Crop yield potential as telltale indice of soil weathering extent and fertility status: The case of East African Highland Bananas

In the African Great Lakes Region, bananas are grown on a diversity of soils with different weathering stages. However studies using the crop yield potential as a mean of soil weathering degree assessment are still scanty. Bananas were grown on five soils types to test if such a relationship could be ascertained. Mineralogical composition, elemental total analysis, routine chemical analysis, oxalates and dithionite-citrate-bicarbonate (DCB) extractions on the 0-20 and 20-40 cm soil layers were used as soil characteristics. Banana yield was higher in Cibitoke where the soil was characterized with relatively high values of total reserves in bases (TRB) and the weathering index of Parker (WIP). In contrast, no yield was recorded in Gitega where the soil had relatively lower values of TRB and WIP and high Fe DCB/Fe total ratio. Furthermore, banana yield was strongly and significantly (p<0.05) correlated with the TRB, the mineral reserves, Fe oxalate/Fe DCB ratio, the silt content and poorly correlated with the soil pH, total carbon and nitrogen, available P, exchangeable bases and the CEC. It was concluded that banana yield potential reflected well the soil weathering extent and in complement to soil properties related the routine analysis, the total analysis provide even more precision to elucidate the snapshot of the soil properties in the light of the observed banana yield potential.

dessert bananas types and plantains, they compose the three banana categories encountered in East Africa (Karamura, 1998).
In that region, bananas are found on a diversity of soils, among which Ferralsols, Nitisols and Acrisols are most common (Davies, 1995;Rishirumuhirwa, 1997;Okalebo et al., 2006) and represent respectively 27, 18 and 8 % of the production area (Eledu et al., 2004).These soils are generally characterized by low levels of fertility (Davies, 1995;Rishirumuhirwa 1997;Eledu et al. ,2004) resulting from their high degree of weathering and leaching, which in turn are the cause of low crop productions in the majority of African countries (Okalebo et al., 2006).Thus for instance, by a simple observation, it is easy to recognize that plant growth and production is poor in one or another site and relatively better in others.As crop production depends on various environmental factors including climatic, edaphic, parasitic and human factors (Godefroy et al., 1991), it is not easy to assign the performance levels at one or another factor affecting the crop yield.Owing to the low inherent fertility of the AGLR dominant soils (Ferralsols, Acrisols and Nitisols) (Davies, 1995) and their more or less advanced weathering stage (Okalebo et al., 2006), the physico-chemical soils' characteristics, may constitute one of the major factors influencing yields variation.As plant growth, vigour and yield are dependent upon the availability of a number of essential element nutrients and other soils properties such as its texture or structure, salinity, acidity, waterlogging, or compaction, Plant growth and yield may be excellent indices of inherent soil fertility and weathering extent.The physico-chemical characteristics of soil can serve in the soil weathering status assessment (Souri et al., 2006).In particular, weathering indices are commonly used to characterize in situ weathering soil profiles as well as for the assessment of soil fertility and soil development (Price and Velbel, 2002).Several types of indices have been used by different authors to characterize the weathering stage of tropical soils: the ratios between acid ammonium oxalate extractable iron and aluminum (Fe o , Al o ) and Na dithionite-citratebicarbonate (DCB) extractable iron and aluminum (Fe d , Al o ), the quantity of DCB extractable free iron and aluminum (Fe d , Al d ) and total iron content (Fet) (Delvaux et al., 1989 ;Mahaney et al., 1994 ;Dalmolin et al., 2006;Zhang et al., 2007;Moody et al., 2008;Neto et al., 2010), the total reserve in bases (TRB) (Herbillon, 1986;Delvaux et al., 1989) and others indices types among which the weathering index of Parker (Delvaux et al., 1989;Price and Velbel, 2002;Patino et al., 2003;Haskins, 2006;Souri et al., 2006).
The ratio Fe d /Fe t , which expresses the transformation of iron-containing silicates into pedogenic iron oxides, is often employed as a weathering index (Kämpf and Curi, 2000).The sum of total alkaline and alkaline earth cations constitutes the TRB which estimates the content of weatherable minerals in mineral soil horizons Syldie 1363 (Herbillon, 1986).TRB includes the bases occluded in primary minerals as well as those located on exchange sites and also possibly in the secondary clay minerals (Delvaux et al., 1989).Among numerous chemical indices used to characterize the profiles alteration degree, Price and Velbel (2002) found that the weathering index of Parker (WIP) had the merit to be valid as well as for profiles on homogeneous and heterogeneous parental material.The WIP has been suggested by Parker in 1970; it allows determining the alteration degree of different rocks.This index is very sensitive to chemical changes in the early stages of alteration (Palma et al., 2003).
There have been many studies on soils characteristics and crop performance relationship since Belgians period in the two aforementioned countries (Burundi and DR Congo) but these works have focused much more on relationship between industrial crops (cocoa, coffee, cotton, hevea, oil palm, tea, sugar cane) and classical agronomic indices of fertility (soil organic matter content, exchangeable bases, soil' nitrogen content, available P) throughout fertilization trials and management practices.Studies dealing with others chemical' soil quality indicators, that is, weathering indices for predicting yield especially on banana in the study area are very rare.
From recorded total banana yield, physico-chemical characteristics and some weathering indices of different soils types, this present study seeks to: 1. Assess the trends of the different physico-chemical soil characteristics and weathering extent of 5 soil types in the AGLR 2. Determine the relationship between the different parameters with focus on observed highland bananas yield levels and the soil chemical properties including the weathering indices; to find out what indicators correlate best with yield.

