Identification and predictability of soil quality factors and indicators from conventional soil and vegetation classifications

Generally, the physical, chemical and biological attributes of a soil combined with abiotic factors (e.g. climate and topography) drive pedogenesis. However, biological indicators of soil quality play no direct role in traditional soil classification and surveys. To support their inclusion in classification schemes, previous studies have shown that soil type is a key factor determining microbial community composition in arable soils. This suggests that soil type could be used as proxy for soil biological function and vice versa. In this study we assessed the relationship between soil biological indicators with either vegetation cover or soil type. A wide range of soil attributes were measured on soils from across the UK to investigate whether; (1) appropriate soil quality factors (SQFs) and indicators (SQIs) can be identified, (2) soil classification can predict SQIs; (3) which soil quality indicators were more effectively predicted by soil types, and (4) to what extent do soil types and/ or aggregate vegetation classes (AVCs) act as major regulators of SQIs. Factor analysis was used to group 20 soil attributes into six SQFs namely; Soil organic matter, Organic matter humification, Soluble nitrogen, Microbial biomass, Reduced nitrogen and Soil humification index. Of these, Soil organic matter was identified as the most important SQF in the discrimination of both soil types and AVCs. Among the measured soil attributes constituting the Soil organic matter factor were, microbial quotient and bulk density were the most important attributes for the discrimination of both individual soil types and AVCs. The Soil organic matter factor discriminated three soil type groupings and four aggregate vegetation class groupings. Only the Peat soil and Heath and bog AVC were distinctly discriminated from other groups. All other groups overlapped with one another, making it practically impossible to define reference values for each soil type or AVC. We conclude that conventionally classified soil types cannot predict the SQIs (or SQFs), but can be used in conjunction with the conventional soil classifications to characterise the soil types. The two-way ANOVA showed that the AVCs were a better regulator of the SQIs than the soil types and that they (AVCs) presented a significant effect on the soil type differences in the measured soil attributes.


60
The multiple roles and functions of soil have resulted in several broad 61 definitions of soil quality. One of the most widely adopted definitions for soil quality 62 (SQ) was proposed by a committee for the Soil Science Society of America (chaired 63 by Karlen) as: "the capacity of soil to function, within natural or managed ecosystem 64 boundaries, to sustain plant and animal productivity, maintain or enhance water and  Smart et al., 2003). 172

Aggregate vegetation class (AVC) +(abrev) Description
1. Crops and weeds (Craw) Weedy communities of cultivated and disturbed ground, including species-poor arable and horticultural crops.

Tall grass and herbs (Tgah)
Less intensively managed tall herbaceous vegetation typical of field edges, roadside verges, stream sides and hedge bottoms. 3. Fertile grassland (Frtg) Agriculturally improved or semi improved grassland. Often intensively managed agricultural swards with moderate to high abundance of perennial rye grass. 4. Infertile grassland (Infg) Less-productive, unimproved and often species rich grasslands in a wide range of wet to dry and acid to basic situations. 5. Lowland wooded (Lwlw) Vegetation dominated by shrubs and trees in neutral or basic situations, generally in lowland Britain. Includes many hedgerows.  Phosphorus was determined by the Olsen P method according to Emmett et al. (2008). 187 Total C and N were determined using UKAS accredited method SOP3102 on an 188 Elementar Vario-EL elemental analyser (Elementaranalysensysteme GmbH, Hanau, 189 Germany) according to Emmett et al. (2008, and2010). from these values.

207
The metabolic and microbial quotients were calculated indices. The metabolic 208 quotient or coefficient was calculated as the ratio between the CO 2 -C from basal 209 respiration and the microbial biomass-C (CO 2 -C resp -to-C mic ), expressed as µg CO 2 -C 210 mg -1 biomass-C h -1 . It is also known as the specific respiration rate (qCO 2 ) (Anderson and Domsch, 1993). The microbial quotient was calculated as the ratio between the 212 microbial biomass-C-to-total organic C (C mic -to-C org ).      Note: *Correlation is significant at P < 0.05 level, and ** at the P < 0.01 level; qMic, microbial quotient; qCO 2 , metabolic quotient; Mic C, 304 microbial carbon (mg C kg -1 ); Mic N, microbial nitrogen (mg C kg -1 ); Mic C:N, microbial C:N ratio; SR, soil respiration (mg C kg -1 h -1 ); SOC, 305 soil organic carbon (mg C kg -1 ); NO 3 -, nitrate (mg N l -1 ); NH 4 + , ammonium (mg N l -1 ); EC, (µS cm -1 ); Phenols, Soluble phenolics (mg l -1 ); Abs  The retained factors accounted for > 61% of the total variance in the measured attributes; see 315   Table 3.     The first factor explained 16.7 % (see Table 3) of the total variance. It was named soil    (Table 5).  (Table 6). The SOM factor had the highest factor scores (P < 0.001) in Heath and Bog.    Table 6).  The cross tabulation of AVCs versus soil types (Table S3) in discriminating the AVCs than the SQF (Table 6) 497

