This study was conducted at the USDA-ARS Small Farms Research Center near Booneville, AR (35.073˚N, 93.983˚W), in which 15-yr old east-west oriented loblolly pine tree alleys of varying width (3.7, 4.9, 7.3, and 9.8 m) were used in this experiment (Figure 1). Background on this stand and experimental area can be found by [6] and [7] . In fall 2011, alleys were manually cleared of woody vegetation then raked with a HR318 hay rake (Gehl Co., West Bend, WI) to remove pine needles and other litter from the alleys. Subsequently in fall 2011, each plot received the equivalent of 4483 kg lime, 90 kg P, and 90 kg∙K∙ha⁻¹ as a 0-20-20 blend. Plots were immediately disked with a Rhino 50 (Athens Plow Co., Athens, TN) at a maximum depth of approximately 13 cm to avoid damage to pine tree roots, to control weeds, and to prepare a smooth seedbed. Control plots were prepared in a similar fashion and were represented by an open unshaded area. On October 14, 2011, three replications of legumes were planted at 22.4 kg∙PLS∙ha⁻¹ for variety-not-stated (VNS) arrowleaf and 11.2 kg∙PLS∙ha⁻¹ ‘Ivory 2’ white clover (DLF Seeds, Roskilde, Denmark), using a Brillion broadcast seeder (Brillion Co., Brillion, WI). Seeding rates were approximately doubled to ensure germination and to mimic seeding rates that are likely to be used under these circumstances. Before planting, arrowleaf clover was inoculated with a rhizobium product to facilitate nodulation (N-Dure rhizobium, Verdesian Co., Cary, NC; Rhizobium leguminosarium biovar trifolii vesiculosum) at a rate of 241 g per 22.7 kg seeds. White clover seed was delivered with a proprietary coating at approximately 32% of total bag weight. The Brillion seeder was equipped with packing rolls in front and back of the seed box so the soil was sufficiently compacted during the planting process. Arrowleaf clover, as an annual, was replanted October 17, 2012, while white clover, a perennial species, was not replanted except in the 4.9-m alleys, a result of sporadic flooding that killed both species. In February 2012, LightScout 3 light sensors (Spectrum, Inc., Aurora, IL) were installed in each plot for the purposes of measuring photosynthetically active radiation (PAR) throughout the entire growing season in 2012 and 2013. Only six measuring units were available, so one unit was placed in the center of one randomly chosen alley in one replication per treatment.

Seedlings were counted 4 weeks after establishment (November 2011) and again in January 2012, November 2012, and January 2013 using a 0.75 × 0.75 m Vogel grid [18] placed randomly in each plot. Four frequency counts (100 cells total) were made in each legume plot. For the fall measurements, emerged arrowleaf seedlings were counted and for white clover, the number of remaining plants from initial establishment was recorded. Dry matter production was measured within 0.25-m² quadrats inside the exclusion cages, leaving a 3.8 cm residue height, with a gasoline-powered hedge trimmer. Plots were clipped approximately every 3 to 4 weeks from April to June in each plot. Samples were placed in paper bags and dried at 50˚C in a forced-air oven to measure DM yield.

A two-factor analysis of variance was conducted using the PROC MIXED

procedure of SAS (version 9.4, SAS Institute, Cary, NC) to determine treatment effects and interactions between alley widths (whole plot) and species (split plot). For all models, each sampling date, alley width, and species were considered fixed effects and replicates were random. Significance was judged at P < 0.05. Data were analyzed separately by date regardless of a lack of significant year effect (P = 0.67) to account for the abnormal amount of precipitation and high temperatures in 2012.

In November 2011, Booneville, AR received nearly triple the normal 30-yr average rainfall following clover establishment (Table 1), and air temperatures were similarly 20% warmer than average. Except for March and August 2012, total precipitation (653 mm) was 66% less than the 30-yr average (1083 mm; Table 1; [19] [20] ). Additionally, 2012 was warmer (18.3˚C) than the average in all months except November and was warmer overall by 2.7˚C (Table 1). In 2013, total precipitation was 8% greater than the 30-yr average, while the average temperature was 1.3˚C warmer than 30-yr average (15.6˚C; Table 1).

Photosynthetically active radiation, similar to temperature, followed an annual somewhat parabolic pattern for all treatments in both years of the study (Figure 2). In the open area (control), mean monthly PAR was greatest in May 2012

(1818 µmol∙m⁻²∙s⁻¹) and September 2013 (1763 µmol∙m⁻²∙s⁻¹). PAR was overall greater in the 9.8-m alley than the 7.3-m alley, which both peaked in June 2012 and June 2013. Similarly, June had the greatest PAR for the 3.7-m alley in 2013, while July had the greatest PAR in 2012 (Figure 2). The 4.9-m alley had greater annual PAR than the 3.7-m alley, but less than the 7.3-m alley. Overall, July (512.4 µmol∙m⁻²∙s⁻¹) had the greatest mean PAR in 2012, while May (467.4 µmol∙m⁻²∙s⁻¹) had the greatest mean PAR in 2013 for the 4.9-m alley (Figure 2).