Experimental sites
Five sites (three in Burundi and two DR Congo) located in different agro-ecological zones (AEZ) were selected as part of the CIALCA project in order to study mulch-based cropping practices (DMC) in banana-bean intercropping system.The sites were selected based on their representativeness of the AEZ regarded as principal banana production areas in those countries.Some characteristics of the experimental sites are summarized in the Table 1.

Agronomic potential evaluation according to Banana performance
A beer banana (Musa acuminata Colla (AAA-EA)), cultivar Igitsiri or Ndundu (respective local name in Burundi and DR Congo) which belongs to the group of EAHB was planted at the above mentioned sites to evaluate the agronomic potential of different soils types.Bananas were intercropped with common bean (Phaseolus vulgaris L.) (a legume).Various no-till treatments with mulching were  2006), *** Annual mean rainfall and temperature for the period 1985-1995and 2007-2009 (IGEBU) (IGEBU); ****Legumes, cereals and roots and tubers crops.compared to a control treatment with tillage and no mulch in a randomized complete block experiment with 4 replicates.The plots without mulch were ploughed twice a year and superficial weeding with hand hoe was performed for the rest of the year to control weeds.No other inputs (fertilizers, manure, lime, mulch) were applied.In the present study, only the control plots were used because they were assumed to reflect as much as possible the natural soil fertility status and hence provide a proper estimation of the agronomic potential and a snapshot of the weathering stage.Banana performance (yield, growth rate) which partially reflects the agronomic potential of sites was evaluated on the basis of yield recorded 30 months after field establishment.Indeed, after this period, bananas were supposed to have completed the first cycle at all sites.This means that banana bunches were harvested and the corresponding weight recorded systematically, as each individual plant became physiologically mature within 30 months.Thus, the total yield per hectare was computed based on the total weight of bunches harvested in the 4 control plots of 100 m 2 each (Burundi) or 96 m 2 (DR Congo) in relation to the surface of 1 ha.

Soil sampling and analysis
In each site and just at the trial inception, soil samples were collected from two different soil layers (0-20 and 20-40 cm) with an auger.For each soil depth, a composite sample was made from 5 subsamples in each of the 4 control plots.For the 5 sites, a total of 10 composite samples (5 from the surface layer and 5 from underlying layer) were therefore subject to different physico-chemical analyses.Particle size analysis was performed on the fine fraction using the pipet method after ultrasonic dispersion in salt-free water with Na + -resins without any pretreatment for organic matter destruction.Soil pH was measured potentiometrically in H 2 O (pH water ) and 1 M KCl (pH KCl ) in 1: 2.5 soil-solvent suspension.The total N and carbon were determined by gas-liquid partition chromatography (GLPC) after a dry combustion using an auto Analyzer CN Flash EA1112.Available P was extracted using Mehlich-3 extraction solution (Mehlich, 1984) and was measured by Atomic Emission Spectrometry (ICP-AES).Cation exchange capacity (CEC) and the exchangeable cations content were determined according to Jackson (1965) (ammonium-acetate pH7 method).Effective cation exchange capacity (ECEC) was the total amount of exchangeable bases (Ca, Mg, K, Na) and the exchangeable acidity (Al and H).Exchangeable acidity was measured by titration with NaOH after H + and Al 3+ ions displacement by a 1 N KCl solution.Al 3+ ion was determined separately by fluoride complexation.Total carbon and total nitrogen content was determined by dry combustion.The clay fraction mineralogical composition was determined at 20°C by X-ray diffractometry after the clays saturation by K + and Mg 2+ ions.
Total elemental analysis was carried out for the 10 soil samples according to Chao and Sanzolone (1992): Soil samples were crushed and powdered (particles diameter < 100 μm).A crushed and powdered sample of 0.1 g was mixed with 1.6 g of lithium metaborate and 0.4 g of lithium tetraborate poured into a graphite crucible.This mixture was then melted at 1000°C for 5 min in a kiln.After this time, the melt obtained was removed from kiln for cooling.The cooled melt was entirely dissolved in 100 ml of 2 M HNO 3 under magnetic agitation at 100°C.The analyses were then performed on the resulting solution.The total  al., 1971).

Weathering indicators calculation
In addition to indicators based on ratios and relationships resulting from different fractions (free, amorphous and total) of the elements (Al, Fe, Mg, Mn, Si), other types of indicators like the TRB, the mineral reserves (MR) and the WIP are also used.TRB was calculated as the sum of the total contents of alkaline and alkaline earth cations Ca+Mg+K+Na (cmolc kg -1 ) (Herbillon, 1986;Delvaux et al., 1989).Mineral reserves in the soil (MR) referred to as "weatherable minerals" (Udo de Haes et al., 2012) were calculated as the sum of the differences obtained between the total quantity of each base and the equivalent exchangeable quantity (Titeux, personal communication, 2012) according to the following equation: Where i corresponds to each of the 4 cations (Ca, Mg, K and Na), X it is the total quantity and X ie is the exchangeable quantity of cation i.The WIP is calculated from the expression (Delvaux et al., 1989): The higher the value of this index, the lower the soil weathering degree.So it would be judicious to consider this index as a "non-weathering index".

