USING INFRARED SPECTROSCOPY FOR DETECTION OF CHANGE IN SOIL

PROPERTIES IN SELECTED LANDUSES IN MT. MARSABIT ECOSYSTEM, NORTHERN KENYA

Caroline Achieng Ouko Research Scientist

CETRAD | P.O Box 144-10400 | Nanyuki | Kenya

Independence: 12th December 1963

Ethnic groups: 43 Area: 592000 km2

Location: 5o north & 5o south & between longitudes 34o & 42o east

Altitude: variable from 0 to 5000m above sea level

Climate (equator, topography, Indian ocean, ITCZ, habitat & ecology)

Economy: heavily rely on natural resources: forests

Introduction

Forests provide essential ecosystems services but have been depleted. The change from natural forest cover to agricultural and pastoral activities is

rampant

Introduction Contd..

Study Area

Mt. Marsabit Forest depletion.

Study Introduction

Study Introduction Contd. The conventional assessment methods to

determine soil degradation are expensive, time consuming and very specific.

The assessment of diverse effects of land use and land use change on soil productivity requires methods that can provide rapid and integrated assessments

Developments in laboratory and field based reflectance spectrometry present a unique capability for rapid, cheap, integrated assessments and routine monitoring of soil productivity status

ObjectiveTo evaluate the use of near infrared

spectroscopy for non-destructive characterization

And prediction of management sensitive soil properties under different land use

systems.

Methodology Three transects cutting across the chosen

land use patterns namely forest, cropped and pastureland (GPS).

Soil sampling for physical and chemical analysis.

Above ground carbon stocks estimated in the different LUS according to Woomer et al 1998.

222 augured soil samples from 0 – 20 and 20 – 50 cm.

Calibration set was a third of the total (74 samples)

Study Area

Methodology cont. The air dried soil passed through a 2-mm sieve

was packed in 12 mm deep and 55 mm diameter Duran Petri dishes.

The samples were scanned through the bottom of the Petri dishes using a high intensity source probe

The probe illuminated the sample giving a correlated color temperature of 3000 K.

Reflectance spectra were recorded at two positions, successively rotating the sample dish through 90o between readings to sample within dish variation.

Methodology cont. X-ray fluorescence (XRF) is used to detect and

measure the concentration of elements in substances. Fluorescence - phenomena of absorbing incoming radiation and reradiating it as lower-energy radiation.

Methodology cont. Reflectance readings of each wavelength

band were expressed relative to the average of the white reference readings.

Spectroscopic transformation was applied to convert spectral reflectance to absorbance

Principle component analysis was implemented in Unscrambler version 7.5

Individual soil variables were calibrated against 214 (0.36-2.49 µm) reflectance bands using Partial Least Squares (PLS) regression

Results and Discussion Mean relative reflectance varied

among the three LUS

The mean soil spectral reflectance from the three LUS exhibited similar pattern indicating similar mineralogy.

Results and DiscussionRelative reflectance averaged across the entire spectrum (albedo) of all the soils ranged from 0.025 to 0.28.

0

0.05

0.1

0.15

0.2

0.25

0.3

0 0.5

1 1.5

2 2.5

3

Wavelength (µm)

Rel

ativ

e re

flect

ance

F

C

R

Near Infrared Reflectance Spectroscopy of forest (F), cropland (C) and pastureland (R) soil samples.

Regression of soil properties measured by standard laboratory procedures and predicted by NIRS – PLS

techniques. pH

y = 0.9504x + 0.3598R2 = 0.9465

6.5

7

7.5

8

8.5

6.5 7 7.5 8

Measured values

Pre

dict

ed v

alue

s

1

Total Carbon

y = 0.97x + 0.0847 R2 = 0.973

0 2 4 6 8

10 12

0 2 4 6 8 10 12 Measured values (%)

Pre

dict

ed V

alue

s

1

Magnesium

y = 0.954x + 0.3733 R2 = 0.9435

5 7 9

11 13 15

5 7 9 11 13 Measured values (ppm/100g)

Pre

dict

ed v

alue

s

(ppm

/100

g)

1

Potassium

y = 0.3329x + 7.6048 R2= 0.3516

0

5

10

15

20

0 5 10 15 20 25 Measured values (meq/100g)

