Conventional detection methodology is limiting our ability to understand the roles and functions of fine roots


  • Alain Pierret,

    Corresponding author
    1. INRA – Climat, Sol & Environnement, Domaine St Paul, Site Agroparc, 84914 Avignon cedex 9, France;
    2. Present address: IRD-IWMI-NAFRI, BP 06, Vientiane, Laos
    Search for more papers by this author
  • Christopher J. Moran,

    1. Centre for Water in the Minerals Industry, Sustainable Minerals Institute, University of Queensland, Brisbane 4072, Australia;
    Search for more papers by this author
  • Claude Doussan

    1. INRA – Climat, Sol & Environnement, Domaine St Paul, Site Agroparc, 84914 Avignon cedex 9, France;
    Search for more papers by this author

Author for correspondence: Alain Pierret Tel: +856 20 550 2680 Fax: +856 21 41 2993 Email:;


  • • We lack a thorough conceptual and functional understanding of fine roots. Studies that have focused on estimating the quantity of fine roots provide evidence that they dominate overall plant root length. We need a standard procedure to quantify root length/biomass that takes proper account of fine roots.
  • • Here we investigated the extent to which root length/biomass may be underestimated using conventional methodology, and examined the technical reasons that could explain such underestimation. Our discussion is based on original X-ray-based measurements and on a literature review spanning more than six decades.
  • • We present evidence that root-length recovery depends strongly on the observation scale/spatial resolution at which measurements are carried out; and that observation scales/resolutions adequate for fine root detection have an adverse impact on the processing times required to obtain precise estimates.
  • • We conclude that fine roots are the major component of root systems of most (if not all) annual and perennial plants. Hence plant root systems could be much longer, and probably include more biomass, than is widely accepted.


Over the past few years there has been a growing realization that our knowledge about roots, which has been applied globally to managing plant systems, has been based on false premises of simplicity (Wells & Eissenstat, 2001; King et al., 2002; Pregitzer, 2002; Pregitzer et al., 2002; Zobel, 2003; Hodge, 2004). In a discussion paper about the fine roots of trees Pregitzer (2002) stressed the fact that our traditional views of fine tree roots – our knowledge and understanding of their length and diameter, structural and functional diversity/complexity and life history/turnover – are probably deeply flawed. More recently, Zobel (2003) revisited and reaffirmed Pregitzer's original points based on examples which demonstrate that we still lack clear anatomical, dimensional, functional and physiological definitions of what a fine root is, not only for trees, but for all plant species.

It is now certain that in both annual and perennial plants, roots <1 mm in diameter form a structurally and functionally complex population which is the dominant component of the root system. While it is widely recognized that fine roots amount to most of the root length in many plant species, it is also acknowledged that this is often underestimated because of their small size and near transparency (Costa et al., 2001). We need to improve our understanding of fine roots to support better prediction and management of biogeochemical cycles from the scale of single plants to the globe (Jackson et al., 1997; McCully, 1999; Norby et al., 2004; Thomas et al., 2004). To do this, it is necessary to come to terms with measuring the quantity and functions of fine roots under a variety of conditions. It is striking that there is no firmly established procedure for the accurate measurement of fine root length and biomass (Vogt et al., 1998).

In trees, the importance of fine roots is well supported by field evidence. For example, King et al. (2002) showed that in loblolly pine, 96% of the mycorrhizal and 77% of the nonmycorrhizal root length occurred in roots with diameters ranging from 0.2–0.6 and 0.4–1 mm, respectively. In a study which demonstrated that overwinter survivorship of apple roots is positively correlated to root diameter, Wells & Eissenstat (2001) showed that the majority of roots (>64.5%) were 0.1–0.3 mm in diameter. Pregitzer et al. (2002) confirmed for eight north American tree species that lateral roots <0.5 mm in diameter appear to account for >75% of the root length. Norby et al. (2004) found that under both ambient and CO2-enriched conditions, ≈80% of the root length of 10–15 yr old sweetgums (Liquidambar styraciflua L.) was in roots <0.5 mm in diameter.

Quantified evidence that fine roots are the principal contributors to root length in annual plants has also been consistently reported. For example, Pavlychenko (1937) found that, depending on the degree of competition between plants, the root length of oats, wheat and spring rye was formed of 45–93% of second-order laterals <0.1 mm in diameter. In a detailed quantitative study of winter rye roots, Dittmer (1937) reported that the length of second- and third-order laterals averaging 0.13 and 0.12 mm in diameter, respectively, made up >99% of the root system length, with over two-thirds of the length consisting of the finest third-order laterals. Kooistra et al. (1992) observed that, depending on soil bulk density, roots <0.2 and <0.3 mm in diameter could encompass up to 40 and 80%, respectively, of the root length developed by maize plants. Pallant et al. (1993) showed that in maize, roots <0.24 mm in diameter made the largest contribution to total root length. Moran et al. (2000) found that ≈80% of the root length of wheat was in roots <0.3 mm in diameter.

The main methods used to measure root length/biomass can be classified as (1) extraction methods (generically known as root washing); (2) mapping techniques; (3) in situ imaging techniques; and (4) other (often sophisticated) imaging techniques.