Some soil chemical characteristics values
Different chemical characteristics such as pH (water and KCl), exchangeable cations content, the sum of bases (S), CEC pH7, ECEC, base saturation rate [(BS=S/CEC) x100], total carbon (C tot ), total nitrogen (N tot ), available P and C:N ratio are presented in Table 2.The analysis of this table shows that soil pH values of the 5 sites were quite different.Mulungu soil was slightly acidic in the two considered layers.In Walungu and Kirundo, soil pH was moderately acidic with values ranging from 5.6 to 5.8; Gitega soil was very strongly acidic with pH values of 4.2 in 0-20 cm layer and 4.4 in the underlying layer (20-40 cm).Cibitoke soil was moderately acidic in the surface (0-20 cm) and slightly acid in the 20-40 cm layer.In all soils and for both sampled depths, pH water was equal or slightly higher than pH KCl and ΔpH values (in modulus values) range from 0 to 0.1 units, which would mean that soils variables charges were dominant in those soils.Indeed, dominance of a soil by variable-charges colloids can generally be assumed when ΔpH is a small negative (less than -0.5), zero or positive value (Rayment and Lyons, 2011; Uehara and Gillman, 1981).Globally, the average values of CEC were very high in soil of Mulungu while the soils of Kirundo and Walungu had a high CEC.CEC was low in Cibitoke and very low in Gitega (Tessens and Gourdin, 1993;Thiagalingam, 2000;Hazelton and Murphy, 2007).The trend was similar for ECEC.The sum of the bases (S) was close to the ECEC at all sites except Gitega, reflecting the high exchangeable acidity at this latter site (Table 2).
With the exception of Gitega where the base saturation rate was very low (< 20%), the average base saturation rate (BS) for both depths in the other 4 sites was greater than 60%, indicating a high (Mulungu, Walungu and Kirundo) to very high (Cibitoke) base saturation status (Table 2).The very low BS at Gitega site appears to be normal since the exchange complex is dominated by H and Al (Figure 1).
The cations contribution to charges (Figure 1) showed that Ca was the dominant cation in Mulungu, Walungu and Kirundo, followed by Mg and finally by K for both depths.In Mulungu the three cations formed respectively 75, 17 and 7% of exchange complex in the surface layer and 75, 20 and 4.7% in the underlying layer.In Kirundo, the respective contribution of the above mentioned cations was 74, 25 and 0.6% in the superficial layer (0-20 cm) and of 75, 24 and 0.5% in 20-40cm soil layer.In Walungu, the saturation percentage was of 67, 31 and 1.4% in surface and 65, 33 and 1% in the 20-40cm layer respectively for the Ca, Mg, and K.In Gitega, the trend was very different from what is observed in previous sites with 57% of exchange sites occupied by Al+H in the surface of which 33% with Al.The latter was taking up approximately 37% in 20-40 cm soil layer of which 8% by Al.In Cibitoke, Ca, Mg, and K respectively contributed to 54, 41 and 4 % in the superficial layer and to 47, 49 and 3% in 20-40 cm soil layer.
In the five soils, it was noticed that total carbon content was relatively higher in the surface layer (0-20 cm) than in the following layer (20-40 cm) (Table 2).Total carbon contents were the highest in Mulungu and Kirundo, and the lowest in Cibitoke.The same trend was observed with regard to total N content.Available P was very high in Mulungu and low in the four remaining sites.One should note the striking low P content in Walungu.C: N ratio was between 10 and 12 for all sites except in Gitega and Kirundo where it was fluctuating around 14. A ratio of 10 is an indicator of good biological activity.The highest ratio observed in Gitega and Kirundo indicates a relative slower degradation of organic matter and therefore a relative lower microbiological activity.

Particle size and mineralogical composition
Soils classification according to their particle size distribution is shown in Figure 2. The soil types fit in to three textural classes: Clay soils (Mulungu, Walungu and Kirundo), sandy clay loam soils (Gitega) and clay loam soil (Cibitoke).The mineralogical composition obtained from the X-ray diffraction (XRD) is presented in  descending order of magnitude that is to say from the most abundant clay to the least abundant (Table 3).Mulungu soil was dominated by halloysite while Walungu, Gitega and Kirundo soils were dominated by kaolinite in the clay fraction.Cibitoke soil was dominated by illite.
Because of the confusion of kaolinite lines and halloysite lines in an XRD, CEC kg -1 clay values (CEC clay ) allowed realizing the dominance of halloysite on kaolinite especially in Mulungu soil.The CEC clay value was determined taking into account the organic matter (OM) The study by Delstanche (2011) in AGLR showed that CEC OM values for this region do not deviate so much from the 200 cmolc kg -1 OM average value.Indeed, this author found that CEC OM values ranged between 180 and 236 cmolc kg -1 OM.Therefore, the CEC clay was worked out by the following equation: (3) A being clay percentage and OM, organic matter rate determined by multiplying C org by 1.728.On this starting point, CEC clay values were of 67 and 47 cmolc kg -1 in Mulungu respectively in the topsoil (0-20 cm) and underlying soil layer (20-40 cm) and thus by far higher to the value of kaolinite.This seems to be a sign of halloysite dominance in comparison to the two other kinds of clay minerals namely kaolinite and illite (Table 3) given that the latter is not very represented.

