotranspiration (ET) covers due to its rapid growth and high water use. This study assessed 30 willow clones over two three-year rotations on a former industrial waste site in Solvay, NY, with alkaline, low-organic substrates and intermittent hardpan. Survival was high after the first rotation (87.9% ± 1.7 SE), but yield was lower and more variable (6.55 Mg ha⁻¹ y⁻¹ ± 0.25 SE) than on mineral soils. In the second rotation, both survival (42.6% ± 3.0 SE) and yield (5.08 Mg ha⁻¹ y⁻¹ ± 0.38 SE) declined. Clone rankings shifted between rotations (Spearman ρ = 0.13, p = 0.48), suggesting that short-term trials poorly predict long-term performance on degraded sites. Survival emerged as the primary driver of yield, with a smaller interaction from hardpan. Clone 05X-295-014 showed notable resilience, maintaining strong performance despite widespread hardpan. Five clones from S. miyabeana and S. purpurea x S. miyabeana groups demonstrated sustained or increasing yield and survival above 60%. These findings emphasize the importance of selecting for survival alongside yield in multi-rotation trials to ensure effective long-term deployment for biomass and phytoremediation in challenging sites.

1. Introduction

Industrial facilities on the west side of Onondaga Lake in Solvay, NY, were a dominant producer of soda ash (Na₂CO₃) from 1881 to 1986 using its namesake, the Solvay method [1]. This process utilized regionally abundant limestone and local salt springs to source brine. Every megagram of soda ash produced generated approximately 10 m³ of liquid byproduct, which contained 1 Mg of calcium chloride and 0.5 Mg of sodium chloride [2]. This waste material was deposited in a series of settling basins in the southwest corner of Onondaga Lake. Ultimately, almost 270 ha of settling basins with depths ranging from 3 to 21 m were created, and due to the characteristics of this material, they supported limited amounts of vegetation and invasive plants [2,3,4]. The high concentration of chlorides in this waste resulted in high concentrations in the surrounding surface and groundwater.

A typical approach to managing the salts leaching from these settling basins would be to construct an engineered cap to block infiltration through this material, thus protecting groundwater and aquatic systems [5]. An alternative approach is to utilize an evapotranspiration (ET) cover to manage the water budget by using plants to increase evapotranspiration, improve water storage at the site and as a result minimize percolation [6,7]. Willow is a productive commercial woody biomass crop utilized worldwide and in the Northeastern US with mean productions from 8.5 to over 14 Mg ha⁻¹ y⁻¹ [8,9]. The plants for a system on this site need to be tolerant of the conditions at the site and have a high rate of transpiration.

Evapotranspiration rates of shrub willow are at the high end of plants in the northern temperate region, with a mean ET rate between 4 and 5 mm d⁻¹ reported across 57 different studies [10]. A previous study of transpiration rates of willow growing at the Solvay site found that transpiration was highest (7.0 mm d⁻¹) in June and dropped over the rest of the growing season to less than 1 mm d⁻¹ in November [11]. Additionally, the ET rate of individual willow clones has been closely correlated to their biomass yield, although the water use efficiency of willow clones can vary [12,13,14]. Therefore, the high biomass production characteristic of shrub willow, along with its long growing season, in which the plant is leafing and engaged in higher evapotranspiration, makes it particularly useful as an ET cover [6,10].

Another advantage of ET cover systems is the lower environmental impact compared to traditional cover systems. Over the lifecycle of cover systems, greenhouse gas (GHG) emissions of traditional compacted clay liners were 2.5 times greater than those of a willow ET cover, and GHG emissions from a geosynthetic liner were 7.2 times greater [15]. Willow stores carbon in coarse roots and stools, and these are typically not removed from the soil throughout the willow’s lifetime, meaning willow crops can act as a carbon sink [16]. The global warming impact (GWI) of a willow ET cover is negative but is also sensitive to root-to-shoot ratio and willow biomass yield, which can vary across sites [15]. Lastly, bioenergy crops such as willow support considerable biodiversity in these highly engineered systems [17].

Starting in 2004, a phyto-recurrent selection approach [18] was used to identify and deploy willow clones that would likely have high survival and biomass production for an ET cover. These selections were based on trials that began as greenhouse experiments, with 38 willow clones and resulted in the selection of 20 clones that were planted in small, replicated trials on the Solvay settling basins. After these initial on-site trials, 5–7 clones were selected for the establishment of approximately 20 ha of ET cover across the site over subsequent years [19].