Extraction methods are based on collecting soil samples of known volume (core or monolith) from which roots are physically separated by carefully washing the soil away, and finally measuring the length of the separated roots using stereological or image analysis techniques (do Rosario et al., 2000). The principle of mapping methods is to record the occurrence of root contacts on a destructively exposed soil surface (van Noordwijk et al., 2000). Root contacts, whether enumerated on a pit face or a core surface with the naked eye, or on soil thin/polished sections using a microscope, are subsequently converted to length measurements according to a calibration procedure. With in situ imaging methods (chiefly involving reflection of visible light on observed objects), roots are observed at transparent interfaces with soil, such as the walls of transparent plastic tubes inserted into the soil for several months (minirhizotrons) (Smit et al., 2000). This allows dynamic monitoring of root growth and measurement of root length either directly, or based on calibration procedures. Finally, other imaging techniques involve probing (using electromagnetic radiation such as light, X-rays or γ-rays, particle beams or variable magnetic fields) of either field specimens or whole root systems confined within the delimited volume of specifically designed containers (the size of which is a function of the probing technique). The result is the reconstruction of either 2D (e.g. X-radiography) or 3D (X-ray CAT scanning, NMRI) images from which a range of root measurements can be derived by means of image analysis (Moran et al., 2000; Pierret et al., 2003).

It is well documented that all these techniques yield highly variable results (e.g. CV > 100% for minirhizotron and washing techniques), and that results obtained using two different techniques are, more often than not, difficult to compare. For example, Kucke et al. (1995) compared the core-break, trench-profile, core and monolith methods. They found good agreement between core and monolith methods, but obtained variable results with the core-break and trench-profile methods, the results from the two latter being poorly correlated with results from the two former. Unlike Heeraman & Juma (1993), they found more consistently lower CVs with monoliths (4500 cm3) than with cores (754 cm3). These authors interpreted the differences between mapping and destructive techniques as the result of (1) preferential orientations of roots, and (2) differences in root visibility depending on contrast with soil matrix. Tierney & Fahey (2002) noted differences between minirhizotrons and a radiocarbon method, but were able to analyse their results making sense of both data sets.

It would be somewhat hasty to dismiss the possibility that the variability of results gained through one technique may reflect real processes such as root proliferation in response to heterogeneous supplies of nutrient (Caldwell et al., 1992; Hodge, 2004). However, the regularity with which root-length measurements appear to vary depending on the technique used (see Literature review below) highlights the methodological, conceptual and scientific issues with which experimenters estimating root quantities are faced. One of the most startling of these issues is that the definition of the object itself – roots in general, and fine roots in particular – is still the subject of debate (Pregitzer, 2002; Zobel, 2003). As described by Zobel (2003), because we basically ignore much of how roots are organized and operate within a root system, we are still studying roots according to arbitrary size (diameter) classes and not according to more logical, function-related parameters. Hence issues as basic as the size distribution of root diameters down to the finest roots remain very unclear. Likewise, precise knowledge about root ontogeny and its morphological expressions, as well as root longevity/turnover, is lacking.

This paper provides a critical appraisal of methods used to measure the quantity (length and biomass) of fine roots. The critique is based on original measurements of fine roots obtained by means of a high-resolution X-ray imaging technique (Moran et al., 2000) and a review of literature. The high-resolution X-ray imaging technique was chosen because it permits the codetection of fine plant roots (down to ≈50 µm in diameter) and calibrated soil structure, and facilitates the quantification of the root/soil couple as a mutually interacting system. We investigated the extent to which root length may be underestimated, using a range of techniques, as well as the technical reasons that could explain such an underestimation. We present evidence showing that, because of the dominance of fine roots in most plant root systems, root-length recovery is strongly dependent on the observation scale at which the length measurement is carried out. Further, observation scales adequate for the quantification of fine root length require long processing times to obtain precise quantity estimates. As a consequence, plants could actually grow much longer (and larger biomass) root systems than is widely accepted. Our ability to understand fine roots will continue to be limited as long as these systematic errors are not taken into account in interpretations of plant functions.

Materials and Methods

Original X-ray root-length measurements

Samples analysed in this paper were taken under a canola crop (Brassica napus L. var Rainbow) grown on a farm located 40 km north-west from Wagga Wagga, NSW, Australia (34°52′ S, 147°05′ E). Mean annual rainfall and temperature are ≈560 mm and ≈15°C, respectively, with winter-dominant rainfall. The soil, a red Kandosol (Soil Survey Staff, 1998), was sampled at depths ranging from 5–115 cm on three occasions during the 1998 growing season: in June (just after sowing); at the end of August/early September (early grain filling); and at the end of October (before harvest). The canola crop measured in this study followed crops of lucerne (Lucerneago sativa L.) in 1996 and 1997. Before sowing, the soil was cultivated to a depth of ≈100 mm. Because of the underlying principles of the X-ray imaging technique (see below), the observations and measurements reported here not only correspond to the roots of the canola crop grown during the 1998 season, but also include information about decaying roots from previous seasons (Moran et al., 2000). More detail about sampling and cropping history is given by Pankhurst et al. (2002).

