1. Introduction
Soil physical and hydraulic properties can change drastically over space [
1] and time [
2] and their evaluation is essential for a rational soil management and, therefore, for increasing crop yields performance [
3]. Moreover, they dynamically affect water balance components and crop yields by relating soil hydraulic functioning to climate patterns and crop water requirements [
4,
5].
Soil properties, such as saturated hydraulic conductivity, total organic carbon content, structure stability index, dry bulk density, as well as capacitive indicators obtained from the soil water retention curve (i.e., plant available water capacity, relative field capacity, air capacity) were widely and successfully applied to investigate soil management effects on soil physical and hydraulic properties [
3,
6,
7,
8]. Keller et al. [
3], for example, have investigated the relationships between crop yield and soil structure in three Swedish fields, applying the simplified falling head (SFH) technique [
9] to evaluate whether field-saturated hydraulic conductivity (
Ks) could be used as a simple and quickly measurable indicator of crop yield. The main findings of their investigation showed that
Ks may be a good indicator of low yielding zones, and degradation of soil structure has been indicated as the main reason for low yield. However, in other studies, spatial and temporal variation of soil water holding capacity was suggested to be a factor partly responsible for crop yield variation, regardless of the amount and timing of the rain contributions [
10,
11]. Consequently, investigations addressed at evaluating new experimental procedures for soil hydraulic characterization, and at establishing the usefulness and sensitivity of soil properties as predictors for crop yield, are needed. This is a current issue in Mediterranean agro-environments, where water resources are scarce and need to be optimally managed [
12].
Soil hydraulic properties assessment, i.e., water retention curve and hydraulic conductivity function (
θ(
h) and
K(
h) relationships, respectively), is expensive and time consuming, since standard methods need specific skills and their spatialization, i.e., prediction on a distributed large scale, which can be burdensome or wholly inapplicable [
13].
Several quick methods are available to obtain
θ(
h) and
K(
h) (or
K(
θ)) relationships. Pedotransfer functions (PTFs), for example, allow for estimating soil hydraulic properties starting with basic soil variables such as soil texture, bulk density or organic matter or hydraulic conductivity [
14]. However, PTFs are not able to accurately quantify the effects of different agronomic options for soil management, unless a site-specific calibration is performed [
15]. On the other hand, the most widely applied laboratory methods as hanging water column apparatus [
16], pressure cells [
17] or evaporation method [
18], may require up to a week (or more) to obtain a single soil water retention curve [
19,
20]. As a consequence, although reliable and accurate, they are not easily applicable for large-scale research and simplified techniques should be chosen for these purposes [
21,
22].
The BEST-procedure by Lassabatère et al. [
23] allows for estimating hydraulic functions repeatedly over space and time with substantially limited experimental burdens. Basically, the procedure requires three sets of experimental information: (i) cumulative infiltration by a simple infiltrometric experiment (i.e., ring infiltration test of Beerkan type), (ii) soil bulk density and volumetric soil water content at the time of experiment by sampling few (generally two) undisturbed soil cores, and (iii) soil particle-size distribution or, alternatively, clay, silt and sand percentages according to the USDA classification. Specifically, the procedure makes use of well-known analytical solutions for
θ(
h) and
K(
h) relationships and estimates (i) shape parameters, which are texture dependent, from particle-size analysis by physical-empirical PTFs, while (ii) scale parameters, which are structure dependent, by a three-dimensional field infiltration experiment at zero pressure head [
23]. Therefore, BEST can be considered an adequate compromise between estimation accuracy and economic-experimental load. For example, some studies have applied BEST to establish the effects of droplet impact on soil sealing and crust formation [
24,
25], to carry out integrated soil physical quality assessment [
26,
27] or to identify the effects of tillage on some soil properties (i.e.,
Ks) under drip irrigation [
28]. A main advantage of BEST is that it can be adopted when a large number of hydraulic measurements must be obtained at the field or at irrigation district scale, for example, for precision agriculture purposes. For instance, Mubarak et al. [
29] assessed the temporal stability of both
Ks and spatial structure of hydraulic properties of a loamy soil. Specifically, they compared
Ks-BEST data with those estimated seventeen years earlier by applying the guelph permeameter method, under relatively comparable soil and agronomic conditions. Results showed that
Ks changed significantly, but observed discrepancies were not higher than a factor of three or four. This suggests that BEST can represent an easy, robust, and inexpensive way for characterizing soil hydraulic behavior and its spatial [
30] and temporal [
29] variability at the field scale. The availability of a large number of georeferenced hydraulic measurements would allow delineating homogeneous sub-areas within the crop field to be submitted to uniform agronomic management [
31,
32]. This can lead to an increase in agricultural resources optimization [
31]. In addition, dense spatial data can be used as auxiliary/covariate information in mixed effects models to improve the estimates of the target variables [
33] or to improve the estimation of treatment significance reducing the risk of misleading or erroneous inferences in analysis of variance [
34,
35]. In other words, the potential application advantages seem attractive, but BEST has not yet been tested for soil hydraulic properties spatialization and the actual reliability for the mentioned purposes must be proven.