By 2010, a new suite of next-generation shrub willow clones had developed in breeding and selection work at SUNY ESF and Cornell University. Concurrently, due to the challenging conditions at this site, some of the initial selections had shown potential weakness in sustaining their performance. A new yield trial was designed and installed in 2013 to evaluate the performance of new clones on the Solvay site. The objectives of this research were to (1) evaluate the survival rate and biomass yield of 30 willow clones over two three-year rotations on the Solvay settling basins, and (2) assess the impact of some of the unique conditions at the Solvay site on willow variety performance.

2. Materials and Methods

The study site was located on a settling basin located southwest of Onondaga Lake near the town of Solvay, NY, USA (Latitude 43.0671991, Longitude −76.2577575). The substrate of the site, byproducts from the Solvay process (Solvay, NY, USA), is comprised of 70% silt-sized particles. The pH in the top 30 cm ranges between 8.2 and 8.5. Between 30 and 60 cm depth, the pH ranges between 8.3 and 9.9. A root-limiting hardpan is present across considerable portions of the site, but varies by location and past activity.

Prior to planting willow, approximately 1000–1200 m³ ha⁻¹ of organic matter consisting of composted horse manure and associated bedding was added. The target mixing depth of this material was 45 cm, but the actual depth was impacted by the presence and depth of the hardpan. This study included 30 willow clones, three of which were checks (SX61, SX64 and SX67) that had previous success at the site (

Table 1. Clone identifier, cultivar name, Salix pedigree, and diversity group for the 30 willow clones planted in a yield trial at the Solvay settling basins.

Clone	Cultivar Name	Salix Pedigree	Group
9970-014		S. miyabeana	MIYA
9870-23	Marcy
9970-036	Canastota
9970-053
9871-31	Sherburne
SX61
SX64
SX67
9882-34	Fish Creek	S. purpurea	PUR
9882-41	Wolcott
99239-015	Allegany	S. koriyanagi S. purpurea	KP
99113-012	Onondaga
01X-265-020
01X-264-062		S. koriyanagi x (S. purpurea x S. miyabeana)	KPM
01X-264-071
99217-015	Millbrook	S. purpurea x S. miyabeana	PM
9879	Oneonta
9980-005	Oneida
05X-292-035
05X-295-014
99201-007	Otisco	S. viminalis x S. miyabeana	VM
99202-004	Fabius
99202-011	Tully Champion
99207-020	Truxton
01X-268-015	Preble
01X-268-016
01X-266-016		S. viminalis x (S. viminalis x S. miyabeana)
02X-326-010		S. miyabeana x (S. viminalis x S. schwerinii x S. viminalis)
02X-326-015
98101-61		S. dasyclados x S. miyabeana	VCC-HYB

). Multiple 25-cm cuttings were hand-planted in double rows in mid-May of 2013 with plants spaced 0.762 m between the double rows and 0.61 m along the double rows. Double rows were planted 1.83 m apart from each other. Plots were 7.8 m by 7.9 m, containing 3 double rows, 13 plants per row, and 78 plants per plot. The experimental design was a randomized complete block design with four replications. Preemergence herbicides were applied following establishment, and the plots were weeded mechanically with a walk-behind mower and by hand during the first growing season. Mowing occurred in the spring after each of the harvests, but no additional weed control measures were deployed following harvests.

After the first growing season, the willow was coppiced. Two three-year harvests occurred during the dormant season after the 2016 and 2019 growing seasons. Survival was determined for the 10 inner plants in the middle row of each plot prior to hand-harvesting. Harvested stems were weighed on hanging scales with an accuracy of 0.1 kg. Yield was annualized by dividing plot productivity (Mg ha⁻¹) by years of growth. Three stems representing the range of harvested diameters were collected and chipped for moisture content determination. The chipped stems were immediately weighed after chipping, returned to the lab, and dried at 60 °C to a constant weight. Moisture content was calculated on a wet weight basis. All yield data are reported on a dry weight basis in Mg ha⁻¹. Fertilizer was applied after each harvest with a mixture of 112 kg N, 45 kg P, 84 kg K, and 73 kg S per hectare. The remaining plants were mechanically harvested with a single-pass cut and chip system.