Soils weathering indicators
Different DCB and oxalate extractible Fe, Al, Si, Mn and Mg contents as well as oxalate extractible Al and Fe ratios are shown in Table 3.Total analysis results, TRB and WIP of the different soil are shown in Table 5.From Table 4, one can note that in all the five sites, the different Fe extractible fractions occurred in the following trend: Fe o < Fe d < Fe t .It was also noticeable that the total iron amount greatly varied according to sites in the following descending order: Walungu, Mulungu, Kirundo, Gitega and Cibitoke.The same trend as in the case of iron was observed in the case of Al with the trend Al o < Al d < Al t except at the Mulungu site.In the latter site, it was noted that Al o and Si o were superior to Al d and Si d respectively (Table 4) and this seems to indicate a particular characteristic of the soil type for DCB and oxalate extractible Al and Si.The Fe o /Fe d ratio which measures the reactivity of the sesquioxides was relatively higher in the Mulungu and Cibitoke and lower in the other three sites; the lowest values being observed in Gitega.Conversely Fe d /Fe t ratio that indicates the soil weathering extent was higher in Gitega site with an opposite trend to that observed in case of Fe o /Fe d ratio.This implies that the two parameters are negatively correlated.The relationship between the two ratios was relatively stronger in the top layer (0-20 cm) than for the 20-40 cm layer with R 2 values of 0.56 and 0.37, respectively (Figure 3).In other words, the higher the Fe o /Fe d ratio, the higher the amount of amorphous iron.Huang et al. (1977) reported that the amount of Fe o as well as Al o amount was related to OM content but not to soil acidity or clay content.
Feo was positively correlated with organic carbon rate.The correlation was relatively stronger in surface layer (0-20 cm) with R 2 = 0.51 than it was in 20-40 cm soil layer where R 2 = 0.26.The positive correlation between the total organic carbon content and Fe o shows the role of iron oxides in the soil organic matter stabilization (Wagai and Mayer, 2007;Wissing et al., 2013).
From Table 5, it was noticeable that TRB, MR and WIP are the highest in Cibitoke and the lowest in Gitega.Walungu site was characterized In bold and italic, values at detection limit of the measuring device.
by TRB and MR values greater than those in Kirundo, but both soil types have identical WIP values of 2.0 if the average of the two depths was considered.However TRB, MR and WIP values related to the different soil layers, enable to separate clearly the two soil types with relatively higher values in Walungu compared with Kirundo in the 0-20 cm soil depth.In the 20-40 cm soil layer the same trend was maintained for both sites with respect to TRB and the MR but a reversed trend was noted for the WIP.Cibitoke site markedly differs from the others by its higher total potassium, magnesium, silicon and sodium contents.Thus, it could be noted that in four out of five soil types namely Mulungu, Walungu, Gitega and Kirundo, total Ca and total Mg were by far two most dominant cations of TRB and MR.This was not the case in Cibitoke where total Mg and total K are rather the two most dominant cations.However the relative importance of the two dominant cations varies from one soil type to another.Thus in descending order, total Ca and total Mg were the most dominant in Mulungu and Kirundo soils.Total Mg and total Ca were the most represented cations of TRB and MR in Walungu and Gitega soils.In Cibitoke, total Mg followed total K and total Na were the cations taking over the TRB and MR.Thus, only based on this criterion, the five soil types boil down to three groups of soils (Mulungu and Kirundo, Walungu and Gitega, and Cibitoke) but the combination with other criteria such as the WIP indicated rather another trend and a clear difference between the soil types as it may be noticed in Figure 4.

Soils agronomic potential
The agronomic potential of different soils was evaluated by comparing the performance level in terms of observed banana yield and different soil chemical characteristics including the computed weathering indices.Total banana yield according to different sites in the control plots is shown in Table 6.
Banana yield per cycle were different according to sites (Table 6) in the following descending performance sequence: Cibitoke ≥ Mulungu > Kirundo > Walungu > Gitega.The relationship between the different soil quality indicators above mentioned (soil textural composition, pH, total nitrogen, total carbon, available P, bases, sum of bases, CEC, Fe o /Fe d and Fe d /Fe t ratios , TRB, MR and WIP) and bananas yield potential in 5 soil types is highlighted in the correlations matrix of Table 7.
The analysis shows that except the pH, exchangeable K and silt content, other physicochemical properties such total nitrogen and carbon, available P, exchangeable bases (Ca, Mg), S and CEC which are measures commonly used to account for the soils fertility status were poorly correlated to observed banana yield in the field compared with the correlation coefficients between total cations (total Ca, total Mg), MR and TRB.The best explanatory factors of the     observed performance banana yield levels were the MR with r = 0.97 (p = 0.007), nonexchangeable Mg with r = 0.93 (p = 0.03), the total Mg with r = 0.92, the TRB with r = 0.95 (p = 0.01) and the silt content with r = 0.87 (p = 0.05).The observed yields were also correlated with Fe o /Fe d ratio in surface with r = -0.95(p = 0.02).
In the underlying layer (20-40 cm), these were also the same parameters namely non-exchangeable Mg, total Mg, MR, TRB which were best correlated with banana yields compared with pH, total nitrogen and carbon, the bases, S as well as the CEC.It is also important to note that R values in the superficial layer (0-20 cm) are clearly greater than those of the underlying layer in the case of Fe o /Fe d ratio (Table 7).
Figure 5 illustrates the relationships between some of soil alteration indicators (silt content, TRB, RM, Fe d /Fe t and WIP) and observed banana yield in 5 soil types.