Soil samples were obtained by manually excavating monoliths ≈200 × 200 × 200 mm. For each sampling date, up to six such monoliths were extracted from the surface down a vertical sequence. In general, two replicates were obtained from each 20 cm depth increment. In the most favourable cases (when the monoliths did not crumble down while being excavated), the two replicates could be X-rayed effectively. However, a certain number of samples had to be discarded because their structure was not preserved intact. Undisturbed soil blocks (≈200 × 200 × 200 mm cubes) were air-dried in the laboratory for ≈12 wk and impregnated with polyester resin using a vacuum drip method (Moran et al., 1989). Subsamples (≈100 × 100 × 100 mm cubes) were extracted after completion of impregnation. From these subsamples, horizontal, 70 × 70 mm sections, 1 mm thick, were produced using a high-precision flatbed diamond-head grinder. Twenty such horizontal sections (six corresponding to the June sampling and seven for both September and October samplings) were imaged using a microfocus X-ray imaging system, the setup of which is described by Moran et al. (2000). To optimize image resolution, a geometric magnification of ×2.86 was applied by placing the sample at some distance from the image detector (Fuji BAS, imaging plates).

Once digitized, X-ray images were calibrated in terms of density using the method of Bresson & Moran (1998). Root traces were clearly visible on the images and were selected using a semi-automatic (computer-aided) procedure. Digital X-ray images were edited using adobe photoshop software and visualized at full resolution on a Wacom PL-400 LCD integrated tablet which allowed direct draw-on-screen interaction with the edited image. Detection of roots and root-related porosity (channels created by roots decomposing or decomposed at the time of exposure) relies on the recognition of a combination of densitometric and geometrical properties. This includes morphometric features of individual objects, but also contextual visual information related to the overall spatial arrangement of the root network imaged in a radiograph. In a previous study (Moran et al., 2000), we recognized the risk of mistaking small cracks for roots, and defined simple morphological criteria to discriminate them. When parallel enough to the section plane, roots and root-related pores appear as fibrous objects of increasing density from the centre to the edges (directly related to their cylindrical geometry). The more perpendicular to the section plane they become, the more roots appear as circular areas of low density. As opposed to these characteristics, small cracks are visible only when they are nearly perpendicular to the section plane, and show sharp edges with no gradual change in intensity from the edges to the centre as observed for roots. Based on these criteria, roots were traced using photoshop's inbuilt pixel-selection tools. The tracing, even though manual, was quite accurate because the operator could adjust the width of drawn lines down to 1 pixel (equivalent to 17.5 µm in the case of the images dealt with here) and could zoom in and out while editing the image.

Once the root images had been obtained, they were processed to derive root diameter distributions, root-length density estimates, and quantified indicators of the interplay between roots and soil structure such as the density of soil adjacent to roots, or the density of soil at increasing distances from roots of different diameter classes (see Moran et al., 2000; Pierret et al., 2000 for detail of procedures). Each X-ray image represents a snapshot, at a given point in time and soil depth, of the long-term co-evolution between successions of roots from crops locally. Objects designated as ‘roots’ can be live roots either recolonizing root channels established by previous crops or creating new pores (thus participating in the progressive refilling of ancient/abandoned root channels in the soil nearby), or they can be decaying and dead roots. The comparison of images corresponding to samples taken at close enough time intervals and similar soil depths can give some indication of the root dynamics corresponding to the growth of a given crop.

Literature-based comparison of techniques used to measure root length

We reviewed a number of papers dealing (whether chiefly or not) with root length measurements to test the hypothesis that there exists a correlation between measuring times and observation scales on the one hand, and root-length recovery on the other. Most references were found using the cab abstracts search engine facility (CABI Publishing North America, with various combinations of the words ‘root’, ‘length’ and ‘density’. Based on this initial search – which yielded >600 references for the expression ‘root-length density’– we retained studies dealing with both annual and perennial species, in which length density was expressed on a volumetric basis (as opposed to surface area) or, alternatively, in which the volume of soil from which root length was recovered was clearly reported. Minimum detail was also sought of (1) soil depth at which measurements had been made and (2) diameters of roots measured. A total of 29 such studies were found. These requisites were set to circumvent the need for excessive interpretation of published data and to provide consistency between the figures compared. As a direct consequence, studies in which root densities were expressed as root biomass per unit soil volume or surface area were not included in our comparison.

The studies selected encompass the five broad families of techniques available to measure root length: minirhizotron tubes (7); root washing (11); micromorphology (8); X-ray computed tomography (1); and projection X-ray imaging (2). A sixth category includes three references corresponding to extremely detailed measurements carried out in the 1930s, and based on a combination of very careful root washing and optical microscopy (Dittmer, 1937). For all but one of these categories of techniques (X-ray CT, because of the lack of references), we calculated the average, minimum and maximum root-length density. In addition, we used the results of a review study of 92 references by van Noordwijk & Brouwer (1991) to compare the root density values most commonly reported for 20 plant species as derived from root washing and minirhizotron observations.

Results and Discussion

X-ray based root measurements

Root diameter and root-length density  Overall, root diameters ranged from 0.052 to 3 mm. Fine roots clearly dominate, with ≈50% of the root length being ≤0.2 mm in diameter towards the end of the growing season (Fig. 1). On average, roots <0.5 mm in diameter accounted for >80% of the root length at all sampling times. Coarse roots (>1 mm in diameter) always represented a very small fraction of the overall root length (≈5–7%), though they could account, in a few instances, for a noteworthy part of the overall root surface area (data not shown).