Depth to hardpan and mixing were recorded through the use of a bucket auger in the center of each plot. Depth to mixing was found visually by looking at the walls of a single hole in the middle of each plot. The contrast of the white or light gray Solvay waste against the dark organic material provided a high-contrast reference for making this determination. The depth to hardpan was determined at the center location and three points on each side of the two double rows of the excavated hole using a metal probe. If the probe hit hardpan, the distance from the surface to the hardpan was recorded. If a hardpan was not detected within 24cm, the depth was recorded as >24 cm.

Effects of weather, especially temperature and rainfall, can have significant effects on clone performance in bioenergy systems [20]. Basic weather information was summarized from online sources. Growing degree days (GDD; base temperature 10 °C) [21] and precipitation [22] were compiled and summarized for the duration of the trial for the early (May–June) and late (July–October) portions of the growing season (

Table 2. Growing degree days (base 10 °C) and precipitation for the early (May–June) and late (July–October) growing season for the duration of the willow trial at the Solvay site [21,22].

Phase		Early Growing Season (May–June)		Rest of Growing Season (July–October)
	Year	GDD (Base 10 °C)	Precipitation (mm)	GDD (Base 10 °C)	Precipitation (mm)
Establishment	2013	457	129	992	455
1st Rotation	2014	465	77	933	383
	2015	491	285	1013	358
	2016	377	66	1061	460
2nd Rotation	2017	362	134	986	401
	2018	445	116	1053	387
	2019	327	131	891	482
Mean		418	134	1006	418

).

Means and standard errors were calculated for annualized yield, survival, and the percentage of the plot where a hardpan was detected within 24 cm of the surface. These values were summarized by year and clone using the MEANS procedure in SAS 9.4 (SAS Institute Inc., Cary, NC, USA). Screening of clones in both rotations was performed using a repeated measures analysis and the LSMEANS statement in the GLIMMIX procedure (alpha ≤ 0.1, method = RSPL, type = arh(1)). To assess the stability of clone ordering across rotations, a Spearman rank correlation was calculated between 2016 and 2019 clone-level yield rankings and the ranking of clones in a mineral soil study using the CORR procedure in SAS 9.4 at the alpha ≤ 0.05 level.

Preliminary graphical analysis indicated a general site relationship among annualized yield in 2019, survival in 2019, and the percentage of the plot affected by a hardpan within 24 cm of the surface. Model selection for the full ordinary least squares (OLS) model (Equation (1)) based on the clone means (combination of 4 observations) was conducted using the REG procedure in SAS 9.4 using backwards selection and Mallow’s Cp statistic [23] at the alpha ≤ 0.05 level. The rationale for using clone means rather than all 119 available observations is to address scaling issues that might affect model inference due to local conditions present at the individual measurement plot. The full model is specified as:

$$Yield={\beta }_0+{\beta }_1·sur+{\beta }_2·{sur}^2+{\beta }_3·hp\%+{\beta }_4·sur·hp\%+{\beta }_5·{sur}^2·hp\%+\epsilon$$

(1)

where:

Yield = Annualized Plot Yield in 2019

Sur = Plot survival in 2019

hp% = Percentage of area where hardpan occurs within 24 cm

ε = Model error

Collinearity among model main components was assessed using the Variance Inflation Factor (VIF), with a target score below 5 [24]. To address the influence of outliers and leverage points observed in the OLS model, a final weighted least squares (WLS) model was produced using the ROBUSTREG procedure using the default settings. The best-performing model was selected based on the correlation coefficient (R²), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC), with further discussion provided in the results.

During the preparation of this manuscript/study, the authors used ChatGPT 4o (OpenAI, San Francisco, CA, USA) for the purposes of reducing the abstract length, assistance in locating applicable reference reading material, and proofreading the document. The authors have reviewed the output, manually edited the document based on suggested changes, and take full responsibility for the content of this publication.

3. Results and Discussion

3.1. General Yield and Survival

At the 2016 harvest, the mean annualized yield was 6.55 Mg ha⁻¹ y⁻¹ (median = 6.95; ±0.25 SE), and individual clones ranged between 2.9 and 8.6 Mg ha⁻¹ y⁻¹ (Figure 1). Six clones had yields in excess of 8 Mg ha⁻¹ y⁻¹, while three clones were below 5 Mg ha⁻¹ y⁻¹; 8 Mg ha⁻¹ y⁻¹ being a lower threshold for commercial clones, and 5 Mg ha⁻¹ y⁻¹ being in the range of stable, non-commercial annualized yields [25]. Survival is a key early indicator of the potential of plants to be effective in phytoremediation applications [26]. In this trial, survival was good across many of the clones, with the first rotation mean survival at 87.9 percent (Figure 2), with only one clone having less than 70% survival.