Soils physico-chemical properties and weathering extent trends according to sites
For the soils as a whole, CEC values are higher in  2), this may be related to the total carbon content and thus to organic carbon content which are similar given that soil pH values are < 7. Indeed, it is well known that soil CEC depends a lot on organic matter quantity as several studies on different soils types (Oxisols, Acrisols, ferric Lixisols) have pointed out (Oorts et al., 2003;Fritsch et al., 2007;Zhang et al., 2007;Mbonigaba et al., 2009;Delstanche, 2011) and to a less extent on soil clay content (Oorts et al., 2003;Delstanche, 2011) and the clay mineral nature.The influence of soil clay mineral nature is well illustrated when confronting Gitega and Cibitoke on the one hand, or Mulungu and Kirundo on the other hand.
It was noticeable that CEC values are higher in Mulungu and lower in Gitega.Soil in Mulungu contains halloysite, as a dominant clay mineral while kaolinite was clearly the dominant clay mineral in Kirundo with less illite, vermiculite being very little represented (Table 3).Therefore, high CEC values in Mulungu could be explained by many parameters among which the high OM content, the halloysite and illite presence as well as the presence of allophanic compounds.Boyer (1978) pointed out that halloysite, illite and allophanes are characterized by a high CEC; in the range of 20 to 30 cmolc kg -1 clay for illite and 20 cmolc kg -1 clay for halloysite.In Gitega, the low values of CEC can be explained not only by the low organic matter content, but especially by the texture and the mineralogical composition.Observed values in Gitega are consistent with those found in the soils where the clay fraction is dominated by kaolinite.Indeed, CEC in this site is about 11 and 5 cmolc kg -1 clay respectively in the topsoil (0-20 cm) and in the underlying layer (20-40 cm).This value of CEC (<16 cmolc kg -1 clay) is characteristic of kaolinite dominance in Ferralsols.Low values of CEC in Gitega bases to TRB may be explained by the high soil acidity which is evidenced by the fact that the bulk of the CEC is engaged in the exchangeable acidity (~57%, Figure 1) at the expense of the bases.There is also the fact that compared with other sites; the soil has very low clay content which further reduces the cations holding capacity.Therefore, the fact that CEC is low while WIP and TRB are still high like in Cibitoke may indicate that CEC use seems to be a good indicator of soil fertility when it is combined with the pH readings, but that others indicators like TRB could be better to judge unequivocally the soil fertility status especially in the sites that have not been affected by fertility management practices, because TRB is unlikely to be much affected by management.
The order of magnitude on the soils weathering stage (based on the TRB, MR and iron oxides ratios, CEC contribution to the TRB) seems to be also in agreement with the silt content values and the soils mineralogical composition (Table 3).The average silt content is higher in Cibitoke (35.2%) while soils silt content are 23.2, 18.9, 17.3 and 10.7% respectively in Mulungu, Kirundo, Walungu and Gitega.The silts mainly consist of primary minerals, they alter more quickly than the sands because of their high specific surface area due to their small size and more than clays which are often neoformed clay minerals and therefore more chemically stable.Therefore one understands why the TRB varies in the same direction than that of silts content given that the complex minerals that make up the silt and sand fractions are basically mineral reserves.As for the case of the silt content, mineralogical composition (Table 3) shows that in Cibitoke, the soil is the least weathered compared with others.Indeed, the mineral fraction is dominated by illite, kaolinite and vermiculite: illite and vermiculite being characteristic of young soils (Brady and Weil, 2002).According to Weaver (1989), illite, vermiculite and interstratified-smectite characterize soils at an intermediate weathering stage while the presence of kaolinite, gibbsite, goethite, hematite and anatase (TiO 2 ) depict a very advanced weathering stage.As such, Gitega and Walungu soils which contain in their mineralogical fraction, gibbsite and high TiO 2 levels (Walungu) in addition to kaolinite are more weathered than others.Gibbsite is appearing indeed in highly weathered soils (van Wambeke, 2002), in addition to iron and aluminum oxides as well as an increase in TiO 2 (Setterholm and Morey, 1995).
The assessment of the total or reserve base cations (Ca, Mg, K and Na) contribution to the TRB or MR indicates that Mg is the main contributor than Ca in Cibitoke, Gitega and Walungu soils.This case in point constitutes another indicator of soils weathering stage and the nature of the soils' parental material.Firstly, the fact that the total Mg is greater than total Ca in Walungu (Mg>Ca>K>Na) and Cibitoke with the sequence Mg>K>Na>Ca (Table 5) would confirm the basic rocks types from which these soils derived.Indeed, Godefroy and Dormoy (1990) indicate that in most soils derived from basic rocks types, Mg is the dominant element of the reserves.According to Jaffré and Veillon (1991), in all soils on ultramafic rocks, calcium and potassium contents are at low levels while magnesium contents are relatively high in particular on alluvium soil, where a strong imbalance Ca/Mg in the exchange complex is found.On the other hand, Boyer (1978) points out that often, the soils having undergone a long pedological process have relatively low reserves more in Ca than in Mg.From this point of view, the soil in Gitega (Mg>Ca>K>Na) (Table 5) would hold the superiority in Mg reserves compared with Ca reserves on account of the highly advanced weathering degree.In the opinion of the same author, Mg can also be dominant in illite and montmorillonite rich soils, and this is the case in Cibitoke (Table 3).
With regard to the different soil properties including TRB, MR, the WI of Parker, or the amount of active iron (Fe d /Fe total ), types of dominant cations of the TRB or the MR, particle size and mineralogical composition, all these parameters converge to indicate that the ascending order of soils alteration degree is Cibitoke (Fluvisol)-Mulungu (Nitisol)-Walungu (probably another type soil and not a Ferralsol)-Kirundo (rhodic Nitisol)-Gitega (typical Ferralsol).The Ferralsol of Gitega having reached an ultimate stage of weathering with regard to the TRB values.

