Applications of soil spectroscopy on Land Health Surveillance

Applications of soil spectroscopy on Land Health Surveillance Ermias Betemariam Hands-on Soil Infrared Spectroscopy Training Course Getting the best out of light 11 – 14 November 2013

Context § There is a lack of coherent and rigorous sampling and assessment frameworks that enable comparison of data (i. e. meta-studies) across a wide range of environmental conditions and scales § Soil monitoring is expensive to maintain § Soil degradation and loss is a challenge § High spatial variability in soil properties- large data sets reduce uncertainty High spatial variability of SOC can rise sevenfold when scaling up from point sample to landscape scales, resulting in high uncertainties in calculations of SOC stocks. This hinders the ability to accurately measure changes in stocks at scales relevant to emissions trading schemes (Hobley and Willgoose, 2010) Soil spectroscopy key for Land Health Surveillance 1

Context (2) • Soil comes to the global agenda: – Sustainable intensification took soil as a x-cutting – Global Environmental Benefits - land degradation and soils are among the priority global benefits (GEF/UNCCD) • SOC as useful indicator of soil health • Importance of soil carbon in global carbon cycle and climate mitigation • Increasing demand for soil data at fine spatial resolution – carbon trading purposes requires high levels of measurement precision 2

Land Health (SD 4) Land Health - the capacity of land to sustain delivery of essential ecosystem services Land health surveillance aims to provide statistically valid estimates of land health problems, quantify key risk factors associated with land degradation, and target cost-effective interventions to reduce or reverse these risks. 3

Land Health Projects No. Name of project 1 Africa Soil Information Service (Af. SIS)/Africa Soils (SSA) 2 Strengthening capacity for diagnosis and management of soil micronutrient deficiencies (SSA) 3 Soil monitoring protocol for the World Bank Living Standards Measurement Study (Ethiopia &. . ) 4 Carbon sequestration options in pastoral & agro-pastoral systems in Africa (Burkina Faso & Ethiopia) 5 Land health surveillance for high value biocarbon development (Kenya, Burkina Faso & Sierra Leone) 6 Land health surveillance system for smallholder cocoa in Ivory Coast 7 Trees for food security in Eastern Africa (Rwanda, Ethiopia, Burundi & Uganda) 8 Land health surveillance for mitigation of climate change in agriculture (Kenya & Tanzania) 9 Land health surveillance system in support of Malawi food security project (Malawi) 10 Land health surveillance system for targeting agroforestry based interventions for sustainable land productivity in the western highlands of Cameroon 11 A Protocol for Measurement and Monitoring Soil Carbon Stocks in Agricultural Landscapes 4

Land Health out-scaling projects Global-Continental Monitoring Systems CRP 5 pan-tropical basins Regional Information Systems Tibetan Plateau/ Mekong Evergreen Ag / Horn of Africa Af. SIS National surveillance systems Ethio. SIS- Ethiopia Project baselines SLM Cameroon Parklands Malawi Rangelands E/W Africa Cocoa - CDI MICCA E. Africa 5

Af. SIS: Soil functional properties Af. SIS ✓ 60 primary sentinel sites ➡ ➡ ➡ 9, 600 sampling plots 19, 200 “standard” soil samples ~ 38, 000 soil spectra 6

Af. SIS: Soil functional properties Spectral diagnostics tools can be used to produce soil maps Prediction map for soil organic carbon for sub-Saharan Africa. (Source: Africa Soil Information Service) 7

Af. SIS: Soil functional properties (1) Af. SIS ✓ 60 primary sentinel sites ➡ ➡ ➡ 9, 600 sampling plots 19, 200 “standard” soil samples ~ 38, 000 soil spectra Ethio. SIS 97 Sentinel sites 8

Af. SIS: Soil functional properties (1) From polygon-based to probabilistic mapping Probability topsoil p. H < 5. 5. . . very acid soils Probability of observing cultivation + Current lime requirement ? ~ min [prob(p. H < 5. 5), prob(cult)] = Grid-based probabilistic maps increases the reliability of the map and its power to be combined with other data sources (remote sensing & terrain data) Taxonomic soil classification systems provide little information on soil functionality in particular the productivity function (Mueller et al 2010) (Walsh, 2013) 9

Living Standards Measurement Study-LSMS-IMS (3) Improve measurements of agricultural productivity through methodological validation and research Low cost MIR soil testing for smallholder farmers Mobile phones for quick soil screening- being tested 10

Carbon sequestration in pastoral & agro-pastoral systems (4) Effects of range management on soil organic carbon stocks in savanna ecosystems of Burkina Faso & Ethiopia Fire (controlled burning -19 years) – Burkina Faso Fire influence: • Carbon allocation - SOC gain • Decrease input - SOC loss Grazing (Exclosures 12 - 36 years) – Ethiopia 11

