Cocoa agroforest multifunctionality and soil fertility explained by shade tree litter traits

1. Manipulating plant functional diversity to improve agroecosystem multifunctionality is a central 27 challenge of agricultural systems worldwide. In cocoa agroforestry systems (cAFS), shade trees 28 are used to supply many services to farmers, yet their impact on soil functioning and cocoa 29 yields is likely to vary substantially among tree species. 30 2. Here, we compared the impact of five shade tree species ( Canarium schweinfurthii (Canarium), 31 Dacryoides edulis (Safou), Milicia excelsa (Iroko), Ceiba pentandra (Kapok tree) , Albizia 32 adianthifolia (Albizia)) and unshaded conditions on the functioning of poor sandy savannah soils 33 within eight cocoa farms in Central Cameroon. We assessed the effects of plant functional traits, 34 leaf litterfall and fine root biomass on a range of soil functions and on cocoa yield. 35 3. Shade trees generally improved soil pH, NH 4+ , NO 3- and Olsen P content, biomass production 36 of bioassays, and soil total C and N content, while leaving cocoa yields unchanged. However, 37 these effects varied largely among species. Improvements of soil functions were low under the 38 two fruit trees ( Canarium and Dacryodes ), medium under the legume tree Albizia, and high 39 under the two timber trees ( Milicia and Ceiba ). Low litter recalcitrance was most strongly associated with increases in soil fertility indicators such as N and P availability, whereas soil C 41 and N content increased with litter Ca restitution. 4. Synthesis and applications . We demonstrate that cocoa agroforest multifunctionality is approach to other belowground traits and other aspects of multifunctionality such as long-term cocoa health and yield.


85
Ecosystems are expected to provide multiple functions and services for human society. Hence, 86 ecosystems' health is now mainly assessed through their multifunctionality (Maestre et al., 2012;Wagg 87 et al., 2014). Ecosystem multifunctionality is assumed to be maintained with high levels of aboveground 88 and belowground biodiversity (Delgado-Baquerizo et al., 2016). However, the identity of species that 89 live in the ecosystem, as well as their functional traits (defined as any morphological, physiological or 90 phenological feature measurable at the individual level; Violle et al., 2007), are at least as important as 91 biodiversity per se in explaining the effects of species richness on ecosystem multifunctionality (Peltzer 92 et al., 2009;Maire et al., 2018). In the agricultural sector, stakeholders and managers are increasingly 93 considering the identity of species that are associated with the crop, and their functional traits, in order 94 to improve agroecosystem multifunctionality (Martin & Isaac, 2015, 2018. For instance, Blesh (2018) 95 recently found that cover crop mixtures with complementary functional traits increased multifunctionality.

96
Likewise, Damour, Navas, & Garnier (2018) proposed a trait-based approach framework which uses 97 traits to select optimal plant community compositions and design agroecological cropping systems.

98
While the use of plant functional traits to improve agroecosystems multifunctionality have been 99 conceptualized in several recent works, it has yet scarcely been put into practice in the field. Improving  Improving agroecosystem multifunctionality by managing plant community composition represents an 105 opportunity to increase the yield in cocoa agroforests (cAFS) from West Africa, where 70% of world 106 cocoa is produced. Farmers introduce shade trees in cAFS to provide an understory shade that reduces 107 cocoa physiological stress, pest and diseases outbreaks (Andres et al., 2016). The shade tree species 108 used in cAFS are very diverse and are selected both for their shade cover and for the provision of 109 additional goods to local populations (firewood, fruit, timber, medicine), which may reach up to 60% of 110 total cAFS plot revenue when adequately managed (Juhrbandt, 2010). Nonetheless, shade trees can 114 introduction may compensate for their light interception effects on cocoa yield (Isaac et al., 2007a), and 115 that (ii) these effects are expected to vary strongly with shade tree species (Wartenberg et al., 2019). In 116 this context, testing whether differences among shade trees functional traits can affect cocoa yield and 117 soil fertility while providing goods for farmers is of high interest.

118
Shade tree effects on cocoa yield, nutritional status and soil fertility has been studied both at the 119 community (Blaser et al., 2017;Niether et al., 2019)   cAFS multifunctionality remains to be tested.

132
We aim to determine whether shade tree traits could be used to identify the shade tree species with the 133 highest improvement of cAFS multifunctionality components compared with unshaded cocoa. We first 134 hypothesize that shades trees promote cAFS multifunctionality through increase of soil nutrient 135 availability, C storage and goods production. We then hypothesize that soil nutrient availability and C

189
After thorough mixing of the remaining soil, an aliquot of about 260 g (dry weight) of soil was sieved at     Table S1 for further details).

231
Total C and N contents were determined by dry combustion using a CHN micro-analyzer (

241
Leaf N and P resorption efficiencies were estimated as described by Freschet et al.

252
where CWM x is the CWM for trait x, p is the relative proportion of either cocoa or shade tree to the total 253 leaf litterfall collected above each soil sampling location, and t is the leaf litter trait value for cocoa or 254 shade tree for the corresponding location (Table S1). We first confirmed with linear mixed-effects models 255 that there were no farm (replicate) effects on CWM traits and soil functions (Table S3 and Table S4).

