INTEGRATING SOIL HEALTH MONITORING WITH HOUSEHOLD SURVEYS TO IMPROVE AGRICULTURAL STATISTICS

Transcription

1 S O I L S A M P L I N G I N H O U S E H O L D S U R V E Y S : E X P E R I E N C E F R O M E T H I O P I A INTEGRATING SOIL HEALTH MONITORING WITH HOUSEHOLD SURVEYS TO IMPROVE AGRICULTURAL STATISTICS ERMIAS AYNEKULU & KEITH SHEPHERD CALOGERO CARLETTO & SYDNEY GOURLAY WORLD AGROFORESTRY CENTRE WORLD BANK

2 ACKNOWLEDGMENTS: The Ethiopia Land and Soil Experimental Research (LASER) study was made possible by generous funding from UK Aid. The authors would like to thank the Central Statistical Agency of Ethiopia and the extraordinary field team for their dedication to the successful implementation of the project. The authors would also like to thank Asmelash Tsegay and Alemayehu Ambel for their valuable contributions to LASER fieldwork preparation and execution. RECOMMENDED CITATION: Aynekulu, E., Carletto, C., Gourlay, S., Shepherd, K. (2016). Soil Sampling in Household Surveys: Experience from Ethiopia. The World Bank, Washington, D.C., and World Agroforestry Centre, Nairobi. Published by The World Bank and World Agroforestry Centre. Copyright The World Bank and World Agroforestry Centre To obtain permission to republish, contact or The findings, interpretations, and conclusions expressed in this guidebook are entirely those of the authors. They do not necessarily represent the views of the World Bank, the World Agroforestry Centre, or their affiliated organizations. Version: August 10,

3 TABLE OF CONTENTS Abbreviations and Acronyms... 4 Introduction... 5 Part I Sampling Design... 7 Part II Soil Sampling From Agricultural Fields Preparation for Fieldwork Collecting Field Samples Plot Layout Soil Sampling Part III Soil Processing & Laboratory Analysis Soil Processing Equipment & Materials for Soil Processing Lab Lab Procedures Laboratory Analysis Part IV. Results from LASER Laboratory Results: Soil Properties in the Study Area Subjective Farmer Assessment Comparison of Objective and Subjective Measures Conclusions References Annex 1: Equipment Cost Estimation Annex 2: Subjective Soil Module

4 ABBREVIATIONS AND ACRONYMS AfSIS AgSS cm CSA EA FAO g GIS GPS ICRAF IR LASER LDSF LSMS LSMS- ISA MIR NIR TXRF UNEP XRD Africa Soil Information Service Central Statistical Agency of Ethiopia s Agricultural Sample Survey Centimeter Central Statistical Agency of Ethiopia Enumeration Area Statistical Food and Agriculture Organization Gram Geographic Information Systems Global Positioning System World Agroforestry Centre Infrared Spectroscopy Land and Soil Experimental Research study Land Degradation Surveillance Framework Living Standards Measurement Study Living Standards Measurement Study - Integrated Surveys on Agriculture Mid Infrared Diffuse Reflectance Spectroscopy Near Infrared Diffuse Reflectance Spectroscopy Total X- Ray Fluorescence Spectroscopy United Nations Environment Program X- Ray Powder Diffraction Spectroscopy 4

