2.1 Study site

This study was conducted at the Biofuel Cropping Systems Experiment (BCSE, http://glbrc.org/ (See Figure S1 for description of experiment)) of the Great Lakes Bioenergy Research Center (GLBRC), located at the Kellogg Biological Station Long-term Ecological Research site in southwest Michigan, USA (Robertson and Hamilton, 2015; 42° 24’ N, 85° 24’ W, 288 m a.s.l.). The BCSE is located in a diverse, rural- to-semirural landscape with cropping systems typical of the upper Midwest US. The climate is humid continental temperate with a 30-year (1981–2010) average annual air temperature of 9.9°C, ranging from a monthly mean of −4°C in January to 23°C in July, and average annual precipitation of 1027 mm evenly distributed throughout the year (NCDC, 2013). Soils at the site are in the Kalamazoo and Oshtemo soil series, fine-loamy, mixed, semiactive, and Mesic Typic Hapludalfs formed under a forested landscape in loamy outwash overlaying sand and gravel (Crum & Collins, 1995; Thoen, 1990). BCSE systems were established in a randomized complete block design, replicated in five 30 meter × 40 meter plots (Gelfand et al., 2020).

A total of eight experimental units comprise this experiment. Six candidate bioenergy cropping systems include no-till energy sorghum (Sorghum bicolor (L.) Moench., an annual, photoperiod insensitive sorghum hybrid, TAM 17900); switchgrass (Panicum virgatum L., variety Cave-in-Rock); giant miscanthus (Miscanthus × gigantea); native grasses (a polyculture of five grasses native to North America—Little bluestem (Schizachyrium scoparium [Michx.] Nash), Big bluestem (Andropogon gerardii Vitman), Canada wild rye (Elymus canadensis L.), Indiangrass (Sorghastrum nutans [L.] Nash), and switchgrass (variety Southlow); early successional vegetation (comprised of grasses and fobs from the pre-establishment seedbank and subsequent colonizers after land was abandoned); and restored prairie (a C3 and C4-species mix of species as described in Sanford et al. (2016) provided by a local prairie restoration contractor). Species composition of the successional and restored prairie systems are available at https://lter.kbs.msu.edu/datatables. Each bioenergy crop was planted and managed according to standard agricultural practices for the region (Table 1; Sanford et al., 2016). Additionally, two reference sites represent the historical and modern landscape: A continuous no-till maize (Zea mays L.) system representing a contemporary row-crop common in the US Midwest Corn Belt, and an 87-year-old managed hybrid-spruce forest located at the Kellogg Experimental Forest (KEF; 42° 21' N, 85° 21'W) depicting a 2.5-acre forest established in 1932 from abandoned agricultural land.

2.2 Data collection and instrumentation

Continuous measurements of albedo at the BCSE and KEF were made from May 2018 to December 2020. At the BCSE site seven measurement stations were deployed at the BCSE in each of six cropping systems. Each station consisted of a tower equipped with a four-component net radiometer (SN-500, Apogee Instruments), two net radiometers (Q.7.1, REBS, USA), and one soil water content reflectometer (CS616, Campbell Scientific Inc. (CSI)). One tower was also equipped with a precipitation gauge (TE525, CSI) and temperature sensor to measure air temperature and relative humidity (HMP 60, CSI). Maize (in 2018) and restored prairie (in 2019) both included an additional four-component net radiometer. The heights of the towers in each plot were adjusted over the study period in order to maintain a field of view above the canopy layer. Sensors were placed ~30–40 cm above the canopies, consistent with other studies in agricultural landscapes (Raupach, 1994; Zeri et al., 2011). At the forested KEF site, we deployed an eddy covariance tower 34 m tall equipped with a four-component net radiometer (CNR4, Kipp & Zonen, Netherlands), precipitation gauge (TE525), and an IRGASON (CSI).

2.3 Statistical analysis

Our quality control protocols consisted of checking data for values within the expected range, e.g., 0 < α_s < 1; 0 < SW_↓ < 1500 W m⁻². Data potentially subject to errors (i.e., instrument tilt; snow cover on an upfacing radiometer dome, temporary tower removal) were eliminated. In the case of maize and restored prairie, where two net radiometers existed simultaneously on the albedo tower, gap-filling was completed where needed. Otherwise, discarded observations were treated as gaps for both incoming and outgoing irradiance at the same interval for all sites. Larger gaps of several hours to up to 30 consecutive days occurred due to instrument failure (see Table S1 for complete tower coverage throughout the study period).