Results No Sig difference in SOC between burned and unburned plots 12

Results (2) No Sig difference in SOC between burned and unburned plots 13

Results (4) No sig. difference in SOC between closed and open plots for all age categories 14

Biocarbon development in East and West Africa (5) Challenges in cocoa production • Develop effective and cost efficient carbon monitoring, reporting and verification systems that can enable smallholders to access carbon markets • Soil spectroscopy will be key component Estimating biocarbon using Li. DAR data- Taita, Kenya (a) indigenous forest, (b) mixed stand of local and exotic species (Eucalyptus sp. ) and (c) cropland with scattered trees Janne et al. , 2013 15

Smallholder cocoa in Ivory Coast-V 4 C (6) LDSF and soil spectroscopy to identify constraints & target interventions in cocoa production Major challenges Disease + pest Soil fertility 16

Trees for food security –ACIAR Characterize land health constraints and assessing Agroforestry intervention outcomes Rwanda Ethiopia 17

Mitigating Climate Change in Agriculture-MICCA (8) Characterize (baseline) and assess impacts of climate smart agriculture practices East African Dairy Development (EADD- Kenya) Conservation agriculture (CARE- Tanzania) 18

Measurement and Monitoring Soil Carbon Stock (11) Can we measure soil carbon cost effectively? 19

Land Health Surveillance Sentinel sites Randomized sampling schemes Consistent field protocol Prevalence, Risk factors, Digital mapping Coupling with remote sensing Soil spectroscopy 20

Measurement and Monitoring Soil Carbon Stock (11) 1 2 3 Why measure carbon? What will the protocol deliver? How much will it cost? 4 Sampling 5 Field work 6 Lab work 7 Data analysis 8 Presenting results 2121

Measurement and Monitoring Soil Carbon Stock (11) Web and excel based tool Sample size determination Sample allocation Soil Carbon stock Moisture content Error …. and reporting DATA INFORMATION KNOWLEDGE WISDOM 22

Monitoring SOC stocks Why cumulative soil mass? Bulk density as confounding variable in comparing SOC stocks Think mass not depth A management that leads to a DECREASE in bulk density will UNDER ESTIMATES SOC stocks & vice versa C conc. (%) Depth(cm) Bulk density (g/cm) SOC stock (Mg/ha) Error 1. 5 150 1. 2 270 1. 5 150 1 225 -16. 67% (Ellert and Bettany, 1995) 23

Cost –error analysis Comparisons of costs of measuring SOC using a commercial lab and NIR spectroscopy Thermal oxidation Sample preparation Soil sampling 4000 Cost per sample (USD) Cost (USD) 6000 2000 0 10 50 100 150 200 250 Number of samples Cost (USD) 8000 6000 Thermal oxidation Sample preparation Soil sampling 12 9 6 3 0 Personnel Others Thermal oxidation NIR spectroscopy Cost IR is cheaper (~ 56%) than dry combustion method for large number of samples 4000 2000 0 -500 15 NIR spectroscopy 500 Number of samples 1500 Throughput Combustion ~ 30 -60 samples/day NIR ~ 350 samples/day MIR ~ 1000/day 24

Cost –error analysis 10. 00 Half 95% confidence interval (t C ha-1) Costs of measurement often exceed the benefits – soil spectroscopy address this challenge 8. 00 6. 00 4. 00 2. 00 0 200 400 600 800 Number of samples 1000 8. 00 6. 00 4. 00 2. 00 0 5000 10000 15000 20000 Cost of carbon measurement (USD) 25

Sources of uncertainty Activity Sampling Sources of uncertainty Sampling design (random, stratified random) Sample size Natural variability (spatial) Sample preparation (e. g. contamination, subsampling) SOC measurement Lab method used (instrument resolution) Human error Field data collection (e. g. soil mass, vol) SOC prediction using IR Imported uncertainties (from reference data) Model (assumption) Instrument and human errors Covariates used Mapping SOC Image pre -processing (geometric and radiometric corrections) Scale/resolution (e. g. farm vs landscape) Model (assumption, strength) 26

Common causes of measurement uncertainty – – – the instruments used, the item being measured, the environment, the operator, other sources Measurements can be precise (repeatable) but inaccurate (off-the mark) G. W. Sileshi, 2013 27

Things to be careful! Lets do it right Avoid contamination Proper labeling 28

Data archiving/publishing Datasaving – dataverse: http: //thedata. harvard. edu/dvn/ 29

Finally… • More research on cost-effective measurement tools • Web services are needed that allow optimised soil information to be automatically exchanged via the internet • Proximal soil sensing • Reduce uncertainties in measurements- error propagates • Develop national capacities, networking and partnership • Baselines are established for important soil properties across Africa • Soil spectroscopy filling the data gaps- at National, Regional & Global levels • Enable decision makers have clear understanding of soil status and trends • Spectroscopy is proved good- adoption and application • Cross sentinel/regional sites analysis 30

Thank you 31