EAHB yield potential as a snapshot of the soil inherent properties and weathering extent
The more TRB or Parker weathering index values were higher or even those of Fe d /Fe t ratio lower, better was banana yield.Hence Cibitoke site with higher values of these indices (TRB and Parker weathering index) is characterized by relatively good banana yield (in absolute terms) followed respectively by Mulungu, Kirundo, Walungu and Gitega (Table 6).The yields taken as a whole were strongly correlated with the MR, TRB, soil silt content, no exchangeable Mg, total Mg, Fe o /Fe d and Fe d /Fe t ratios, pH water and WIP (Figure 5) and poorly correlated with the common routine analysis parameters such as C tot (C org ), N tot , P, CEC (Table 7) although the correlations coefficients are not always significant.The fact that banana yield is better correlated with the TRB or MR than the CEC or the sum of exchangeable bases (S) would be due to the fact that low values of CEC (or even those of S) are observed in less altered soils (with high primary minerals content) as well as in highly weathered soils.Conversely, TRB, MR, non-exchangeable cations and total fraction of cations, the silt content (Yakubu and Ojanuga, 2013) and banana yield values are higher in young soils than in very old soils and this explains the better correlation between those parameters.At the same time, this would indicate that the prediction of a soil potential production based only the routine analysis parameter, that is, the CEC or the sum of the bases is distorted and the distortion appears to be reduced when the results of total analysis are considered.Subsequently, one is tempted to say that the measurements in routine analysis (e.g.exchangeable cations, CEC) do not seem to be fair indicators to reflect the potential level of chemical soil fertility and therefore explain perceived banana yields levels in field.On the other hand, the fact that banana yield is much lower in Walungu than the yield observed in Kirundo (Table 6) while values of TRB or MR are relatively high in Walungu (Table 5) points out that the plants performance is a result of a lot of complex interactions of numerous factors rather than simple indicators.
For instance, the very low available P compared with the critical value for banana of 15 mg kg -1 (Okalebo et al., 1993) at Walungu combined with the high level of Mn in the topsoil (0-20 cm) or even the toxicity level observed in the 30-50 cm soil layer (Muliele, 2014) appears to be the most limiting factors hampering good banana yield in the site.
The relationship between active iron and banana performance is negative and more pronounced in the 20-40 cm layer.The negative influence of iron oxides, more marked in the underlying layer in comparison to the surface layer (Table 7) would be due to the influence of organic matter.The higher is the organic matter, the less is active iron (Fe d ) and more amorphous iron (Fe o ).As there is more organic matter in surface (0-20 cm) than in the 20-40 cm layer, there is theoretically more active iron in the underlying layer (20-40 cm) which has a negative influence on fertility because of the anionic exchangeable capacities (AEC) exhibition associated with the surface of kaolinite, iron and aluminum oxides as well as amorphous materials (Pansu and Gautheyrou, 2006;Brady and Weil, 2014).This means that iron oxides as well as aluminum oxides become positively charged and attract, retain, and supply negatively charged anions, such as sulfate, phosphate, nitrate, and chloride (Brady and Weil, 2014) but in reverse, soils with high AEC experience leaching of positively charged nutrients, such as calcium, magnesium and potassium.Moreover, phosphorus as an easily exchangeable anion may be limited, because it tends to form strong bonds with the oxides (Uehara and Gillman, 1981;Wandruszka, 2006;Brady and Weil, 2014) and therefore not readily available for the plant uptake (Brady and Weil, 2014).The presence of high quantity iron oxides being a sign of weathering, it is easy understand therefore that the relationship between the active iron and the yield is negative and more pronounced in the 20-40 cm layer where there is less organic matter than in the surface layer.
With respect to certain common routine analysis parameters (pH Water , Ct ot , N tot , available P and exchangeable cations) and if all 5 sites are taken as a whole and correlation coefficients magnitude was considered, the yield was respectively correlated with exchangeable K, pH, P, Mg, Ca and finally total nitrogen.The strong correlation between yield and exchangeable K seems to be consistent with common observations.Sure enough, many studies (Turner and Barkus, 1982;Smithson et al., 2004;Moreira et al., 2009;dos Santos et al., 2009;Nyombi et al., 2010;Galán Saúco et al., 2014), converge to indicate that with regard to cations, K is the most limiting factor and therefore the most important, followed by Mg.Even if the correlation coefficient are not very strong and significant, a positive correlation between available P and the yield was expected since, with the exception of Mulungu, the soil content in this variable is far away below the critical value for banana (Okalebo et al., 1993;Godefroy et al., 1991) and hence may overall constitute a limiting factor.Likewise, the absence of correlation between banana yield and C tot (C org ) may be due to the fact this parameter is in sufficient concentration (Okalebo et al., 1993) in all sites but Gitega and therefore globally not a yield limiting factor.The strong correlation between yield and the soils pH while pH values in all sites except Gitega that fluctuate around 6.0 are much greater than the critical value of 5.2 suggested by Okalebo et al. (1993) may indicate that this parameter is still limiting and that the optimum pH value for EAHB banana production would be above the lower limit of the 5.8 to 6.5 optimum pH values range indicated by Galán Saúco et al. (2014).