Mean survival by the 2019 rotation dropped significantly to 42.6 percent (±3.0 SE; p < 0.0001; Figure 2). Only four clones had survival more than 70%, and only one exceeded 80%. As a result, the mean annualized yield was only 5.08 Mg ha⁻¹ y⁻¹ (median = 5.36; ±0.38 SE), which was significantly lower than in 2016 (p = 0.0002). However, yield for individual clones ranged between 0 and 12.0 Mg ha⁻¹ y⁻¹ (Figure 1). In the 2019 rotations, the yield of seven clones exceeded 8 Mg ha⁻¹ y⁻¹, while fourteen fell below 5 Mg ha⁻¹ y⁻¹. As a group, these low performers were significantly below the top three performers in the trial (05X-295-014, 05X-292-035, and 9970-053; p < 0.0001). Three clones (98101-61, 99113-012, 99202-011) experienced near-complete stand failure by the end of the second rotation, as evidenced by yields < 0.5 Mg ha⁻¹ y⁻¹ and large inter-plot variance, suggesting mortality rather than merely growth suppression.

3.2. Individual Clone Performance

Breeding/genotype selection clearly matters; a third of the material meets the 8 Mg ha⁻¹ y⁻¹ “rule-of-thumb” for economic viability on marginal land [25,27]. The willows at this site were generally successful in the establishment phase and through the first rotation. At the end of the first rotation in 2016, all but one willow clone (01X-265-020) had a survival greater than 70% and 19 of the 30 clones had survival greater than 90% (Figure 2). Survival dropped significantly for 23 of the 30 clones, but yield only dropped significantly for nine of them. The survival rates for several clones at this site were acceptable, but not quite as high as the survival reported after the first rotation for yield trials planted on mineral soils in the region. For example, the survival of 18 clones at two sites in central NY planted on mineral soils ranged from 86.1% to 100%, with all but two clones at one site exceeding 90% [28]. Johnson et al. [29] found that survival was greater than 90% for 30 clones on mineral soil in NY and over 70% for 22 clones in southern Minnesota over three rotations. In contrast, in this study, survival decreased in the second rotation, with only four willow clones (05X-295-014, 9871-31, SX64, and 9980-005) having survival greater than 70%. Nineteen of the clones had survival less than 50% and of these, 12 had survival less than 30%.

Mossler et al. [30] reported that initial survival was high for willow species that had undergone screening before testing on a coal mine spoil, but declined over time. At age 1, all but one species have greater than 90% survival; by age five, survival has decreased considerably and ranged from 94 to 61%. The willow grown at this site and in the Mossler et al. [30] had good initial rooting and survival in screening trials on mineral soil, and this was reflected in the early success of the plants when they were grown on mine spoils. However, as the plants developed and aboveground biomass increased, the requirements for moisture and nutrients from this site may have been limited by soil conditions and variation in weather patterns, particularly dry periods, may have added stress, resulting in decreases in survival.

There were dramatic changes in the ranking of clones based on yield from 2016 to 2019 (Spearman ρ = 0.13, p = 0.48); the change in rank among the individual clones ranged from minus 18 to plus 20 positions. This result emphasizes the need to monitor trials on these sites for multiple rotations to optimize decisions for clone deployments in the phytoremediation system. This is particularly important on former industrial or highly disturbed sites where the impact of variation in weather patterns may not show up for several years and become more apparent as the plants get larger and have greater water and nutrient demands. Of the top six clones with yields greater than 8 Mg ha⁻¹ y⁻¹ in 2019, only one (05X-292-035) was also ranked in the top six in 2016 (

Table 3. Ranking of clones based on 2019 yield and ranking for 2016 yield. The 2016 and 2019 codes¹ summarize yield and survival selection thresholds for both the 2016 and 2019 harvests.