Pre

dict

ed v

alue

s (m

eq/1

00g)

1

CEC

y = 0.7072x + 8.9154 R2 = 0.7629

20 25 30 35 40 45 50

20 30 40 50 Measured values (Cmol/g)

Pre

dict

ed v

alue

s

(Cm

ol/g

)

1

Total Nitrogen

y = 0.9884x + 0.0077 R2= 0.9537

0 1 2

0 1 2 Measured values

Pre

dict

ed v

alue

s (%

)

1

Calcium

y = 0.8309x + 1.5033 R2= 0.7626

5

10

15

20

5 10 15 20 Measured values

Pre

dict

ed v

alue

s

(ppm

)

Measured N

Linear (PredN)

NIRS Prediction of Soil Properties using Partial Least Square Regression

Carbon, nitrogen, pH, exchangeable magnesium & calcium, and CEC were reliably predicted (r2 > 0.76) by NIRS-PLS.

Cross validation models with high regression (r2) values such as those obtained with N, CEC and exchangeable Ca also yielded large validation r2 values (r2 > 0.76).

These properties are highly correlated with carbon (r2 = 0.97).

pH (r2 = 0.95) and exchangeable Mg (r2 = 0.94) were more accurately predicted by NIRS-PLS than would be expected based on their correlations with carbon.

Correlation coefficients of extractable K was very low (r2 = 0.35) probably due to its luxury consumption and leaching as weathering advances.

Estimated total A&BG carbon stocks (ton/ha) under different land use

systemsCarbon stocks

Forest Cultivated Rangeland

Above C 4.3x109 1.5x109 1.7x109

Below 1.4x109 1.2x109 0.9x109

Total 5.7x109 2.7x109 2.6x109

Mean SE 24.5 20.3 19.2

Conclusion and Recommendations

Soil carbon content is useful to assess rate and extent of land degradation.

This study has shown that conversion of forests to agricultural use affects the soil properties.

The carbon stocks were especially affected in that the carbon sinks were reduced in the converted land use systems and a decline in carbon stocks of 45.6% and 47.4% in the pasture and cropped land was observed.

Conclusion and Recommendations

NIRS was sensitive to changes in soil properties caused by forest conversion and gave good estimates of management induced changes in soil properties including total C and total N, CEC, exchangeable Ca and Mg, and particle size distribution.

Significantly, these are primary soil constituents for which a theoretical basis for reliable NIRS prediction exists.

NIRS spectroscopy offers great potential for estimating and monitoring variations in these constituents under different land use land management scenarios.

Conclusion and Recommendations

Future direction for research is to develop a spectral library of referent or benchmark sites at a landscape scale. The spectral separability of managed systems relative to an undisturbed benchmark offers a new research vision

Future studies should explore approaches that combine Discriminant analysis and strategic spectral libraries of pre-agriculture (benchmark) soil conditions with information on physical, chemical and biological properties.

The reliable methods will accelerate the development of risk-based approaches that explicitly account for site history and land use management.

Future Outlook

Looks rocky and steep

Look for opportunitiesDivine intervention

Acknowledgment

Research funds were provided by AGREF

Thank you for your attention

References Analytical Spectral Devices Inc. (1997). FieldSpecTM User’s guide.

Analytical Spectral Devices Inc., Boulder CO.

Food and agriculture organization (FAO), Forest Resources Assessment (1990). Global Synthesis, FAO Forestry Paper 124, FAO, Rome, Italy, 1995.

McCarty, G.W., Reeves, J.B., Follett, R.F., and. Kimble, J.M., (2002). Mid-infrared and near-infrared diffuse reflectance spectroscopy for soil carbon measurement, Soil Sci. Soc. Am. J. 66 (2), pp. 640–646).

Naes, T., Isacksson, T., Fearn, T. and Davies, T. (2002). A user-friendly guide to Multivariate Calibration and Classification. NIR Publications Chichester, UK.

Shepherd, K.D. and Walsh, M.G., (2007). Review: Infrared spectroscopy-Enabling an evidence-based diagnostic surveillance approach to agricultural and environmental management in developing countries. Journal of Infrared Spectroscopy 15, 1-19.