256
CWM traits and soil functions differences between cocoa -shade tree associations were then assessed

262
We considered the following agroecosystem functions: total soil organic C, total N, NO 3 -, NH 4 + , Olsen P 264 by shade trees: fruit or timber. An agroecosystem multifunctionality index was calculated under each 265 treatment. Briefly, values of each function were standardized by its maximum across all treatments, and 266 thus ranged between 0 and 100%. Multifunctionality was then defined as the number of standardized 267 functions under each cocoa -shade tree association that had a value above a threshold T (30, 50, 70 268 and 90%). C mineralization, which is considered as a negative process relative to cAFS functioning, 269 was inverted before being standardized. Fruit and timber production were either attributed a value of 0

271
Beforehand analyses showed that soil functions under cocoa -shade tree associations were better

290
Among shade trees, Dacryodes was the smallest and the thinnest species, while Ceiba was the tallest 291 and widest species (Table 1). Litter N and P contents were higher for the deciduous (Milicia, Ceiba and 292 Albizia) than for the evergreen species (Canarium and Dacryodes), the latter being at the same level as 293 cocoa. Cocoa had higher leaf N and P resorption efficiencies than shade trees (except for Dacryodes N 295 (< 6), higher tannin content and Lignin:N ratio than litter from the deciduous shade tree species (Table   296 1). Among the deciduous species, litter from N 2 -fixing Albizia had the highest N content, and the lowest 297 lignin:N ratio, Mg and tannin contents.

298
Association with shade trees affected strongly leaf litterfall but did not modify patterns of fine roots 299 biomass significantly (Table 2). Cocoa litterfall was decreased by half under shade trees, regardless of 300 the shade species considered. Shade tree litterfall varied between species and ranged from 2.6 to 4.7 t 301 DM ha -1 ( Table 2). The total amount of litterfall was lower for unshaded cocoa and cocoa shaded with

302
Canarium than for the other associations. Cocoa leaf litterfall amounted for only 21 to 36% of total leaf 303 litterfall under shade trees ( Table 2). As a result, community weighted mean litter traits in association 304 were thus mostly driven by the characteristics of the shade tree species (Table S2).

305
Community weighted mean litter quality was improved in cocoa -deciduous species associations with 306 increased litter pH and decreased lignin:N ratio ( Figure 1a; Table S2), as well as increased litter N, P, 307 K, and Ca contents (with the steepest increase of Ca content with Ceiba and and N content with Albizia, 308 Figure 1b; Table S2). Association with evergreen species decreased the averaged litter quality through 309 an increase of tannin content and a decrease in litter pH and Mg content ( Figure 1, Table S2).  (Table 3). Soil pH was only improved under Milicia and Ceiba, while soil 314 under Albizia was acidified compared to the other associations (Table 3). Only association with Ceiba 315 led to a significant increase in soil C and N contents. Overall, shade trees had relatively little impact on 316 soil C mineralization, nitrification and cocoa yield.

317
Agroecosystem multifunctionality was higher under cocoa -shade tree associations than under 318 unshaded cocoa, yet depended on the shade tree species and the threshold considered (Figure 2).

319
These improvements were more obvious at the threshold value of 50%, where multifunctionality index 320 increased from 2 for unshaded cocoa, to 4 for associations with Dacryodes, 6 with Canarium and Albizia,

321
and 8 for associations with the two timber trees. Higher multifunctionality under the deciduous trees as 322 compared to unshaded cocoa corresponded to higher NO 3 -, Olsen P content and bioassay production,

323
as well as the additional fruit or timber production of all shade trees except Albizia ( Figure S1). Impacts Dacryodes, and to timber production and soil pH increase under Milicia and Ceiba (Figure 2 and S1).  impact of associations with deciduous species on soil N and P availability may be linked to the high 389 quality of the deciduous species litters (low lignin:N ratio and tannin content) as much as to its high N 390 and P content. Indeed, despite similar N and P content, litter from the evergreen tree Dacryodes, with 391 lower overall quality, improved less soil N and P availability than the deciduous tree species.

392
Our finding that litter low recalcitrance is associated with higher soil C sequestration is in line with the

403
further studies will need to go beyond classical measurements of standing biomass and to focus more 404 specifically on root turnover, exudation rates and mycorrhizal associations.

405
Overall, cocoa -Ceiba was the association that increased most cAFS multifunctionality, along with cocoa     RI varies from 0 to 1 and represents the sum of the Akaike weights of the models in which each variable is used. RI represented in red and blue correspond to significant variables ( † P < 0.10; *: P < 0.05; **: P < 0.01; ***: P < 0.001) with positive and negative coefficient, respectively (see Table S6 for further details).

TABLE S3
Linear mixed-effects models of farms identity impacts on cocoa -shade tree associations community weighted mean attributes, with associations identity set as random effects. The analyses were performed on the treatments represented by one modality per farm (i.e. unshaded cocoa, cocoa -Dacryodes edulis and cocoa -Ceiba pentandra associations).