We calculated values for the growing season (GS), non-growing season (NGS), and annually. GS was defined as May through October (DOY of 121-304) following previous studies of similar bioenergy species in our region (Sciusco et al., 2020; Zeri et al., 2011), where plant emergence occurs in early May, and harvesting at the KBS BCSE is completed in November, after a killing frost. NGS includes all other days not defined as GS. The daily mean albedo for each site was computed by aggregating 5-minute data into half-hourly time steps.

Differences in annual and seasonal albedo were analyzed by mixed models analysis of variance (ANOVA) using the statistical package R (R Development Team, 2013), with crop type as fixed effects and years as random effects. For all tests, the statistical significance using Tukey HSD was evaluated at p < 0.05. Diurnal changes in ∆_α for each site were also explored by analyzing daily averages between our reference sites of maize and forest. Finally, temporal variances in surface reflectivity affected by daily changes were modeled with a local polynomial regression (LPR). This nonparametric technique was used to determine a weighted average in order to fit a smooth curve between our variables. This allowed any daily estimates near “outlier” points from being highly biased while still ensuring a smooth fit (Cleveland et al., 1990).

3.1 Annual row crops

The average annual albedo of our annual row crops ranged from 0.212 ± 0.005SE in 2018, to 0.270 ± 0.010 in 2019 and 0.271 ± 0.010 in 2020 (Table 2). Energy sorghum (0.270 ± 0.010) and maize (0.247 ± 0.010) had lower albedos than most other cropping systems throughout the study period. During the growing season, maize albedo (0.184 ± 0.002) was lower than energy sorghum (0.208 ± 0.004) (Figure 1). During the non-growing season both crops had slightly higher albedos than most other cropping systems at 0.362 ± 0.018 and 0.373 ± 0.013, respectively.

3.2 Perennial crops

Combined, perennial crops had consistently higher annual and growing season α_s compared to annual crops (p < 0.05) (Figure 1). Annual average α_s was highest in early successional (0.288 ± 0.010), and miscanthus (0.271 ± 0.010), intermediate in switchgrass (0.265 ± 0.01) and restored prairie (0.264 ± 0.010), and lowest in native grasses (0.254 ± 0.010). Among the perennial crop types, miscanthus had the highest mean growing season α_s in 2018 (0.251 ± 0.003) and 2019 (0.227 ± 0.002) and native grasses lowest (0.186 ± 0.002) (Figure 1a). Albedo was more variable during the non-growing season and ranged between 0.24 and 0.85. The non-growing season α_s was much higher than the growing season α_s but with statistically insignificant differences among crops.

3.3 Comparisons across systems

Differences in the growing season α_s and annual α_s were statistically significant (p < 0.05) among crops and between forest and perennial crops, as well as between perennials and the maize system. The annual mean albedo for all systems ranged from 0.134 in the reference forest, 0.247 for the reference crop maize, to 0.264–0.288 for the six bioenergy crops (Table 2). Mean diurnal variation of α_s during different seasons were apparent for all sites, but varied in magnitude and by season and cropping system (Figure 2).

3.4 Seasonality

Variations in α_s expected upon converting either maize or forest to a candidate bioenergy crop (Equation 2; Figure S2), were similar among crops across both growing season and annual time frames. Conversion of forest to bioenergy crops resulted in higher RFs than conversion of maize to bioenergy crops (Figure 3). Average cooling effects from modeling the conversion of maize to another bioenergy crop yielded −3.83 ± 1.00 W m⁻² (Table 3), while modeled conversions from forest to another bioenergy crop showed a − 16.75 ± 3.01 W m⁻² cooling effect. Highest daily averages in mean growing season RF were observed in miscanthus (RF_FOREST: −20.99 ± 3.45 W m⁻²; RF_MAIZE: −9.49 ± 1.66 W m⁻²) and switchgrass (RF_FOREST: −17.37 ± 2.68 W m⁻²; RF_MAIZE: −6.07 ± 0.96 W m⁻²). There was also a clear seasonality in RF when modeling the conversion from annual row crop maize to another candidate crop. During the non-growing season, average cooling effects of maize were small, on the order of −0.95 ± 0.88 W m⁻² (Table 3) for all perennials cropping systems versus a − 4.54 ± 0.93 W m⁻² during the growing season. Overall, early successional, miscanthus, and native grasses resulted in the greatest average annual cooling relative to the no-till maize system.