5 INTRODUCTION Renewed interest in increasing agriculture productivity to meet food security needs and increasing resilience of agricultural systems in developing countries, especially in sub- Saharan Africa, makes understanding soil fertility constraints and trends ever more important. Measurement and monitoring of soil health are fundamental to developing a sound knowledge of problems and solutions for sustainable crop production and land management (Aynekulu and Shepherd, 2015). Much of the current analyses on agricultural productivity is hampered by the lack of consistent, good quality data on soil health and how it is changing under past and current management. Direct systematic measurement of soil fertility as part of household level data collection has rarely been attempted due to the high costs of soil sampling and analysis. Linking soil health information to socio- economic household surveys provides an important opportunity for enhancing our understanding of trends in soil health and the impact on crop productivity among smallholders, as well as the coping mechanisms adopted by farmers faced with deteriorating soil conditions. Surveillance frameworks for systematic and consistent monitoring of soil and land health recently developed based on new digital sensing technology offer the potential to be linked with socio- economic surveys (Shepherd & Walsh, 2002, 2007). In particular, new rapid low cost technology for assessing soil characteristics using infrared spectroscopy has made soil fertility characterization feasible in large studies (Shepherd & Walsh, 2007). These techniques are now being supplemented by other light- based techniques using laser and x- ray spectroscopy and are being applied to large area sampling schemes in sub- Saharan Africa under the Africa Soil Information Service project (AfSIS). Soil health information is of interest to many different audiences for different purposes. The ideal soil health measurement scheme would be tuned to the exact purpose and context for which it is required. However, it is rarely feasible to develop and refine methods for every situation. Hence, we describe a 5

6 procedure based on that recently implemented in a methodological experiment in Ethiopia, which is suitable for systematic measurement of soil health as part of a household- level data collection operation. This procedure may be modified to fit specific study objectives and data needs, but should provide a general sense of survey design requirements and implementation considerations. The results of the study also ought to illustrate the benefit of including objective measurement of soil health in household surveys with an agricultural focus. The target users of this guide include survey practitioners, technical staff of relevant government agencies and research institutes, and field data enumerators who will implement field data collection. Building on the on- going Living Standards Measurement Study- Integrated Surveys on Agriculture (LSMS- ISA) effort and as part of a broader World Bank research agenda, the LSMS team implemented a multi- faceted research project aimed at improving the quality and relevance of agricultural statistics. The project agenda includes seven distinct components: (1) land area measurement, (2) soil fertility, (3) water resources, (4) labor inputs, (5) skill measurement, (6) production of continued and extended- harvest crops, and (7) computer- assisted personal interviewing for agricultural data, all supported by generous funding from UK Aid. This guide is based on the soil fertility component of the project where both the feasibility of integrating soil quality analysis into household socio- economic data collection operations and the local knowledge of farmers in assessing their soil quality were evaluated. The Land and Soil Experimental Research (LASER) study, which will serve as the foundation for this document, tested the integration of soil analysis in the household survey context in three agro- ecological zones in Ethiopia. The LASER study was implemented by the LSMS team of the World Bank in partnership with the Central Statistical Agency of Ethiopia (CSA) and the World Agroforestry Centre (ICRAF). Fieldwork consisted of the implementation of a variety of subjective farmer- estimated indicators of soil quality as well as conventional and spectral soil analysis, resulting in a unique plot- level dataset. The data collected in the 6

7 LASER study allows for the analysis of the impacts of relying on subjective farmer estimates of soil quality for policy- based decision- making through comparison of subjective and laboratory (also referred to as objective) measures of soil properties. Results from the methodological experiment data reveal the inability of smallholder farmers to clearly distinguish between soil fertility levels, which we hypothesize, may partially explain the slow adoption of improved agricultural practices and inputs often observed in Africa. This document uses the protocols, experiences, and lessons learned from the Ethiopia LASER study to provide a step- by- step guide to integrating soil collection and analysis in household surveys. Wherever possible, implementation steps and strategies are generalized to the fundamental requirements and complemented by the specific approach used in LASER. The soil sampling and measurements implemented in the LASER study, and in this guide, were derived from Africa Soil Information Service (AfSIS) protocols but adapted to fit sampling from agricultural fields. The remainder of this guide describes soil sampling, processing and analysis in the following sections: (I) Sampling Design, (II) Soil Sampling from Agricultural Fields, and (III) Soil Processing and Laboratory Analysis. Part IV provides key results from the LASER study. PART I SAMPLING DESIGN Sampling frameworks are designed to reduce sampling error, avoid biased selection of sampling sites, and guide where and how many samples to take. It is vital to be clear on the decision or sets of decisions that the measurement will support before designing the sampling framework, bearing in mind that the main purpose of measurement is to reduce decision uncertainty. Randomizing sites within the target area or sampling strata is important to provide unbiased estimates of soil properties within a stratum and allow inference to be made to the whole area. Providing unbiased data on the statistical distribution of 7

