collecting the dirt on soils
TRANSCRIPT
Ermias Betemariam†, Calogero Carletto˚, Sydney Gourlay˚, Keith Shepherd†
† World Agroforestry Centre˚ The World Bank
29th International Conference of Agricultural EconomistsMilan, Italy – August 8-14, 2015
Collecting the Dirt on Soils: Advancements in Plot-Level Soil Testing and
Implications for Agricultural Statistics
Support• LSMS Methodological Validation Program, funded by UK Aid
Objectives• Test subjective approaches to measurement vis-à-vis
objective methods for land area, soil fertility & crop production
Partnerships• Central Statistical Agency, World Agroforestry CentreStatus• Fieldwork: Completed March 2014• Soil Testing: Completed March 2015• Analysis & Dissemination: On-going
Land and Soil Experiment Research (LASER): Ethiopia
Methodologies tested:Land Area • Traversing (i.e., compass and
rope)• GPS measurement (Garmin)• GPS measurement (Android
tablet)• Farmer self-reported area• Clinometer• Farmer self-reported incline
Completed duringthe post-planting
visiton up to two fieldsper household
Soil Fertility
• Spectral Soil Analysis • Conventional Soil Analysis • Farmer self-reported soil
quality
Samples collected during the post-planting visit, processed at
regional labs and shipped
to ICRAF Nairobi for analysis
Maize Production
• Crop-cutting using a 4m x 4m subplot and 2m x 2m subplot
• Farmer self-reported harvest
Completed by field
teams when alerted
by household
LASER Methodologies1018
households interviewed
1799 fields
selected for objective
measurement and soil testing
3791 soil samples collected*
205 fields with
crop-cutting*2 samples were collected from each field (different depths and sampling procedures), an additional sample was collected on fields with crop-cutting.
Survey Design• 3 household visits:
– Post-Planting– Crop-Cutting– Post-Harvest
• 5 mobile field teams
• CAPI administration
Elevation Rainfall AEZ
LASER Sample
85 EAs12 HH Each
Soil Sampling Protocol
Soil Sampling Protocol
Objective Data• 100% of samples were
tested with:– Mid-infrared diffuse
reflectance (MIR) spectroscopy
– Laser diffraction particle size distribution analysis (LDPSA)
• 10% of samples tested with:– Conventional wet
chemistry analysis– X-ray methods for
mineralogy (XRD)– Total element analysis
(TXRF)
Objective Data
Predictive power of
MIR spectrosco
py
Objective DataMean SD
Physical% Sand 12.3 7.1% Clay 65.1 12.8% Silt 22.6 7.4
ChemicalpH 6.3 0.6
Macronutrients:Total Carbon 3.4 1.3
Total Nitrogen 0.3 0.1Exchangeable Calcium+ 3454 1821
Potassium+ 735 278
Exchangeable Magnesium* 539 201Micronutrients:
Iron+ 160 63
Zinc+ 6 4
Phosphorous+ 46 44
Exchangeable Manganese+ 182 51+ Extracted with Mehlich 3 method* Extracted with wet method
Top Soil
Objective Data
Distribution of soil organic carbon by
administrative zone.
Objective Data
• variation of soil properties within zones…
02
46
8S
oil O
rgan
ic C
arbo
n (%
)
excludes outside values
West Arsi Zone
Enumeration Area, West Arsi Zone
…and within enumeration areas
Objective Data
Subjective DataWhat is the quality of soil on [PLOT]?
Subjective Soil Quality
Good 42%
Fair 53%
Poor 5%
ET21%
UG7.7% TZ
6.8%
MW11%
source: worldbank.org/lsms
Subjective DataFarmer assessment of soil quality included: overall quality, texture, color, and type
020
4060
Per
cent
FINE BETWEEN COARSE AND FINE COARSE
SR Good Quality SR Fair Quality SR Poor Quality
Farmer Identified Soil Texture & Quality
010
2030
4050
Per
cent
BLACK RED LIGHT
SR Good Quality SR Fair Quality SR Poor Quality
Farmer Identified Soil Color & Quality
16
Comparison: Subjective vs. Objective
13% of the top-soils have SOC < 2%
Of these, respondents only classified 6% as poor quality.
Comparison: Subjective vs. Objective
0.0
2.0
4.0
6D
ensi
ty
0 20 40 60 80 100CEC
SR Good Quality SR Fair Quality SR Poor Quality
kernel = epanechnikov, bandwidth = 2.2310
Cation Exchange Capacity
Comparison: Subjective vs. Objective
Comparison: Subjective vs. Objective
0.0
2.0
4.0
6.0
8D
ensi
ty
0 10 20 30 40Predicted % Sand
Reported as "very fine" or "fine"Reported as "between coarse and fine"Reported as "coarse" or "very coarse"
Respondent Assessment of Soil Texture
0.0
2.0
4.0
6.0
8D
ensi
ty
0 10 20 30 40Predicted % Sand
Reported as "sandy" soilReported as "clay" soilReported as "mixture of sand and clay" soil
Reported Soil Type
Scalability & Implementation Challenges• Fieldwork timeline implications:
– Avg 38 minutes per plot– Additional driving time for lab delivery– Consider crop-type and soil sample timing
• Logistics:– Sample labeling– In-country lab infrastructure
• Currently testing in-country MIR– Capacity/timeline at analytical lab
• Cost implications– Not cheap, but getting cheaper (advantage of
spectral analysis)
• Subjective soil data exhibits little variation
• Respondents (in this country context) often optimistic about state of the soils– Unclear on which
properties farmers are basing their assessment (and questionnaire design adjustments).
• Spectral soil analysis a feasible option to improve soil quality data (in the right context, and with the right budget).
Concluding Thoughts
Next Steps• Further analysis on optimal respondent for
subjective questions.– Also, improve questionnaire design by
identifying subjective questions with strongest correlation to objective measures.
• Does heterogeneity of soil properties in an EA affect the ability of respondents to assess their soil quality?
• Compare plot-level soil results with national soil maps (such as the Harmonized World Soil Database).
Questions?
Ermias Betemariam†, Calogero Carletto˚, Sydney Gourlay˚, Keith Shepherd†
† World Agroforestry Centre˚ The World Bank
29th International Conference of Agricultural EconomistsMilan, Italy – August 8-14, 2015
Collecting the Dirt on Soils: Advancements in Plot-Level Soil Testing and
Implications for Agricultural Statistics