The general objective of this study was to test the BEST-procedure for the spatialization of soil hydraulic properties. In particular, the spatial distribution of the measured-estimated by BEST variables (soil texture, bulk density, saturated hydraulic conductivity, plant available water capacity, relative field capacity, air capacity) and ancillary soil properties (total organic carbon content, structure stability index) was investigated in a typical wheat cropping system in Southern Italy. Cross-correlation analysis was applied to establish strength and the extent of the spatial relationships between selected soil properties and crop yields.
4. Discussion
Overall, the observed correlations showed physically plausible relationships among soil properties (
Figure 4). In particular, inverse relationships between
ACe,
TOC,
SSI and
BD were detected, as well as between
ACe,
Ks and
RFCe. As expected,
Ks was positively correlated with
ACe, as increasing values of soil aeration should match with increasing saturated hydraulic conductivity (
Ks); a reasonable positive relationship between
Cl + Si fraction and
Ks was also detected. Expected correlations have been verified between
SSI and
TOC (positive) or between
SSI and
Cl + Si (negative), since
TOC and
Cl + Si terms appear, respectively, in the numerator and in the denominator of Pieri’s equation [
6]. However, uncertain results were obtained for the significant correlations of
PAWCe, because indirect relationships between
PAWCe and
Cl + Si, as well as
BD, were detected. In fact, although the latter was quite verified in the literature for medium-high
BD values [
6,
57,
58], the former relationship does not seem entirely plausible because higher
PAWCe values should correspond to higher contents of fine particles of the soil. A similar explanation can also be given for the relationship (both general and spatial) between
PAWCe and crop yield.
Agronomic results of this investigation showed medium-high wheat yields on average (
Table 1). However, according with the hypothesis that good conditions of soil physical quality should correspond to good yields [
3,
59], findings showed non optimal soil conditions in terms of
TOC,
SSI, and
RFCe (
Figure 3), i.e., in 43% of cases, and some considerations are due, particularly for optimal
TOC ranges. For instance, the suggested lower critical
TOC limits for agricultural soils (i.e., about 20 g kg
−1) by Reynolds et al. [
6] were obtained for urban soils, namely for sustainable establishment of plants in constructed landscaping soils used in urban parks, playing fields, curbside plantings, etc. [
6,
60]. Despite this critical threshold being applied in the literature e.g., [
46,
61,
62], different limits should be considered for agricultural soils, and specifically for crops grown in Mediterranean environments. For example,
TOC mean values observed in this investigation (i.e., about 16 g kg
−1) were relatively high as compared to those reported in
Table 1 by Ventrella et al. [
63], for eight experimental farms and soils with different texture, located from north to south of Italy (5.2 ≤
TOC ≤ 15.3 g kg
−1). In addition, in other investigations where a variety of plant species was considered, optimal
TOC levels, as suggested in the literature (i.e., 30 <
TOC < 50 g kg
−1), were seldom reached. For instance, for approximately two-hundred soil cores collected in three areas and 18 spot sites of agricultural and forest Sicilian environments, Castellini and Iovino ([
14];
Table 2) reported
TOC values equal to 10.8 g kg
−1 on average (0.9–37 g kg
−1 as a measured interval). To provide more concrete examples,
Table 7 summarizes other results of some investigations carried out in the Mediterranean environments of Apulia, Sicily and Sardinia (south-central Italy). With reference to such investigations, listed
TOC values were equal, at most, to 16.7 g kg
−1 for conservation soil management of durum wheat, equal to 22.6 g kg
−1 for a citrus groove or slightly higher for a grassland, suggesting that it is not common to find considerably higher
TOC levels in Mediterranean agricultural environments. On the other hand,
TOC values close to the upper limit suggested by Reynolds et al. [
6], 50 g kg
−1, were only detected for a sandy loam soil under high maquis (holm oak), i.e., when organic matter accumulation and probably slow organic matter mineralization rate can lead to such high levels (
Table 7). As an example,
Figure 3 shows how the
TOC classification by Sequi and Nobili [
44], specifically referred to fine textured soils (clay, clay–loam, silty–clay and silty–clay–loam), is more suitable for the specific characteristics of Mediterranean agricultural soils. As a consequence, more accurate optimal or critical values should be provided for specific agro-environments, for example performing ad-hoc new research or by reviewing the available literature (e.g., through meta-analysis). Similar considerations can be made for
SSI, as it is obtained from
TOC and texture, and is not related directly to the soil structure, but is rather related to the soil resilience [
6]. On the contrary,
RFCe limits may be considered relatively more reliable as they were successfully verified in reference to other soil physical properties (i.e.,
BD,
TOC,
ACe, macroporosity) by Reynolds et al. [
64] for a clay loam soil, as well as in comparison with literature guidelines.