8 variables is not only useful for reporting the prevalence of land health problems (e.g., low soil carbon content) but also provides a means of setting local reference values (defining what is low, moderate or high), which can in turn be conditioned on various factors (e.g., soil texture). A small probability sample generally provides much more useful information than a large biased sample. The sampling design should be driven by the objectives of the study. Before designing the sample, it must be clear what level of representativeness you are striving for (if any) and the geographic area of interest. If the study is targeting specific crops and/or populations, that too must be made explicit. Once the objectives are clear, it is necessary to decide: a) How many households will be included in the sample? b) How many agricultural fields from each household will be subject to soil testing? c) How many soil samples will be collected from each field? d) Where will the soil samples be collected in the field? Points (a) and (b) are study- specific decisions and must be made with the objectives in mind. Points (c) and (d), however, are more generalizable. Depending on the expected variation in the soils, the size of the agricultural fields in the study area, and the method in which the sample collection sites are identified (point (d)), it is likely sufficient to test a single top- soil sample (0-20 cm depth) and a single sub- soil sample (20-50 cm depth) from each selected field. Soils provide different types of support to crops at different depths, and crops vary in their soil depth requirements. Maize, for instance, requires deeper soil than teff (Calviño et al. 2003; Evert et al., 2009). Thus, generally it is advisable to test both a top- and sub- soil samples but this should be reviewed in light of the crop of interest and overall study objectives. Point (d), the location of sample collection points 8

9 Soil sample Study zone Enumeration Area 1 (EA1) EA Household1 (HH1) HH Crop field 1 Crop field cm depth Soil sample cm depth Figure 1. Nested sampling procedure followed in Ethiopia LASER study. within the field is critical, and, fortunately, the most straightforward to implement. If the aim is to assess the soil quality of the field on the whole (and not a specific point within the field), the sample collection points must make up a representative sample. Therefore, it is strongly recommended to use composite samples. The strategy for collecting such composite samples, and addressing point (d) in general, is detailed in Part II. The objectives of the LASER study were multifaceted and included indicators related to soil properties, crop type, and socio- economic characteristics, among others. Additionally, the focus was on methodological validation rather than producing nationally representative statistics. Because there were multiple indicators, calculating the sample size based on the variance of a single indicator was not the preferred approach. Instead, a multi- stage nested approach using a practical allocation of enumeration areas (EAs) across agro- ecological zones was used (Figure 1). Given the methodological focus of the study, it was preferable to implement a smaller- scale survey operation that could be closely supervised. Yet, it was also imperative that the study area included substantial variation in soil properties such that the comparison of soil quality measurement methods 9

10 Figure 2. Location of the study area: markers indicate agricultural fields where soil samples were collected in Borena, East Wellega, and West Arsi zones of the Oromia region, Ethiopia. could be made in different contexts. Therefore, the study was limited to a single administrative region. Oromia region was selected because it represents a large area of Ethiopia and encompasses areas with variation in rainfall, elevation, and agro- ecological zones. 1 The sample was restricted further to three administrative zones of the Oromia region, which were selected based primarily on agroecology and geographic diversity. Secondary consideration was made for the availability of local soil research centers that could be used for soil processing. The three selected zones are East Wellega, West Arsi, and Borena. Using the CSAs Agricultural Sample Survey (AgSS) as the sampling frame, a total of 85 EAs were randomly selected using a practical allocation and implicit 1 According to FAO (1986), the three agro- ecologies are locally classified as Dega (highland altitude > 2300 m: West Arsi), Weyna Dega (middle altitude m: East Wellega), and Kolla (lower altitude < 1500 m: Borena). 2 For details on the practical allocation of EAs across agro- ecological zones, refer to the Basic Information Document 10