Figure 1.

Evolution of root diameter distribution throughout the growing season. Error bars represent 95% confidence intervals for each distribution. Note the dominance of fine roots, as indicated by the percentage of overall root length (>80%) contributed to by roots <0.5 mm in diameter. Main plot, detail of the 0–0.6 mm root diameter range; inset, full root diameter range.

Fig. 1 shows the cumulative distribution of root diameter expressed as a percentage of total root length. For each diameter class we show the upper and lower 95% confidence limits which indicate that there is little evidence for statistically significant differences. To test for differences in the overall distribution functions for the three sampling times, we conducted two-way Kolmogorov–Smirnov (K–S) tests for the whole root diameter range (<3 mm) and for the fine roots (<0.6 mm). The K–S tests indicated that there is a shift towards a finer root length distribution from June to October. However, none of the distributions was statistically significantly different, irrespective of the upper diameter limit. We can interpret these results in (at least) two ways.

First, we can assume that the distributions are the same. Consequently, we cannot discount the possible implication that the lucerne and canola crops have the same proportions of root radii across the root system, even though they are very different types of plant, with different overall length densities, and have their roots in different parts of the soil (Fig. 2). Consistent interpretation would also have us state that this is true of the fine roots. This seems worthy of further investigation.

Figure 2.

Evolution of root-length density as a function of soil depth throughout the growing season (vertical error bars represent estimated error on soil depth; horizontal error bars are 95% confidence intervals for root-length density, when more than one sample was available).

Second, we can assume that the sample size (70 × 70 mm) is so small that variation in the root system is not fully described by our sampling. In this case we interpret the distributions as though they are different – the root system in October is finer than that in September which is, in turn, finer than that in June. Our interpretation is that, throughout the growing season, the proportion of fine roots tended to increase. In Fig. 1 it can be seen that roots <0.2 mm in diameter represented 33% of the overall root length in June, while this proportion rose to 46 and 54% in September and October, respectively. This suggested a shift towards smaller root diameters could reflect the replacement of the remnant root distribution of the lucerne crop, which had developed over the two previous years (represented by the initial root diameter distribution measured in June) by, on average, finer canola roots.

Not unexpectedly, given the massive amounts of fine roots detected in the images, root-length densities (Lv) derived from the X-ray images were very high (range 0.11–1.7 mm mm−3 throughout the growing season). From our data we can infer that, over the growing season (June–October), there was an overall increase in root-length density at soil depths ranging from 20 to 70 cm. The magnitude of this increase in root-length density could be of the order of 1 mm mm−3 (100 cm cm−3) in the 30–50 cm soil depth range (Fig. 2). These measurements are variable between samples taken at similar times and soil depths, which indicates heterogeneous root development (e.g. September sample at 55 cm soil depth, CV = 28%). It also appears that Lv measured in September at depths >50 cm were higher than in October. Given the limited number of samples and the destructive nature of the technique, it is virtually impossible to decide whether this pattern reflects an actual change in root exploration at this specific depth between the two sampling times or, more generally, the spatial variability of root exploration at any given time and soil depth. All we know is that the majority of roots – whether dead or alive – are registered in X-ray images as local variations in soil bulk density corresponding to the channels they created as they grew through the soil matrix (which does not apply to roots that grew in pre-existing macropores). Therefore roots grown only a few weeks before sampling would have left, at least, a detectable imprint in the soil matrix, even if they had died, dried and decayed in between the two sampling times. In other words, at a few weeks’ interval, without a sufficient number of replicate samples it is not technically possible to discriminate between the instantaneous spatial variability of root exploration and a temporal change in root exploration. Consequently, if the higher Lv values found in September without doubt reflect some sort of opportunistic foraging strategy, i.e. that roots proliferated where nutrients/water were more readily available (Smucker, 1984; Hodge et al., 1999; King et al., 2003), this could correspond to either a temporal response to shifting soil (or other) conditions, or to a synchronized response to soil heterogeneity.

Root spatial distribution and root/soil interplay  One advantage of the X-ray imaging technique is that it is not only restricted to the measurement of root length. By combining images of roots with the corresponding density-calibrated image, we are able to quantify the interplay between roots and soil structure. The subsoil (depth 20–100 cm) in which the roots of the canola crop developed had an average bulk density of 1.68 Mg m−3 (50 and 90% of the subsoil volume had bulk densities ranging from 1.47 to 1.88 Mg m−3 and 1.02–2.17 Mg m−3, respectively). Soil bulk density ρ increased linearly from 1.39 Mg m−3 at 5 cm to 1.82 Mg m−3 at 70 cm (ρ = 148.46 × depth (cm) − 203.75, r2 = 0.98). X-ray-derived soil bulk densities corroborated well with gravimetric measurements within the 5–50 cm soil depth range (Fig. 3), but at 50 cm the values derived from X-ray were, on average, 0.12 Mg m−3 denser. In June, before the canola had established any roots in the soil, the highest Lv values were found, counterintuitively, in the denser soil samples. This positive correlation between soil density and Lv disappeared in September and October (Fig. 4). Overall, from June to October it is possible (although difficult to establish firmly given the limited number of samples) that a shift in rooting pattern occurred, corresponding to the replacement of lucerne roots (represented by the June root-length density − soil bulk density function in Fig. 4) by canola roots (represented by the September and October root-length density − soil bulk density functions in Fig. 4). Overall we concluded that there was no relationship between soil bulk density and root-length density at the scale of whole soil sections (70 × 70 mm).