Clone	2016 Code for Yield and Survival 1	2019 Code for Yield and Survival 1	2016 Yield Rank	2019 Yield Rank	Combined Yield Rank
05X-295-014	2A	1A	21	1	1
05X-292-035	1A	1X	4	2	2
9970-053	2A	1C	18	3	3
SX61	XA	1B	28	4	13
99217-015	2A	1A	11	5	5
9871-31	2A	1A	19	6	6
SX64	2A	1A	26	7	11
9980-005	2A	2A	20	8	12
9870-23	2A	2X	15	9	10
9970-014	1A	2B	6	10	4
01X-268-015	2A	2B	10	11	9
99201-007	2A	2X	17	12	14
01X-264-071	2A	2C	16	13	15
9970-036	1A	2X	3	14	8
SX67	1A	2B	2	15	7
01X-264-062	2A	2B	12	16	17
99207-020	1A	XX	5	17	16
01X-268-016	XA	XX	29	18	24
99239-015	1A	XX	1	19	18
02X-326-010	2A	XX	7	20	19
9882-34	2A	XX	13	21	21
01X-266-016	2A	XX	9	22	20
9879	2A	XX	23	23	23
98101-61	2A	XX	25	24	25
9882-41	2A	XX	27	25	27
99202-004	2A	XX	8	26	22
99113-012	2A	XX	14	27	26
01X-265-020	XC	XX	30	28	30
02X-326-015	2A	XX	24	29	29
99202-011	2A	XX	22	30	28

¹ For the 2016 and 2019 codes, 1 indicates yield > 8 Mg ha⁻¹ y⁻¹, 2 is yield between 5 and 8 Mg ha⁻¹ y⁻¹, and X is yield less than 5 Mg ha⁻¹ y⁻¹. The letter is an indication of survival, with A being > 70%, B being 60%–70%, C being 50%–60% and X being < 50%. “X” is a disqualifying level for yield or survival.

). The other top five ranking clones in 2016 decreased in yield by 2019, with values ranging from 4.0 to 6.0 Mg ha⁻¹ y⁻¹. Some declines were acute; the top yielding clone (99239-015) in 2016, with a yield of 8.61 Mg ha⁻¹ y⁻¹, dropped to the 19th position in 2019, and the yield dropped to 3.09 Mg ha⁻¹ y⁻¹. Seven clones with yields greater than 9 Mg ha⁻¹ y⁻¹ in 2019 were ranked from 4 to 28 in terms of yield after the first rotation. If selection for future expansion of the ET willow cover system at this site had been based on first rotation data, there would likely be serious failures in the system within a few years. The pattern on these highly disturbed sites is often different than what is seen in trials on mineral sites, where once willow is established, it often does well over time, especially when the plant material being tested went through an initial screening program. This is the case for data from 360 willow test plots across five sites with mineral soil [31], where yield tended to converge over two rotations, with lower yielding clones in the first rotation generally having larger yield increases and higher yielding clones having no change or a slight decrease. The clones utilized in this study had been selected as part of a breeding and selection program, so those clones that had poor yield or survival were screened out at an earlier stage.

The patterns for untested clones on the current trial seemed more analogous to the “stable” and “decline” biomass patterns observed for Santucci et al. [25], which evaluated untested non-commercial clones planted in their first selection trial in the region in 1993, with very limited or previous screening. Yields from these previously untested clones varied quite a bit over multiple rotations, with several of them showing distinct declines in yield. This trial was a suite of largely untested willow clones, so this variation in yield patterns over time was anticipated.

These results suggest that due to the substrates on the Solvay settling basins, although capable of supporting some clones when amended with substantial amounts of organic matter, the response among clones can be quite variable. Several clones showed sharp increases in yield from the first to the second rotation, but only two of them were statistically significant. Ultimately, 05X-295-014 was ranked 21st in 2016 with a yield of 6.07 Mg ha⁻¹ y⁻¹ (±1.8 SE), but in 2019, it was the top-ranking clone with a yield of 11.97 Mg ha⁻¹ y⁻¹ (±1.69 SE). SX61 had a yield of 4.86 Mg ha⁻¹ y⁻¹ (±1.22 SE) in 2016 and was ranked 13th, but by 2019, the yield had increased significantly to 8.53 Mg ha⁻¹ y⁻¹ (±1.52 SE) and it was ranked 4th. The average yield of the top seven clones identified in 2019 was 7.34 Mg ha⁻¹ y⁻¹ in 2016, and was 10.72 Mg ha⁻¹ y⁻¹ in 2019, which was a 46% increase. Variation in performance is to be expected and is often hard to predict on these highly disturbed sites, necessitating the intentional implementation of a phyto-recurrant strategy to identify the correct plant material.