Exchangeable Ca and Mg values (Table 2) illustrate that both cations are not limiting.Indeed, in all sites except in Gitega, Ca and Mg values seem high and therefore less limiting.In addition, the soil content in these two elements seems be in agreement with soil pH values that are all superior 5.5; Ca deficiencies occur in soils with low CEC and pH≤ to 5.5 (Landon, 1991).So, the soil in Gitega is the only one where Ca deficiencies are probable.In a study on banana systems in Kibungo (Rwanda) (dominance of Nitisols), Lassoudière (1989), Godefroy et al. (1991) indicate that critical values are respectively of 6, 2.5, 1.5 cmolc kg -1 soil respectively for exchangeable Ca, Mg K.According to Landon (1991), the deficiency threshold of exchangeable Mg in tropical regions is fixed at 0.5cmolc kg-1.On this basis, only the ferralsol (in Gitega) is deficient in exchangeable Mg (Lassoudière, 1989;Godefroy et al., 1991;Landon, 1991) and in exchangeable Ca (Lassoudière, 1989;Godefroy et al., 1991).
The nil banana yields in Gitega can be explained in part by the poor physico-chemical properties of the soil.Low CEC, lower rate of base saturation (BS), the low level of organic carbon, nitrogen and available P, highly acidic pH (Table 2) and higher aluminum saturation rate (33%, in 0-20cm layer) are soil fertility indicators far from conducive conditions for good banana production.A suitable land have a CEC > 16 cmolc kg -1 clay, a BS > 35% (Sys et al., 1993;Delvaux, 1995) and a pH Water range of 5.8 to 6.5 or pH KCl in the range of 5.0 to 5.8 for optimum banana Syldie 1375 production (Galán Saúco et al., 2014).In addition, with respect to the soil pH of the soil (around 4), textural composition (much more sand, Figure 2) and the low C org (SOM) content of the soil in Gitega, it seems coherent to have low levels of exchangeable Ca (Table 2) and in addition to aluminum toxicity expectation.Indeed, Fenton and Conyers (2002) reported that soil Ca content less than 0.5 cmolc kg -1 ) was among others usually associated with sandy soils that are low in SOM and with a pH < 4 and that in addition, such soil properties will have a greater consequence on plants growth than the Ca deficiency, in most situations.Many authors (Boyer, 1976;Delhaize and Ryan, 1995) indicated that exchangeable aluminum becomes toxic from pH water < 5 or pH water ≤ to 5.5 (Meriño-Gergichevich et al., 2010), it is therefore easy to understand that this factor may constitute a yield limitation in Gitega because of its toxicity, but is not a problem in the other sites (Mulungu, Walungu, Kirundo and Cibitoke) whose pH values lie around 6 (Table 2).In fact, the content of exchangeable Al in this site of 33% is greater than the value of 30% reported by Gauggel et al. (2003).These authors reported that production experiences indicate that in soils with a pH range from strongly to extremely acid (which the case of Gitega), exchangeable aluminum concentrations are greater than 30%, and together with high soil manganese (Mn) concentrations, yields are reduced, usually to less than 2,000 boxes ha -1 year -1 (~30 tha -1 year -1 ).In optimal conditions, for a potential banana yield of 70 to 93 tha -1 year -1 in the same area (Latin America) (López and Espinosa, 1998;Chia and Huggins, 2003), this corresponds to a yield loss of 50 to 63%.The fact that banana yield in Gitega is zero meaning a loss about a 100% while in Latin America yield loss range from 50 to 63% in similar conditions, would indicate a less tolerance to extremely acidic soils of EAHB than the varieties grown in Latin America.Although manganese toxicity was also expected in this site, this thesis does not seems be plausible.Indeed, there is naturally little manganese in the soil of which the total Mn content is approximately 160 ppm, a quantity which is much less (20-fold less) to 3275 ppm of total Mn indicated by Fouré and Marchal (1983) to observe a manganese toxicity in banana.These soil quality indicators (very low pH, exchangeable aluminum, very low base saturation, very low CEC and TRB or MR) in the site of Gitega could explain the poor banana performance subject to other many harmful phenomena to the plant at low soil pH such as the decrease in nitrification, the phosphorus deficiency and the availability of some heavy metals (Landon, 1991).The fact that exchangeable K is relatively well correlated with banana yield (Table 7) is possibly also an expression of a cationic imbalance.For all the sites, the values of K /(Ca + Mg) ratio are less than 10%, critical value proposed by some authors who worked in this zone (Lassoudière, 1989;Godefroy et al., 1991).Except the site at Gitega, the banana yield data analysis, site by site indicates that very low Ca/K or Mg/K ratios or high K /(Ca+Mg) ratios correspond to better banana yields (case in Cibitoke and Mulungu), which reflect the prominent role of cation balance and K in banana nutrition and yield.However, in Gitega, Ca/K or Mg/K ratios are low and even the K/ (Ca + Mg) ratio is better than other sites but Gitega is the place where banana performance is the worst.This situation is to be linked to the very low soil fertility status in the site.According to Boyer (1978), balanced ratios of Ca, Mg and K are not of great significance for the plants nutrition in a soil with a BS < 10% as it is the case at Gitega; the very acidic pH and the exchangeable aluminum hamper the plants growth of in such cases.Soil acidity as the main handicap in Gitega seems more likely since this site presents good physical properties conducive to banana cultivation.Indeed, sandy clay loam soils, as it is the case of soil in Gitega (Figure 2), are the best for bananas cultivation (Robinson and Galán Saúco, 2010).Moreover, the average on the two depths for textural composition is 29.6% clay, 10.7% silt and 59.7% sand (Figure 2).These proportions do not differ from the optimum soil texture values reported by these authors (both for bananas and plantains) of 30% clay, 10% silt and 60% sand, because of a better aeration, a good infiltration coupled with a good soil drainage.