Figure 3.

X-ray and gravimetric measurements of soil bulk density as a function of soil depth.

Figure 4.

Root soil interplay: absence of relationship between bulk soil density and the presence of roots at the scale of whole soil sections (70 × 70 mm), as indicated by root-length density as a function of average soil bulk density in which roots developed.

The absence of a relationship between bulk soil density and the presence of canola roots at the scale of whole soil sections (70 × 70 mm) does not preclude the existence of an association between roots and soil structure at a finer scale. Exploratory measurements indicated that the maximum distance at which soil adjacent to roots differed in density from that of the bulk (computed for each individual soil section) was <0.1 mm for 85% of the roots (data not shown). Based on this result, the interplay between roots and soil structure was assessed at the local scale by measuring soil density values in a 5 pixel (87.5 µm) wide zone adjacent to the roots (the width of the measurement zone around roots was not altered as a function of root radius). This measurement revealed that, in the subsoil (soil depth >20 cm), roots of all diameters were surrounded by soil looser than the average density of the bulk soil (the 1.68 Mg m−3 average value derived from 19 images; Fig. 5). This rooting pattern was consistent throughout the growing season. At the time of sowing, roots and/or root pores corresponding to the previous (lucerne) crop were already preferentially located in the least dense parts of the soil volume, and the roots from the new (canola) crop appear to have followed the same growth pattern (September and October distributions, Fig. 5). Such root distributions indicate that roots preferentially explore and develop in the most porous parts of the soil volume. Previous microbiological measurements (Pierret et al., 1999; Pankhurst et al., 2002) also showed that these regions supported more abundant and active microbial populations than the denser matrix. The spatial association between roots of all diameters and less dense soil is observed at all soil depths down to 105 cm, and appears to be concomitant with the lesser decline in microbial activity observed in macropores as compared with the bulk soil (Pankhurst et al., 2002).

Figure 5.

Root–soil interplay: density of soil in contact with roots as a function of their radius. ◆, average values; solid line, smoothed average (spline interpolation); dotted lines, 95% confidence interval; horizontal dashed line, average subsoil bulk density (1.68 Mg m−3).

A feature of Fig. 5 is that the thinner the roots, the denser the soil in which they were located. This is in contrast to what might be expected given evidence from controlled experiments in homogeneous soil of varying density (Bengough et al., 1997). In such studies, roots faced with denser (stronger) soil tend to be thicker than those in less dense soil. It is tempting to speculate that this points to a growth pattern of exploration of the soil structure/water/oxygen/nutrient complexity as opposed to a conceptual model of roots as battering rams. The plant initially preferentially grows into previously formed macropores, accessing sufficient resources by progressing vertically though the soil. Periodically, finer roots extend into the surrounding soil matrix. These roots can penetrate smaller pores, particularly if moisture conditions favour reduced soil strength (differentially controlled in the same soil by soil water potential) without excessively restricting the flow of gases. The proximity of these initial forays to macropores ensures that oxygen supply is nonlimiting, as the diffusion rate of oxygen in air-filled pores is much faster (≈104 times) than in water-filled pores. This has been demonstrated previously for the growth of fungal hyphae (Harris et al., 2003; Otten et al., 2004).

If such fine root ‘explorations’ are successful at accessing resources, the plant may respond by prioritizing allocation of assimilate into that region, with the consequence of producing more fine roots. Eventually, limiting factors will constrain further proliferation in a zone. Pregitzer et al. (1998) demonstrated that in trees, among roots, fine roots in zones of high nutrient availability had the highest respiration rates. This illustrates the importance of oxygen consumption and/or displacement of available oxygen by CO2 from respiration. Armstrong (1979), using a model parameterized by reasonable assumptions for soil and root conditions, showed that in aggregated soil roots <0.05 cm could experience anaerobic conditions if the effective aggregate diameter is as small as 1 mm. In our subsoil material, it is likely that fine roots will frequently be exploring beyond this distance from larger ‘interaggregate’ pores. Equally, consumption of the available nutrient and/or water may result in limiting conditions. Similarly, greater success elsewhere in the root system (perhaps as a result of diminishing return caused by nutrient or water depletion) may result in reduced priority access to the assimilate required to continue to proliferate fine roots. This sort of feedback, combined with spatial (and potentially temporal) heterogeneity of resource distribution, could result in what would appear, in retrospective measurements, to be excessive, chaotic bursts of fine root growth. This sort of growth pattern would result in large local variations in root-length/diameter relationships if observation specimens are small relative to the extent of the root system (Fig. 1).

A similar model of soil exploration, in which root growth and uptake are driven by the local balance between supply and demand for water and nitrate, was simulated by Dunbabin et al. (2002a,b). Given the evidence of differential biological environments between the bulk soil and the regions of root exploration (Pierret et al., 1999; Pankhurst et al., 2002), it is possible that a synergy, or even a symbiosis-like relationship, exists between plants and soil flora through exploitation of heterogeneous soil resources. This additional level of rhizospheric feedback could be added to simulation models to assess the extent to which this may further drive fine root proliferation.