Among those investigated, only five soil physical properties were found spatially structured, i.e.,
Cl + Si,
BD,
TOC,
SSI and
PAWCe, but their relationships with grain yield did not appear to be always convincing (
Figure 5). In detail, in comparison to the yield map (i.e., lower or higher yielding zones), relatively coherent spatial distributions were identified in terms of fine soil texture components (
Cl + Si) and
BD; this suggests that, especially bulk density, can represent a reliable physical indicator to manage within-field sub-areas with different productivity. For this soil indicator, the map also shows that the sub-areas with relatively lower productivity correspond to those with very low
BD values (as specified by the critical lower value of 0.9 g cm
−3 of
Figure 3), suggesting that this limit seems quite realistic for the case under study. On the contrary, conflicting information between yield and
PAWCe maps were obtained (i.e., an inverse spatial structure), as already shown by the overall correlation. These results can be attributed to uncertainties of the
PAWCe estimates obtained by BEST which would require further deepening. As expected, a relatively similar spatial distribution was observed for
TOC and
SSI, but not consistent with that of crop yield. In other words, findings provided by variables directly measured, both for overall correlation and for spatial analysis, were more convincing than those derived by variables estimated by BEST, for which further investigations are probably needed to quantify the accuracy degree of estimated soil water retention curve.
To deepen the aforementioned results and corroborate the relationships between yield and considered soil properties, a stepwise analysis was carried out to identify the variables most affecting wheat yield among those directly quantified (
TOC,
BD, fine textural components) or derived from laboratory measurements (
SSI). Results of stepwise analysis showed that
BD was the only variable selected (
P = 0.0054), thus confirming the key role of soil porosity and compaction in affecting crop yield. Consequently, for investigated soil variables, only the
BD map seems utilisable for agronomic management, i.e., for precision agriculture applications, because several factors may have contributed to produce unexpected or uncertain results for the other variables. Among these factors, particular relevance can be attributed to: (i) the uncertainties of the
PAWCe estimates obtained by BEST; (ii) the relatively poor correspondence between observed physical–chemical fertility of the soil and agricultural yields in the specific conditions investigated; (iii) a possible spatial variability at a scale smaller than that experimentally measured (i.e., lower than about the mesh side). About the last statement, as
Ks showed no significant spatial structure, we could make a plausible conjecture suggesting that spatial variability of
Ks occurred at a smaller scale than that investigated [
67]. However, no significant relationship between
Ks and crop yield (or
BD) was detected. Overall,
Ks is reported to be a better soil structure indicator [
68] than
BD, as the latter does not provide any information on soil pore distribution, i.e., architecture, connectivity and tortuosity of soil porosity [
3]. In our case study, a very good spatial correspondence between crop yield and soil bulk density was detected, and the cross-correlation analysis also showed a significant positive relationship with the crop yield for the lower distance, of about 25 m. Although this result could be considered questionable from the perspective of precision agriculture application (i.e., for implementation on platforms with on-the-go soil sensors), it is quite significant as
BD measurements can be obtained more easily if compared to
Ks, and it can be easily related to penetrometric soil measurements.
As a final remark, it should be noticed that SSI showed promising characteristics to be used as a representative indicator to identify homogeneous within-field areas, because of its tight relationships with chemical (TOC) and hydrological (fine texture components, and in turn PAWCe and BD) indicators and its predisposition to spatialization, thus suggesting a possible significant correlation with yield under a denser spatial sampling scheme. If this were the case, SSI would candidate itself as a key indicator for precision agriculture applications.
5. Conclusions
In this investigation a set of eight physical and hydraulic soil properties directly measured or estimated by BEST was obtained and spatialized to investigate correlations and identify intervals corresponding to medium-high levels of wheat yields, in order to provide useful information for site-specific agronomic management.
According to the guidelines of literature, a soil physical quality evaluation highlighted that soil under study had optimal bulk density, plant available water capacity, air capacity and saturated hydraulic conductivity values, but that the total organic carbon content, structure stability index and relative field capacity suggested too low levels of organic carbon or excessive soil compaction. However, both a literature analysis for different types of Mediterranean vegetation cover and the correlation analysis (overall and spatial correlation) suggested that a review of the optimal or critical TOC values for typical crops of Mediterranean environments should be made, as TOC value around 20 g kg−1 is hardly achievable even under conservation agriculture systems, and literature values are probably not entirely realistic for Southern Italy’s cereal crops. In addition, capacitive indicators estimated from the soil water retention curve of BEST provided both expected and uncertain correlations, especially with regard to the inverse relationship between plant available water capacity and wheat yield. Therefore, for the soil–crop system studied, application of the simplified BEST-procedure did not return completely reliable results and further investigations are needed to quantify the accuracy and reliability of estimated soil water retention curve. These main results open up new research perspectives to improve our knowledge on this topic.
Among measured soil properties, BD showed a spatial distribution in agreement with that detected for crop yield, and the cross-correlation analysis also showed a significant positive relationship only for short lags. Finally, SSI showed promising characteristics suggesting a possible significant correlation with yield under a denser spatial sampling scheme, and a potentiality as a key indicator for precision agriculture applications.
Further research on this topic is needed for Mediterranean agro-environments, by deepening: (i) the reliability of available measurement methods for accurately estimating representative physical and hydraulic soil properties, and (ii) the temporal stability of observed spatial relationships between soil properties (soil bulk density or soil texture) and crop yields along a larger time interval.