While yield is often the primary characteristic used for ranking willow or other species in these kinds of trials, it is important to screen choices with other data, such as survival and the trend in survival over time, if that data is available. In this trial, if the top five clones based on just 2019 yield were selected, there would be two with survival that declined considerably between 2016 and 2019. Clone 05X-292-035 survival dropped from 92.5% (±4.8 SE) to 47.5% (±8.5 SE) in the second rotation, and 9970-053 dropped from 78% (±11 SE) to below 60% (±21 SE). While their yields remained high in the second rotation, the decreases in survival raise questions about the potential of these clones to perform effectively long term at this site, so they should not be recommended for inclusion in future expansion efforts. At the end of the 2019 rotation, five of the top seven clones based on biomass also had acceptable survival (greater than 60%) that did not significantly decrease: 05X-295-014, SX61, 99217-015, 9871-31, and SX64 (Figure 2; Table 3). Two of the top seven clones had the only significant increases in yield observed between 2016 and 2019 (Figure 1). This suite of clones includes the two check clones (SX61 and SX64), indicating that there were a limited number of other clones that performed as well as these clones, which have proven good performance at this site over time.

3.3. Site Factors

Results show that the Solvay site can produce commercial-scale biomass once plants are established if survival remains adequately high. Across 30 clones, mean annualized yield increased from approximately 6.0 to 7.5 Mg ha⁻¹ y⁻¹ (20%–30%) between the first and second three-year rotations, but the response was highly genotype-dependent (Figure 1) and was likely also impacted by seasonal climate during the rotation.

Harayama [20] demonstrated the importance of adequate rainfall and higher temperatures during the early growing season on willow performance. The average GDD for the early part of the growing season across the seven years of the trial was 418 (Table 2). Both GDD and precipitation were close to averages over the length of the trial, and the consistent precipitation without an excessive amount of heat may have contributed to the establishment success of the willow at this site. Two early growing seasons in the second rotation were cooler (362 GGD in 2017 and 327 in 2019). The warmest early growing season was in 2015 (491). The average GDD for the late portion of the growing season was 1006. The later growing seasons were cooler in 2014 (933) and 2019 (891). Average precipitation for the early part of the growing season was 134 mm, with 2014 (77 mm) and 2016 (66 mm). The early growing season in 2015 was exceptionally wet with 285 mm of rainfall, more than double the amount in the other years of the trial. For the latter part of the growing season, average precipitation was 418 mm, and the only year that was more than one standard deviation from the mean was 2019, with 482 mm. Ultimately, there does not appear to be any remarkable differences between the first and second rotations in terms of heat and precipitation, which might explain the decline in overall mean survival in the two rotations.

There are multiple factors at this site, including the substrate conditions such as pH, salt content, lack of nutrients and limited structure that can inhibit the establishment and growth of many plants. Many of these challenges are improved with organic amendments [32,33,34]. However, there is an intermittent hardpan that has been identified at various locations across this site, which are known to impact tree growth [35], although the effects of compaction may be localized [36], These are not easily ameliorated without cost-prohibitive measures using heavy equipment, and some uncertainty as to how that will change the hydrologic conditions the willow ET cover was developed under. In the case of this trial, the hardpan interfered with successful organic matter mixing in locations where it is shallow and ultimately limited rooting depth. Therefore, it is a possible explanation for changes in yield and survival over the two rotations. Its effects might not be reflected in the 2016 yields because the crop is still developing its root systems. Previous studies on mineral soil have shown that root biomass in willow peaks after the third or fourth rotation [37].

Regression analysis indicated that there was a general site relationship among annualized yield in 2019, survival in 2019, and the percentage of the plot affected by a hardpan within 24 cm of the surface. Several candidate regression models were developed from the full model (Equation (1)) to characterize this relationship. The structure of the best performing model is Equation (2), in which the yield in 2019 was described by survival in 2019 and the percent of the plot with a hardpan within 24 cm of the surface. The model had an R-square of 0.8368; however, the Akaike Information Criterion (AIC) and Bayesian Information Criterion) BIC was more than double that of a WLS model with the same structure, implying that outliers and leverage points were having considerable influence on the coefficients and artificially increasing the R-square.

$$Yield={\beta }_0+{\beta }_1·sur+{\beta }_2·sur·hp\%+\epsilon$$

(2)

Thus, the final model (Equation (3)) had an R-square of 0.7144, with survival and the survival*hardpan interaction significant at p < 0.0001 and p = 0.0297, respectively. The majority of the variation was explained by survival alone, and the effect of the survival*hardpan interaction was significant but slight (<5% of the 71% of the variation explained by the model as determined by F-values during model selection). The VIF score was 2.48, which was under the conservative 5.0 threshold set for the model [24].