Soil quality indicators across soil types
Since qMic and bulk density were moderately correlated (r=0.46**), they may be 498 redundant as indicators to be used together. If only one attribute were to be used to monitor soil 499 quality in soil types and AVCs, qMic and BD respectively seems to offers the greatest potential 500 judging from their high weights on the respective prediction models. However the qMic may be a 'MUST be included' soil attribute in the minimum data set, due to its important role in several 502 soil functions, being a fraction of soil carbon. Soil C influences a wide range of soil functions 503 including bulk density, infiltration, pesticide buffering, aeration, aggregate formation, pH, buffer 504 capacity, cation-exchange properties, mineralization, and the activity of soil organisms (Larson 505 and Pierce, 1991; Seybold et al., 1997). However, since the measurement of bulk density is 506 reasonably easy to obtain, it is therefore reasonable to consider it together with SOC, microbial 507 and biomass C as minimum data set for assessing soil quality across average vegetation classes 508 in the study area.

537
The clusters from multivariate classification are "natural" groups, which uses the 538 "minimum-variance" solution; where a population is partitioned into cluster subsets by 539 minimizing the total within group variation while maximising between groups variance (Wishart,540 1968). The groups/cluster formed from the multivariate analysis need to have no significant 541 overall spread. The clusters therefore, should correspond to data modes (distinct modes).

542
However, most of our cluster modes defined by soil types were not always distinct. Most of them 543 were separated from each other by significant "noise" data, making it impossible to resolve all 544 the clusters. Thus, the definition of the reference values for each soil type or AVC was 545 ambiguous, since most soils types or AVC groups could not be differentiated (Fig. 2). Forming, only a few could be differentiating (Soil Survey Staff, 1999). Even when the soil quality 548 factors/indicators and attributes were used in combination, some groups/clusters could still not 549 be resolved. Therefore, the soil quality indicators and attributes identified in this study can only 550 be used to characterise soil types and AVCs groups rather than for prediction or classification.

551
From the discriminant plots and the dendrogram in Fig. 3 three groups can be defined in soil 552 types and four groups in the AVC. AVCs) as well as significant differences in the effect of soil type on the soil attributes between 577 the AVC (significant interaction of soil type × AVC; Table 7). The 'practical' significance of 578 each term from Partial Eta Square values indicates that AVCs (with a large Partial Eta Square = 579 0.42), were a better regulator of the SQIs than soil types (with a weak Partial Eta Square = 0.09).

580
The effect size for the interaction was equally relatively weak (Partial Eta Square = 0.16). The 581 conclusion of the significant (P < 0.01) interaction effect of soil type × AVC is that the soil type were not at all represented (See Table S3). This problem can contribute to the complexity and 591 accuracy in the interpretation of the interaction effect observed above.

594
The dominant SQFs/Is and attributes varied by both soil type and AVC. The SOM factor 595 was the most discriminating factor for both soil types and AVCs with microbial quotient and 596 bulk density as the most discriminating measured attributes. The discriminant analysis on the 597 important measured attributes comprising the SOM factor produced three fairly homogenous 598 groups for soil types and four groups for AVCs. It was however, impossible to define reference 599 values in the SQF/I or attributes for separate individual soil types or AVCs, as property ranges 600 greatly overlapped due to large between group variability (probably due to integrating large 601 spatial areas). Some of the differences observed in soil types with regard to soil attributes were in 602 part dependent on the AVCs differences. conditions. Therefore, the search for SQIs which can be predicted by soil types continues.

610
For future work it might be worthwhile to make special consideration for the climatic,