Aller au contenu
Aller au menu
Aller à la recherche
Aller au pied de page

Spectral indicators of biological soil quality (SpecBio)

Spectral indicators of biological soil quality (SpecBio)

Summary
There is increasing demand for diagnostic and monitoring tools of soil quality. Quantitative infrared spectroscopy has demonstrated its potential for providing fast, inexpensive and accurate determinations for a range of basic soil properties, for instance carbon (C) and nitrogen contents. Its interest for addressing more complex soil properties is less established. The SpecBio project aimed at assessing the potential of quantitative infrared spectroscopy for determining C protection in soils and the vulnerability of this protection to temperature (warming).
Two 100-sample topsoil populations were studied:
- one originated from France, from the Réseau de mesures de la qualité des sols (RMQS), which is a national soil quality monitoring network; this population included acid soils (Brittany, Massif Central, Pyrenees, Alps, etc.) ;
-the other population originated from the northern half of Tunisia, and included mostly carbonated soils.
All samples were scanned in the near infrared (NIR, 1100-2500 nm) and mid infrared (MIR, 2.5-25 μm, i.e. 4000-400 cm-1); for both ranges, spectrum acquisition is less than one minute and does not require any consumable. Soil samples were then incubated (without preincubation) during three weeks, either 2- or 0.2-mm crushed, at 18 or 28°C (i.e. four treatments, with three replicates per treatment, leading to a total of more than 2400 incubations). Protected C was defined as the difference between C mineralized during incubations of 0.2- and 2-mm crushed soils (soil microorganisms cannot access organic matter protected within < 2-mm aggregates, but can access it more easily once these aggregates have been crushed). The vulnerability of C protection to warming was defined as the difference between C protected at 18 and at 28°C.
Prediction models were built for about 60 variables (i.e. soil respiration, C protection, vulnerability, in several conditions, over several periods, expressed as proportions of total soil or soil C, etc.), using four types of soil spectra (NIR or MIR spectra from 2- or 0.2-mm crushed samples), testing ten mathematical pretreatments of spectra (e.g. mean subtraction and standard deviation division, derivatization, etc.), which represented more than 2000 models for each soil population studied.
Using NIR or MIR spectra, accurate predictions of soil respiration were achieved for the Tunisian samples, either 2- or 0.2-mm crushed, at 18 or 28°C (R² ≥ 0,71 between predictions and measurements); the most accurate predictions were achieved using NIR spectra of 0.2-mm crushed samples (R² ≥ 0.85). Protected C was accurately predicted at the beginning of incubation (R² ≥ 0.65 over 7 days), especially using NIR spectra of 0.2-mm samples (R² ≥ 0.84), but less accurately then (R² ≥ 0.54 over 21 days), especially at 28°C; data might have been less reliable at the end of 28°C incubation. Vulnerability of C protection to warming was poorly predicted; it was calculated as a difference (18 vs. 28°C) between differences (respiration of 0.2- vs. 2-mm soil), which resulted in uncertainty propagation and might explain why spectra were poorly fitted to these data.
Looking at spectral regions that contributed heavily to prediction models provided useful information on chemical groups involved in the processes studied: so soil respiration appeared to be promoted by soil richness in nitrogenous organic compounds (proteins, amides, amines); and also by richness in aliphatic organic compounds, at 18 and 28°C, and aromatic organic compounds, at 28°C especially (aliphatic substrates, which are more labile, are used more rapidly at 28°C, then soil microorganisms use aromatic substrates, which are more recalcitrant). Carbon protection was also promoted by richness in aliphatic and nitrogenous organic compounds. All these influences could be seen on both MIR and NIR spectra. Organohalogenated compounds, organochlorinated especially, seemed to have a negative effect on soil respiration and C protection, which could be seen on NIR spectra (many pesticides are organochlorines). Carbonates also seemed to affect soil respiration and C protection negatively, which suggested they were poorly involved in carbon dioxide emission (at the time-scale of incubations).
For the French samples, predictions using NIR and MIR spectra were disappointing in general for soil respiration (R² ≤ 0.71) and poor for C protection (R² < 0.3). Several hypotheses might help explain worse results than for the Tunisian soil population:
- higher population heterogeneity, leading to poorer calibration;
- mineralogy less favorable to spectrum fitting, for instance affecting (and providing noise in) spectral regions involved in soil respiration;
- sample collection over a 10-year period, which might increase population heterogeneity (e.g. as regarded microbial communities) thus impede spectrum fitting to reference data, respiration data especially (by contrast, sample collection in Tunisia was carried out within a few weeks).
It was difficult to decide between NIR and MIR ranges: in general NIR yielded better predictions for the Tunisian population but MIR often tended to yield better predictions for the French population, using spectra from 0.2-mm crushed samples in both cases. Predictions were often less accurate using spectra from 2-mm crushed samples but not often (e.g. with NIR on French samples), anyway they were interesting as this soil preparation is less tedious than fine crushing; using spectra from 2-mm samples, more accurate predictions were achieved with NIR than with MIR. Thus the NIR domain could be considered slightly more interesting than MIR.
The SpecBio project confirms the interest of quantitative infrared spectroscopy as a time- and cost-effective, non-toxic, technology for providing soil quality indicators, even complex ones (carbon protection). This approach should thus develop as a tool for diagnostic and monitoring of soil quality indicators; this will require calibration data building for quality indicators considered relevant, as the SpecBio project has begun to do for C protection in some soil types of France and Tunisia. Nevertheless this approach has limitations: reference data used for calibration should not include large uncertainty (e.g. resulting from successive calculations), and sample populations should not be too heterogeneous.

haut de page

 

Contact
IRD Montpellier (Bernard Barthes)

haut de page