Our results on soil/root associations are in good agreement with our previous X-ray-based measurements of another soil/crop combination (Moran et al., 2000) and with micromorphological observations of plant roots and macropores (Stewart et al., 1999). Our observations of differential root growth depending on soil structure/local soil density are also consistent with the results obtained using a variety of techniques. For example, based on the profile wall method (Böhm, 1979; van Noordwijk et al., 2000), Ehlers et al. (1983) detected a broad association of oat roots with biopores 2–10 mm in diameter. Smucker & Aiken (1992), using root washing and minirhizotrons, found reduced length density of maize roots in dense peds (Lv << 3 × 10−3 mm mm−3 at distances >2 cm from the peds surfaces) compared to that in the softer matrix (Lv ≈ 1–1.5 × 10−2 mm mm−3) in which the dense peds were distributed. These authors noted that root growth is often confined to volumes between peds and to their surfaces, and that interiors of larger peds are sources of water which are not available to roots growing in bulk soil. Although this must be interpreted with the proviso that root washing is unlikely to have identified the finest roots, which, as argued below, are more likely than coarser, identifiable roots to have been present in such conditions. More recently, Pardo et al. (2000) and Amato & Ritchie (2002) also reported substantially lower Lv in distributed clods compared with looser bulk soil based on core washing measurements. Micromorphological observations reported by van Noordwijk et al. (1992) also indicate that the most common growth habit of maize roots is to explore ‘… voids between the aggregates or … places which had been voids before the soil was compacted’ and that ‘… rarely did small roots grow through aggregates’.

Literature review of comparative performance of root observation techniques

General findings  When comparing the respective performances of the principal techniques used to date to measure root length, our aim is to draw attention to the observations that:

  • 1there are important discrepancies between average lengths obtained with different measurement techniques;
  • 2root length measurements obtained with any technique are inherently and sometimes outstandingly variable for the same species/crop;
  • 3there appears to be a clear correlation between the size-detection threshold of a technique (the smallest root diameter that is effectively measured) and the overall root length measured.

The paucity of published data on root-length density for canola is one reason that has motivated us to widen the range of references in this review to studies dealing with other plant species. Indeed, we found only one reference (Merrill et al., 2002) including measurements (derived from minirhizotron observations) of root-length density for field-grown canola. Consequently, some of the deviations we report here are likely to be influenced by the bulking of results obtained under different growth conditions for a range of plant species differing in their root growth habits (variability external to the differences between techniques which we initially wanted to discuss). For example, Hamblin & Tennant (1987) reported that cereal species consistently had five to 10 times the root length of grain legume species. However, despite this, Lv values reported in the literature (as indicated by their order of magnitude) appear to depend on the technique with which they were obtained. Fig. 6 shows the average, minimum and maximum root density values derived from our literature review for each of the six categories of technique available to measure root length (see Materials and Methods). This figure (note log scale) shows that the average Lv values increase according to the following sequence: minirhizotron ≤ root washing << micromorphology ≈ CAT scanning ≈ X-ray projection < root washing work conducted in the 1930s. The difference between the most commonly used techniques and more complicated alternatives is far from anecdotal, as it averages at least one order of magnitude (8 × 10−2 mm mm−3 for root washing vs 7.5 and 9.7 × 10−1 mm mm−3 for micromorphology and X-ray imaging, respectively).

Figure 6.

Comparison of Lv values obtained using the principal techniques that have been used to measure this parameter. Note logarithmic scale for Lv values. Bars, original data extracted from 29 studies published between 1937 and 2004. Horizontal plain and dashed lines represent, respectively, the 95% confidence interval and maximum and minimum Lv values reported by van Noordwijk & Brouwer (1991) for 20 plant species, using minirhizotron and root washing (see Fig. 7). The references used are as follows. (Exponent numbers in front of references indicate the technique(s) dealt with in the quoted papers: 1minirhizotron; 2root washing; 3micromorphology; 4CAT scanning; 5X-ray imaging; 61930s studies.) 1,2Asseng et al. (1998); 3Babel et al. (1995); 2Bouma et al. (2000); 1Buckland et al. (1993); 2Costa et al. (2002); 2Cresswell & Kirkegaard (1995); 6Dittmer (1937); 6Dittmer (1938); 2Ehlers et al. (1983); 2Hamblin & Tennant (1987); 2Heeraman & Juma (1993); 2,4Heeraman et al. (1997); 1Kirkham et al. (1998); 3Krebs et al. (1994); 3Melhuish & Lang (1968); 3Melhuish & Lang (1970); 1Merrill et al. (2002); 1Meyer & Barrs (1991); 5Moran et al. (2000); 6Pavlychenko (1937); 5this study; 1,2Smucker & Aiken (1992); 3Stewart et al. (1994); 3Stewart (1997); 1Vamerali et al. (1999); 3van Noordwijk et al. (1992); 3Veen et al. (1992); 2Xue et al. (2003); 2Yano et al. (1998).