4.1 Hypothesis 1: Albedo of perennial versus annual cropping systems

For the three-year period of this study, annual mean α_s ranged from 0.134 for the reference forest to 0.247 for the reference maize and from 0.264 to 0.288 for our annual and perennial bioenergy cropping systems (Table 2; Figure 1). These values are consistent with earlier α_s comparisons of perennial grasses and annual row crops of maize and energy sorghum (Campbell & Norman, 2012; Fritschen, 1967; Krishnan et al., 2012; Miller et al., 2016).

Although our perennial crops overall (0.269 ± 0.01) had a higher α_s compared to annual crops (0.259 ± 0.01), energy sorghum (0.270 ± 0.01) was an exception in that its α_s was more similar to that of the perennial crops, likely due to a later planting date than maize. It is also worth noting that our no-till maize system was a continuous rotation rather than rotated with soybean. This provides a more conservative estimate of α_s change from conventional cropland, first because soybean has a higher surface reflectance than maize, and secondly because no-till management (used for soil and water conservation) is more reflective than conventional tillage due to more surface residue and undisturbed soil. Energy sorghum, which is also a no-till annual crop, also displayed a higher α_s (0.270 ± 0.01) compared to maize, but was still lower than perennial crops. Energy sorghum is usually planted several weeks later than maize in our region, which appeared to be the major factor contributing to its greater α_s than maize. With the soil's being bare throughout the months of April, May and early June, planting date can affect α_s as the irradiance reaching surface soil occurs for a much longer period compared to perennials, which typically emerge in early May in our region, and thus energy sorghum can have a greater warming effect in early spring. Thus, crop-species can influence surface-induced α_s which can result in substantial climate change effects (Chen et al., 2019). In general, differences among cropping systems were greatest early in the growing season. Differences were less apparent at peak growing season when all systems had closed canopies, resulting in greater irradiance interception.

Our reference forest was a mix of both coniferous and deciduous species. Bonan (2008) showed that forests have lower surface albedo than most other cover types, which contributes to climate warming. Sciusco et al. (2020) used satellite imagery to show that α_s within forest landscapes in southwest Michigan were ~ 3% lower than α_s of those of bioenergy croplands. We also found consistently lower albedos in our reference forest system relative to both perennial and annual row crops (Figure 2, α_s: 0.135).

4.2 Hypothesis 2: Seasonal differences in albedo among cropping systems

Seasonal albedo patterns were observed within our study sites. (i.e., growing season, winter, monthly, annually). Differences in α_s were most likely influenced by differences in agronomic practices including planting densities, planting times, harvest dates, and stover retention, as well as differences in plant morphology and canopy growth. In winter months (January–March), temporal variations among all bioenergy crops appeared quite high, especially when snow was present. In early spring (February–April), the perennial grass cropping systems had consistently higher albedos than no-till maize. However, during the growing season (May–October), all bioenergy crops had similar α_s values. After harvests (November–December), α_s was elevated in all bioenergy croplands. The reference forest had the lowest monthly α_s of all study sites, reflecting its conifer composition, and changed little throughout the year (Table S2). The effect of tall, complex forest canopies on α_s during winter periods can affect the amount of radiation absorbed or reflected. Similar to other studies (e.g., Betts & Ball, 1997; Robinson & Kukla, 1984), albedo in winter mixed forests are estimated to be around 0.110–0.150. Thus, changes in α_s observed during both the growing and non-growing seasons can vary during periods of high snow, winter thaws, weather, and sunlight intensity, and can have significant cooling of seasonal mean and annual temperatures.

4.3 Hypothesis 3: Climate, agronomic practices, and plant species effect on albedo

Finally, we hypothesized that albedos of different bioenergy crops are time dependent as each species and ecosystem are affected by climate, seasonality, and agronomic practices. Modifying surface albedo through alternative avenues, such as crop residue management and landscape conversion from climate and agronomic practices, may add additional cooling benefits to the warming climate.