$$Yield=-0.0727+0.1221·sur-0.0004·sur·hp\%+\epsilon$$

(3)

The model indicates that willows that can maintain high survival will generally perform better (Figure 3 and Figure 4); however, certain individual clones appear to have high resilience to adverse conditions such as hardpan. The model suggests that a 10% increase in survival will correspond to a >10% increase in yield across all clones. Other phytoremediation studies have noted the importance of this connection between survival and growth, which are indicators of the effectiveness of the plants. Zalesny and Bauer [26] found that there was no correlation between first-year survival and biomass, but did find a strong correlation between survival and biomass by year 11. A similar pattern was found here, where first-rotation survival did not predict biomass at the end of the second rotation, but second-rotation survival was a key parameter. These results emphasize the importance of monitoring screening trials for multiple years, especially in phytoremediation applications.

Souch et al. [36] attributed a 12% decrease in biomass production to their most compacted loam soils. The model presented in this study also suggests that when survival is high, every 20% increase in hardpan extent will collectively decrease mean yield by approximately 5% (Figure 3). In addition, it suggests that extensive hardpans impact survival. However, the observed hardpan effect is still weak; at small, plot-size scales, more plastic or adaptive clones could thrive despite the presence of a hardpan. The impact of compaction on tree growth at various plot and operational scales can be highly variable [38]. The overall negative impact of hardpans may be offset at small scales by a variety of possible factors (e.g., breaks in the root limiting layer, microclimate, or topographic features that cause water to concentrate).

There is at least one notable exception to the pattern associated with yield, survival and percent of the plot impacted by hardpan, clone 05X-295-014 (S. purpurea x S. miyabeana diversity group) that is labeled as 1 in Figure 3 and Figure 4. This clone almost doubled its yield from 6.07 Mg ha⁻¹ y⁻¹ in 2016 to 11.97 Mg ha⁻¹ y⁻¹ in 2019, and survival increased from 75% to 87.5%, which is likely due to a new plant sprouting from pieces of stem that were left on site after harvesting. The high yield and survival occurred despite hardpan being present on about 46% of the area of these plots. Ultimately, 05X-295-014 was identified as a top-producing clone in a trial on mineral soils with yields of just under 15 Mg ha⁻¹ y⁻¹ [28] and was the top-ranked clone among clones that were common to both these trials. The performance of this clone should be examined further to understand other factors that contribute to its success across a range of conditions.

There were three other clones in the S. purpurea x S. miyabeana diversity group in this trial (Table 1), and two of them (05X-295-035 and 99217-015) were also in the top five clones in terms of 2019 yield and the third (9980-005) was ranked 8th. Specifically, 05X-295-035 had yields of 8.19 Mg ha⁻¹ y⁻¹ in 2016 and 9.41 Mg ha⁻¹ y⁻¹ in 2019, but survival of this clone dropped from 92.5% in 2016 to 47.5% in 2019, suggesting that future performance could be compromised. Clone 99217-015 had yields of 6.77 Mg ha⁻¹ y⁻¹ in 2016 and 8.48 Mg ha⁻¹ y⁻¹ in 2019, which placed it fifth in the ranking, with a survival level at 77.5% at the time of each harvest. The other clone from this diversity group (9980-005) had yields of 6.18 Mg ha⁻¹ y⁻¹ in 2016 and 7.74 Mg ha⁻¹ y⁻¹ in 2019, with survival of 75% in 2019. It appears that this diversity group might include individuals that do well on this site, and it would be worth testing other individuals and trying to understand if there are unique characteristics among these clones that make them suitable for ET cover applications at this site and others with similar stressful characteristics. Serapiglia et al. [28] noted that based on prior breeding and selection work that the S. purpurea x S. miyabeana crosses should have good yields and noted that it might be the triploid nature of the individuals rather than this specific diversity group that contributes to their high yields, so testing other triploids at this site may be another avenue to explore further.

While the performance of clones in previous trials at other sites may provide information to help select clones to screen for a phytoremediation application, they are not a reliable source of information to make final decisions. There was a total of 11 clones that were included in both the Serapiglia et al. [28] trial and the current trial, but the Spearman rank correlation was low and not significant (Spearman ρ = 0.20, p = 0.55), which further supports the importance of working through the steps in phyto-recurrent selection [18,26] and maintaining trials for long enough for the effectiveness of clones to be expressed under a range of conditions, especially on highly disturbed sites.