As the most sophisticated/time-consuming techniques have been covered in only a limited number of papers, we believe that the part of our review focusing on these techniques is reasonably comprehensive. This contrasts with the case of the two most commonly used techniques, root washing and minirhizotron, for which there exists a huge body of literature. For example, a search with the key expression ‘root-length density’ on the cab abstracts database yielded 611 references for the period from 1973–2004, with the highest score (13.4%) in the journal Plant and Soil. If the search is broadened to ‘root length’ then the corpus of references becomes more than 10 times greater. A detailed analysis of such a vast literature is not necessary to gain a consistent picture. Based on four previous review papers (van Noordwijk & Brouwer, 1991; Jackson et al., 1996, 1997; Zuo et al., 2004), we feel confident that the values presented here as representative of the average yield of the root washing and minirhizotron methods are not anecdotal. First, in a recent study aimed at producing a generalized function of wheat root-length density, Zuo et al. (2004) examined 89 data sets from 10 papers which indicate a maximum Lv of the order of 6 × 10−2 mm mm−3 for root washing and minirhizotron observations. Jackson et al., 1996, 1997) compiled 250 field studies dealing with root measurements in a variety of biomes, from which they selected six references for crops that indicate a maximum root biomass of the order of 2 kg m−3. Assuming an average specific root length of 118 mg−1, these authors estimated that, for grass species, a maximum value of 2.36 × 10−1 mm mm−3 can be expected using root-washing techniques. Finally, a review study of 92 references by van Noordwijk & Brouwer (1991) shows that root-length density values derived from root washing and minirhizotron fall within the range 0.01–0.1 mm mm−3 for 14 plant species with a maximum of 2.6 × 10−1 mm mm−3 for wheat (Fig. 7). These data are in good agreement with the range we derived from our own literature review for root washing and minirhizotron measurements (Fig. 6).

Figure 7.

Root-length density values derived mostly from root washing and minirhizotron measurements for a range for 20 plant species (Note logarithmic scale for Lv values). Original data: van Noordwijk & Brouwer (1991).

Pros and cons of the root-washing and minirhizotron methods  In practice root washing, whether implemented manually or with the aid of the so-called hydropneumatic elutriation apparatus (Smucker et al., 1982), which allows faster, automated processing of samples, is without doubt the most widely used method to quantify root length and biomass. Its consensual adoption through time results more from pragmatic considerations (basically an optimal trade-off between rapidity, cost and precision) than from indisputable accuracy and reproducibility (Kucke et al., 1995; Jackson et al., 1996). Several papers focusing on root washing (Amato & Pardo, 1994; Livesley et al., 1998; Gathumbi, 2004) provide quantified evidence that there is a clear inverse relationship between sieve mesh size and root recovery, even when care is taken to separate the finest roots from the soil (through repeated cycles of sedimentation and decantation). In their original description of the hydropneumatic elutriator, Smucker et al. (1982) clearly recognized this inverse relationship, which is why they recommended the use of a 0.075 mm sieve for maximum root-length recovery. Consistent with this recommendation, Livesley et al. (1998) reported a root-length recovery of <40% with a 2.0 mm compared with a 0.25 mm sieve, associated with a twofold increase in mean root diameter recovered when using the 2.0 mm sieve. Amato & Pardo (1994) reported an even more extreme relationship between root recovery and sieve size, with a total root length collected with a 2 mm sieve of only 10% of that retrieved with a 0.2 mm sieve. Considering that the most readily available commercial version of Smucker and coworkers’ hydroelutriation apparatus, the Delta-T root washer (, is delivered with a 0.5 mm mesh filter, there appears to be a significant risk that unquestioning use of such a device results in root length being vastly underestimated. There are also concerns that many fine roots are broken off and washed away with the soil, leading to biased length estimates (Pearson & Jacobs, 1985; Livesley et al., 1998).

Minirhizotron tubes have consistently been reported as an effective means of monitoring root development, at least qualitatively (Smit et al., 2000). However, there are still difficulties in converting minirhizotron root counts into more usable values such as volumetric root densities or root biomass. Several authors have attempted to cross-validate the simple conversion equation proposed by Merrill & Upchurch, 1994) by calibrating minirhizotron data with independent root-length density measurements (core sampling method, De Ruijter et al., 1996; root washing, Heeraman & Juma, 1993). However, high CVs typical of all root measurements, in particular caused by the widespread variability of root system architecture and its local expression (response to soil constraints, to hotspots of nutrients, etc.), render the validity of such a calibration exercise extremely limited (Heeraman & Juma, 1993). In addition, root identification in minirhizotron images is known to be ambiguous/difficult under some circumstances (Liedgens & Richner, 2001). As direct, automated root extraction from images is almost always impossible, the minirhizotron method relies, to some extent, on the user's ability to recognize roots. Finally, in a review paper on fine root research with minirhizotrons, Johnson et al. (2001) also provide experimental and simulated evidence that, because of their relatively rapid turnover rate, a large proportion of fine roots could be missed with this observation technique if successive samplings are carried out more than 1 wk apart. This is true of all techniques that have longer sampling times. On the other hand, minirhizotron tubes have been reported as an effective means of detecting roots as fine as 0.09 mm in diameter (Wells & Eissenstat, 2001). Hence it appears that this technique is potentially well suited for assessing root length/biomass accurately, provided there is enough sampling and effective image analysis is available for automated data processing. Currently, automatic retrieval of root data from minirhizotron images still appears to be an issue, which might explain why, to date, this technique has been successfully used in root turnover/survival studies, but not so much for the assessment of root length/biomass.