Seedling count and DM yield differed (P < 0.05) between clover species and alley widths on at least two measuring dates (Table 2; Figure 3 and Figure 4). Arrowleaf and white clover differ in shade stress adaptability [21] . In November 2011, seedling count was greatest in the 9.8-m alley for white clover (Figure 3), which did not differ (P > 0.05) from the counts in the 3.7- and 7.3-m alleys or the open area for white clover. The lowest seedling count in November 2011 (0.54 seedlings∙m⁻²) was observed in the 4.9-m alley for arrowleaf clover, which did not differ from the 7.3- and 9.8-m alleys for arrowleaf clover (0.99 and 0.96 seedlings∙m⁻², respectively). The relatively high counts for both species in the 3.7-m alley (2.75 and 1.22 seedlings∙m⁻², respectively) was counterintuitive based on PAR levels, but was likely due to less weed competition or greater moisture retention compared with other treatments [6] [22] [23] [24] . In January 2012, the greatest (P < 0.05) seedling count for all treatments (2.02 seedlings∙m⁻²) was measured in the 7.3-m width alley for white clover (Figure 3). No seedlings were

counted in the 3.7-m alley or open area for white clover due to the extreme precipitation that occurred in November and December 2011 (Table 1). Seedling counts were greatest (1.47 seedlings∙m⁻²) in the open area for arrowleaf clover, which did not differ (P > 0.05) from the 3.7-, 7.3-, and 9.8-m alleys (arrowleaf) and 4.9- and 9.8-m alleys (white clover). The lowest seedling counts occurred in the 4.9-m alley for arrowleaf clover. Seedling count was greatest (1.58 seedlings∙m⁻²) in November 2012 in the 9.8-m alley for arrowleaf clover, which did not

differ (P > 0.05) from the 4.9- and 7.3-m alleys and open area for arrowleaf clover and the 4.9-m alley for white clover. The lowest counts observed in November 2012 (0.48 seedlings∙m⁻²) occurred in the 3.7-m alley of arrowleaf clover. There were 0 seedlings for other treatments of white clover due to severe drought throughout 2012 (Table 1; Figure 3). In January 2013, the greatest (P < 0.05) numbers of seedlings for all treatments were counted in the 4.9- and 9.8-m alleys for arrowleaf clover (0.98 and 0.95 seedlings∙m⁻²), respectively). The lowest number of seedlings was counted in the 3.7- and 4.9-m alleys for white clover, which did not differ (P > 0.05) from the 7.3- and 9.8-m alleys for white clover and the 3.7-m alley for arrowleaf clover. Because there was no evidence of white clover growth the previous fall, any white seedlings counted possibly emerged from seeds that were shed in 2012 or as growth from dormant stolons. As a perennial, white clover flowers and sets seed continuously during the growing-season [16] . No white clover seedlings were observed for the open area, as the drought likely had a more severe effect than in the shaded alleys where soil conditions remained more favorable for growth.

Dry matter production varied among treatments and time of sampling (Figure 4). Overall, yield was greatest (P < 0.05) in the open area for arrowleaf clover compared to all other treatments for all dates in the study (Figure 4). Dry matter yield was greatest in June for most treatments in both years (Figure 4). In March 2012, the second-greatest DM yield (0.90 Mg∙ha⁻¹) was measured in the 7.3-m alley for arrowleaf clover, which did not differ (P > 0.05) from DM yield in the 9.8-m alleys for arrowleaf and white clover and the open area for white clover. In April 2012, DM yield (2.48 Mg∙ha⁻¹) was second-greatest in the open area for white clover, which did not differ (P > 0.05) from the 7.3- and 9.8-m alleys for arrowleaf clover. Although DM yield for the arrowleaf open area (4.51 Mg∙ha⁻¹) was numerically greatest in June 2012, it did not differ (P > 0.05) from DM yield in the 7.3- and 9.8-m alleys for arrowleaf clover, which were greater (P < 0.05) than all other treatment combinations (Figure 4). Throughout all of 2012, the 4.9-m alleys for both clover species yielded the smallest DM, probably due to repeated flooding. Tree-forage interactions influence soil temperature and moisture, leading to differences in susceptibility to environmental extremes such as drought or flooding [21] [25] . Because arrowleaf clover concludes its annual life cycle by the end of June, further sampling did not occur. In 2013, DM yield decreased dramatically, and for the 4.9-m alley and open area for white clover, no growth was observed for the year, most likely due to the drought of 2012. Additionally as a result of the drought, white clover plots became increasingly infested with weeds (data not shown), which also influenced the persistence of this perennial clover. This differs from previous research [26] that confirmed cool-season legumes such as clover are good choices for use as living mulches in controlling weed pressure. By June 2013 (Figure 4), the 7.3- and 9.8-m alleys for arrowleaf clover had achieved significant growth, nearly equal to 2012, but yield was still less (P < 0.05) than in the open area; however, yields for all other treatments were lesser (P < 0.05). As expected, DM production was greater in the open area compared with the alleys, and differences were more pronounced later in the growing-season. This is in contrast to other shade-tolerant species such as Amphicarpaea bracteata L. and Lespedeza virginica L., which also have high yield and forage quality [21] .