Zeri et al. (2011) and Miller et al. (2016) noted the differences in albedo in perennials as well as annual row crops were highly influenced by planting density, plant morphology, and canopy architecture. Maize had much lower α_s during the growing season than did the perennial systems (Figure 1). This was due to the immediate growth of perennials at the start of the last frost in March. As maize was planted at the sites around early May, the land surface was still moderately exposed due to late seeding and small seedlings during the same period. Bare earth left exposed to the atmosphere during the first few weeks of the growing season allows the maize system to absorb more solar radiation than vegetated fields until their canopies fully develop; hence overall α_s of the perennial systems remains much lower.

Similar agronomic variables such as phenology, cover crops, crop residue management, planting date, row spacing, crop variety, and canopy duration also play an integrative role in affecting albedo (Campbell & Norman, 2012; Luyssaert et al., 2014; Moore et al., 2021; Odum, 1984). Energy sorghum's growth habit is similar to that of maize, but with more side shoots, a more extensively branched root system, and it is planted in conventional 0.38 m rows rather than 0.75 m for maize, which allows for optimum use of moisture and sunlight, resulting in a higher surface reflectance more similar to the albedo of perennial crops than to that of maize. After approximately 6 weeks to fully establish a closed canopy, energy sorghum remains green until fall harvest. Consequently, its average albedo (0.270) was comparable to that of perennials.

Albedo during the non-growing periods was markedly higher when the landscape was completely or partially covered with snow. During the winter periods, stover breaks up the snowpack, leading to changes in energy reflection. Though not part of this study, previous research has shown that winter cover crops can also induce a localized cooling effect by reflecting more incoming radiation back into the atmosphere (Lugato et al., 2020). Taken together, our results suggest that large scale conversions of landscapes by expanding bioenergy cropping systems may significantly affect local climate in the Midwest U.S. due to altered albedo values (Georgescu et al., 2009, 2011; Mykleby et al., 2017).

4.4 Radiative forcings from forest are more pronounced compared to maize

When comparing RFs for alternative representations of historical (forest) and contemporary (maize) converted land uses, RFs had strong seasonal variations. Overall, converting forest to cropland led to substantial decreases in RF, and therefore climate cooling, with the degree of cooling dependent on crop type. The peak difference in annual mean RF between maize and other bioenergy croplands (except forest) averaged −3.83 ± 0.78 W m⁻², with early successional, miscanthus and native grasses having the highest cooling potentials at −6.77 ± 1.16 W m⁻², −5.13 ± 1.02 W m⁻² and − 4.26 ± 0.91 W m⁻², respectively.

This was similar to prior findings of Miller et al. (2016) who used highly resolved albedo measurements to compare miscanthus, switchgrass, and annual row crops of rotational maize/soybean, and found that the perennials switchgrass and miscanthus had a daily cooling potential of −5 W m⁻², and miscanthus of −8 W m⁻², compared to maize. Similarly, Sciusco et al. (2020) and Sciusco et al. (2022), who integrated spatial and temporal changes as main drivers of albedo variations, showed that cropland had higher albedo and intra-annual variabilities, with an average RF between −5.6 W m⁻² to −1.2 W m⁻² when compared to forests. Abraha et al. (2021) also assessed the biogeophysical climate impact of albedo using multiple modeled conversions from an unconverted reference CRP grassland using radiation measures from eddy covariance towers, which showed that switchgrass and restored prairie fields provided albedo-induced cooling.

Forest conversion can lead to significant albedo-induced cooling, but harvesting forest can create large initial carbon debt requiring long payback periods before net gains in albedo-induced cooling are achieved (Mykleby et al., 2017). Harvesting large forests and planting bioenergy crops can reduce or remove carbon stores within the crop and soil itself (Chen et al., 2004; Noormets, 2016), which could potentially cause higher emission of GHG, i.e., warming effects. Changing forest cover can further affect climate change through complex forest-atmosphere dynamics including plant phenology, land changes, and climate (Duveiller et al., 2021). Fu et al. (2021) observed negative RF due to albedo-induced GWP by deforestation, while Bastable et al. (1993) also noted that the widespread deforestation for croplands could lead to positive feedback effects' dampening the cooling effects from elevated albedo. Conversions of non-agricultural land such as forests with high initial carbon stocks to the cultivation of another bioenergy crop could also result in large carbon debt before net mitigation is achieved as much as a century later (Amiro et al., 2010; Field et al., 2020; Robertson et al., 2017). However, forests also aid in generating cloud cover, and reflect more radiation back into the atmosphere, adding to the long-term net cooling effect on landscapes. Thus, as negative RFs can contribute to climate change mitigation by increasing reflected surface radiation (Caiazzo et al., 2014), understanding how these changes can alter the Earth's surface properties are crucial for developing land-based mitigation policies.