How do alternative techniques compare with root washing and minirhizotron?  In the light of most literature dealing with root-length density measurements, the X-ray-derived Lv values reported here (0.11–1.7 mm mm−3) may appear overestimated. However, for comparison, several other authors, using different imaging techniques, obtained root-length density estimates of the same order of magnitude as those reported here. Using a computer-aided micromorphological system, Krebs et al. (1994) measured root-length densities ranging from 0.5 to 3 mm mm−3 for pasture soils, at depths of 0–5 cm. In this case, roots <200 µm were measured on soil polished sections. Using a similar technique, Stewart (1997) reported root-length densities ranging from 0.12 to 0.31 mm mm−3 for wheat at soil depths ranging from 7 to 33 cm. Using X-ray microtomography, Heeraman et al. (1997) obtained root-length densities ranging from 0.75 to 0.78 mm mm−3 for bush beans grown in small pots for only 14 d. These authors also reported that the values obtained with microtomography were about 1.5× greater than those derived from destructive measurements (0.44–0.59 mm mm−3 using core washing). Experiments conducted in the 1930s, based on careful (hence tedious) root-washing exercises, yielded Lv values of the same order of magnitude (Pavlychenko, 1937) to one order of magnitude higher (Dittmer, 1937) than those obtained using modern imaging techniques. Dittmer's (1937) study, which combined root washing with microscopic measurements using a graticulated eyepiece, is strongly in favour of the hypothesis of a correlation between spatial resolution and root recovery (Dittmer measured the length of all roots down to third-order laterals only 0.12 mm in diameter).

Image-based techniques such as micromorphology or X-ray projection imaging are, of course, not free of limitations and/or possible biases. For example, regarding root counts by means of micromorphology, Melhuish & Lang (1970) indicated that ‘The utility of the method … rests strongly on the user's ability to recognize roots [and that] for fine roots, this can be difficult …’ This, to some extent, also applies to the X-ray method used in this study, and the uncertainty that might ensue is discussed by Moran et al. (2000). Nevertheless, root recovery of senescent and decaying roots and of fine, near translucent, roots is still likely to be better with micromorphology and X-ray projection imaging than with conventional techniques. In effect, while these two categories of roots are easily overlooked during minirhizotron observations (because of a lack of contrast with the soil matrix), or lost during root washing (because they are particularly fragile and easily broken), they most often create imprints detectable by micromorphology or X-ray imaging, even though part of or all the root material has vanished by the time of sampling (Moran et al., 2000). Conversely, the extra length measured by means of the micromorphology and X-ray techniques (as compared with conventional techniques) is unlikely to relate to the unwanted inclusion of root hair and endomycorrhizal (or arbuscular mycorrhizal) hyphal length, as the diameters of these are 12–15 and <5 µm, respectively (Staddon et al., 2003), well beyond the spatial resolution of the X-ray technique used in this study (which in practice is ≈50 µm) and of micromorphological observations. (Despite 15 µm not being beyond the reach of optical microscopy, micromorphological observations are usually conducted at low magnification.)


A review of the literature has demonstrated that root-length recovery varies by as much as one order of magnitude depending on the technique used. As a general principle, there is a clear correlation between fine root detection and/or processing times on the one hand, and root-length recovery on the other. There exists ample evidence that profuse fine roots can easily be overlooked when using root washing and minirhizotron, the most convenient, affordable and rapid, and hence most widely used, techniques to assess root length and biomass.

Based on original root measurements and a review of an array of literature spanning more than six decades, this paper confirms that the fine roots of an annual crop (canola) contribute significantly to its overall root length. In our study, roots <0.2 mm in diameter represent up to >50% of root length, a result that is in agreement with the findings of Pallant et al. (1993), who showed that in maize, roots <0.175 mm in diameter can account for more than 56% of root length.

Statistical analysis of the X-ray-based root-length and diameter information leads us to two potential interpretations. First, the lack of statistical difference between root distributions indicates that two different crops, and one crop at two different times, have the same relationship between root length and root diameter. This is consistent for coarse and fine roots. Such a result, if true, warrants further investigation. Second, if the lack of significant differences is caused by local spatial variation, the canola crop has a finer overall root distribution than lucerne, and this becomes more emphasized throughout the growing season. The latter interpretation is consistent with visual inspection of the distribution functions.

We conclude that plant root systems are likely to be much longer and include more biomass than is widely accepted. Despite sustained technological progress leading to improved understanding of root function and biogeochemical cycles, we believe there is an urgent need for careful studies aimed at accurately quantifying root length and biomass under a range of biophysical conditions and for a wide array of plant species.


This research was financially supported by CSIRO Land & Water, the Australian Grains Research and Development Corporation and INRA. We thank Mr Colin McLachlan, CSIRO Land & Water, Canberra, for invaluable technical assistance with the production of most of the original measurements presented in this work. We are grateful to Dr John Passioura, CSIRO Plant Industry, for support and guidance during the several laborious years of work required to gain confidence that the early results from X-ray imaging were not an aberration of the technique. We thank Dr Yves Dudal, INRA Climat, Sol & Environnement, Avignon, for the helpful comments on the manuscript.