Conclusion
The production potential reflects the degree of soil alteration and generally speaking, the soil fertility status.In this study, it was found that banana yield levels were associated to the soil weathering stage and the other studied soil parameters.Banana yield levels were better and positively correlated with the total reserves in bases (TRB), the mineral reserves (MR), the Fe o /Fe d ratio and the silt content than the other routine soil analysis' parameters such as pH, total nitrogen, available P, exchangeable cations, the sum of the bases and the CEC.The more was the alteration degree of the soils, the less was the yield potential.Furthermore, soil weathering extent was better apprehended and associated to the level of production if the parameters from the soil routine analyses were completed by the total analysis parameters and iron oxides consideration.This was particularly patent for the most weathered soil of Gitega.Moreover, since the plants nutrition in low inputs agriculture system to is done at the expense of the primary minerals weathering which can be estimated by the TRB or mineral reserves, it appears appropriate to estimate the potential of soil chemical fertility and the type of interventions (fertilization and/or amendment) based on this parameter in addition to N and P levels assessment.Therefore, one might as well think about the plants nutrition balance by N, P, K; one might as well think about also, how to improve the soils mineral reserves including the reserves of Ca and Mg as they appear at low levels in weathered soils which make up the majority in the AGLR.

Figure 1 .
Figure 1.Different cations contribution to ECEC (sum of exchangeable cations and the exchangeable acidity).

Figure 2 .
Figure 2. Soils textural classes of the soils under study.

Figure 3 .
Figure 3. Relationship between reports Fe o /Fe d and Fe d /Fe t (established from 5 types soil).

Figure 4 .
Figure 4. Sites relative position on the basis of TRB and WIP mean values.The vertical dashed line represents the upper limit of TRB = 40 cmolc kg -1 soil of highly weathered soils.

Figure 5 .
Figure 5. Relationship between observed banana yield and (a) non-exchangeable and total Mg (b) mineral reserve and TRB, (c) the quantity active iron in the soil, (d) the quantity amorphous iron in the soil.

Table 1 .
Some general characteristics of the five experimental sites.

Table 2 .
Selected chemical characteristics of 5 soils types of the RGL (Burundi and DR Congo).

Table 4 .
DCB and oxalate extractible elements contents (Al, Fe, Mg, Mn, Si), total iron content and different iron oxides ratios in the fine earth.

Table 5 .
Elemental total content, TRB and WIP of the five soil types in the fine earth (<2 mm).

Table 6 .
Total banana yield observed in control plots.

Table 7 .
Pearson correlation coefficients between observed banana yield and: (a) exchangeable cations, the sum of exchangeable bases (S), non-exchangeable bases (MR) and the sum of total cations (TRB); (b) other weathering soil indicators (data from 5).The arrow indicates that correlations are established between the yield and the features of part (a) or (b) of the table Note that ∑No exchangeable bases=MR and ∑Total bases=TRB The *, ** and *** indicate the level of significance corresponding respectively to 001<p-value<005; 0001 < p-value < 001 and p-value<0001 The absence of asterisk indicates that the correlation is not significant (p-value>005).