4.5 Assumptions and uncertainty

Change in α_s is highly correlated with land surface properties such as vegetation type and cover, snow cover, soil moisture (Ahmad & Lockwood, 1979; Weidong et al., 2002), and changes in climate such as drought or floods (Bright et al., 2012). Modifying the surface albedo of croplands through different agronomic practices such as planting date and density, crop residue management, tillage, and harvest may affect warming and cooling influences (Davin et al., 2014). The vegetative structure of crops can also affect radiation absorption and reflectance; crops with complex geometries and textures can have lower albedos than those with single stalks and leaves such as maize. Seasonal changes from bare-ground to seeding, growth, and finally senescence and harvest can also affect surface albedo through changes in crop height, crop cover, leaf texture, and leaf age (Henderson-Sellers & Wilson, 1983; Monteith, 1959).

Management effort, economic costs of inputs, and landowner preference for a specific crop will also affect crop choice, as will environmental impacts unrelated to climate mitigation (Robertson et al., 2017). Other environmental impacts include nitrogen and phosphorus pollution from excess fertilizer use, as well as potential biodiversity impacts on pollinators, insect pests and pest predators, and other taxa. Energy sorghum, for example, takes much more management effort and cost than an equivalent perennial crop, and receives more herbicides, is often fertilized more heavily and often tilled, leading to soil and nutrient loss. Likewise, miscanthus, despite its high productivity (Gelfand et al., 2020; Heaton et al., 2008), as a non-native grass supports far fewer insect and vertebrate taxa than switchgrass or other native grasses, and is also invasive (Lowry et al., 2022; Williams & Feest, 2019). Thus, environmental and economic trade-offs must be considered in addition to albedo considerations when choosing among alternate bioenergy crops.

Accurate quantification of RF from a specific land surface depends on reliable measurements of atmospheric transmission. This can be determined using the atmospheric transmittance (T_a) (Chýlek & Wong, 1998; Lenton & Vaughan, 2009) or clearness index (K_T) (Bright et al., 2012; Paulescu et al., 2021; Sciusco et al., 2022). Both measurements account for changes in solar irradiance measured on the surface and its counterpart measured at the top of the atmosphere. Muñoz and Kravchenko (2011) and Cherubini et al. (2012) assume a global constant of 0.854 for upward atmospheric transmittance from clear sky conditions. However, this is usually initiated at a 60° angle, and is not a good representation in regions of highly variable weather. Instead of using the global mean we employed a daily calculation of T_a as the ratio of SW_↓ /SW_TOA from daily measurements of T_a obtained from our albedo towers in each cropping system. By calculating T_a for each individual day, we reduced bias during periods of highly variable weather and cloud cover, and thereby reduced error estimates in radiative forcing by up to 30% (Sciusco et al., 2020) As the differences in RF with the use of K_T and T_a were negligible, we determined the use of T_a for calculations in RF were sufficient.

Larger gaps of several hours to up to thirty consecutive days existed within our dataset due mainly to the agronomic management needs of croplands (i.e., removal of towers during seeding, herbicide spraying, fertilizer input, harvest, etc.), as well as unforeseen instrument failure. In addition, the pilot year was initiated in mid-May of 2018, such that late winter and early spring measurements for our first year were missing (Table S1). This may have caused slightly higher surface reflectivity than on average, compared to those measured at eddy covariance towers, which are usually permanent fixtures within landscapes. However, despite this drawback, the use of multiple mobile micrometeorological towers was highly effective in addressing the potential of changes in surface albedo in multiple candidate bioenergy landscapes.

Future research should examine the interactions of albedo with other biogeophysical, biogeochemical, and micrometeorological changes. Additionally, long-term observation-based, quantitative estimates of annual to decadal-scale changes of shortwave radiation may also be useful for capturing rare events that are not detected by shorter timescale methods.