3.4. Management Implications

When deploying clones in phytoremediation or biomass production systems, it is advisable to use a collection of 3–5 or more clones to mitigate risk from pests, diseases or dramatic changes in weather that might impact one particular genotype [39,40]. In this study, clone rankings were not significantly correlated between the first and second rotations (Spearman ρ = 0.13, p = 0.48), indicating that early screening trials based on single-rotation data would likely misidentify superior clones. This trend aligns with Fabio et al. [41] and Berthod et al. [42], who found substantial genotype × environment interactions in shrub willow biomass traits, suggesting that genotype performance can shift dramatically between trial sites and conditions.

With yield as the selection priority, selecting clones for use on this site for an evapotranspiration cover after two rotations would suggest that a suite of the five top yielding clones (05X-295-014, 05X-292-035, 9970-053, SX61, and 99217-015) could be those chosen (Table 3), all of which are from the from the MIYA and PM diversity groups (Table 1). However, while yield is important and related to the amount of water that these plants would transpire in this ET system, these results suggest it is likely important to consider maintained survival as well, which has also been noted in other studies on highly disturbed land [43].

Clone 05X-292-035 was ranked second in terms of yield, but its survival dropped from 92.5% in 2016 to 47.5% in 2019. Likewise, 9970-053 was ranked third based on 2019 yield, but its survival dropped from 77.5% in 2016 to 50.0% in 2019. These drops in survival suggest that the potential for these clones to perform well over multiple rotations at this site is uncertain. These results suggest that the sixth and seventh highest-yielding clones in 2019, 9871-31 and SX64, may be better choices for the long term. They both have survival values above 70%. After two rotations, the top five clones to recommend for a larger deployment might be 05X-295-014, SX61, 99217-015, 9871-31, and SX64.

Zalesney and Bauer [26] demonstrated a screening tool for evaluating clones in phyto-recurrent trials that plots clonal means relative to trial means for two measurement periods. In their plot, utilizing diameters as their performance metric, they suggest clones positioned to the right of the origin were the best candidates, with those plotted in the upper right being optimal. In this trial, biomass and survival were plotted using the same approach, and the top seven clones fell in these same regions relative to the origin (Figure 5). However, it could be argued that the regions relative to the origin might have more nuanced indications about the optimal clones, as a 1:1 line also shows inherent stability of individual clones relative to the trial mean at each time step. In addition, using survival as a metric along with biomass, the 8 Mg ha⁻¹ y⁻¹ and 60% survival cutoffs that were imposed can also be plotted. Thus, by adding these constraints to this approach, the same or a very similar conclusion pertaining to the top five performing clones can be drawn. The importance of survival as a selection criterion is clearly illustrated by this instrument, especially if the survival rank orders are unstable.

Beyond yield, other criteria may prove important in selecting clones for use in deployments like ET covers or other objectives that are not considered in this study. Water use efficiency could be an important consideration in other ET covers [44]. A clone with high yield and high water use efficiency might be less desirable than one with low yield and high water efficiency. Alternatively, other applications of willow biomass crops may prioritize form [43], carbon sequestration [45], feedstock quality [46] biodiversity [47,48] or other factors. However, regardless of whatever plant attributes are prioritized, survival remains a critical factor in achieving project objectives.

4. Conclusions

This study evaluated 30 willow clones over two rotations on a former industrial site with alkaline waste substrates and a frequent hardpan layer. Substantial shifts in survival and yield between the first and second rotations underscore the importance of multi-rotation testing when selecting clones for deployment on challenging sites. Early performance was not a reliable predictor of continued performance, as clone rankings changed markedly between rotations.

Survival emerged as the primary driver of yield across clones. A regression model incorporating survival and its interaction with hardpan extent explained over 70% of yield variation. Although hardpan reduced yield where survival was already high, its overall effect was modest. Certain clones appeared to tolerate or avoid hardpan-related constraints, with 05X-295-014 performing especially well despite 46% hardpan coverage.

A subset of clones (05X-295-014, SX61, 99217-015, 9871-31, and SX64) exhibited strong yield and survival across both rotations and are recommended for further use in the evapotranspiration cover system. Clones from the S. miyabeana and S. purpurea x S. miyabeana diversity groups appeared to show the most promise. These results highlight the importance of integrating survival metrics, not just yield or other performance criteria, into clone selection and suggest that diversity in deployment can improve resilience to site-specific stressors.