11 stratification of EAs across agro- ecological zones (Figure 2). 2 Finally, within each EA, 12 agricultural households were randomly selected from the AgSS household listing completed in September Thus, the process for determining point (a) above was complete. A total of 1020 households were selected for inclusion in the LASER study. The decision- making process for point (b), the number of fields subject to soil testing per household, was driven by the desire to allow for a comparison of subjective and objective measures of soil quality while controlling for household characteristics (therefore, requiring more than one field to be selected per household) but also the low expected variation of soil properties across fields cultivated by the same household (assuming geographic proximity). Up to two fields were measured per household. First, if any fields contained pure stand maize, one was randomly selected. Then, a second field was randomly selected from the remaining cultivated fields irrespective of crop type. If no fields contained pure stand maize, two fields were randomly selected. If the selected household cultivated only one plot, only that plot was subject to testing. The preference for maize in the first field selection was purely to satisfy the maize crop- cutting component of the study. The field selection was completed using a household- specific random number table to prevent any selection bias. In LASER, the determination of the number of samples per plot, point (c) above, was based on the existing AfSIS Land Degradation Surveillance Framework (LDSF). From each crop field, one composite sample was collected from the top- soil (0-20 cm depth) and one central sample from the sub- soil (20-50cm depth). Ideally, composite samples would be collected for both the top and sub- soils, rather than a single central sub- soil sample. If resources allow, a composite sample at each depth is recommended. 2 For details on the practical allocation of EAs across agro- ecological zones, refer to the Basic Information Document for the LASER study, found on the LSMS website (direct link: 11

12 The process of physically collecting the soil samples, as well as the identification of where the samples should be collected, point (d), are discussed in detail in the following section. The sampling strategy above is simply one example of an approach to soil sampling and analysis in household surveys. The approach to sampling may vary with the objective of the survey, the sampling frame, and resource availability, among other factors. The sample design must be such that it leads to an unbiased sample of households and their agricultural lands, and achieves the desired level of representativeness. 3 Depending on the scope of the study and/or the homogeneity of soils in the study area, one may collect soil samples from one or more agricultural fields per household. Previous studies, such as the digital soil map of Africa made available under the AfSIS project (africasoils.net), could be used to estimate soil variability in the study area. Similarly, the degree of clustering may vary (i.e., number of households selected per EA). Survey practitioners are encouraged to engage with a sampling expert in the survey design phase. PART II SOIL SAMPLING FROM AGRICULTURAL FIELDS The importance of having a consistent field measurement protocol that can be applied under all expected conditions cannot be over- emphasized. After the sampling design has been completed and it is clear how many households will be visited, how many parcels or fields will be selected for soil sampling, and how many soil samples will be collected from each selected parcel or field, you can prepare for fieldwork implementation. 3 The Ethiopia LASER study was not designed for representativeness of households or soils. Rather, due to the methodological nature of the study, the sample design was simply aimed at testing the various methodologies under a variety of soil conditions. 12

13 2.1 PREPARATION FOR FIELDWORK Proper preparation is critical to ensure a successful soil sampling campaign, and for the well- being of the field team. Prior to fieldwork, it is important to have a good understanding of the area to be surveyed, including its topography, climate and vegetation characteristics, accessibility, and security situation. It is important to consider the following points before commencing fieldwork: Collate existing information about the area to be surveyed including maps (topographical, geological, soils and/or vegetation), satellite images and/or historical aerial photographs. Confirm you have the necessary tools required for soil sample collection (see below). Train the field teams (enumerators) and pilot all procedures. Prepare logistics in terms of transport, local guides, interpreters, and accommodation (if needed). Inform local government officers and community leaders about your activities and obtain permission from the land owner(s) to sample a given area, making sure that that all parties understand what you are doing. For most enumerators, soil sampling will be a new concept. It also requires much more physical exertion than enumerating a typical household survey. This fact should not be underestimated when planning fieldwork timelines as teams may not be able to complete as many households in a single day as they would in standard household survey operations. Extra days should also be planned for training. On average, LASER enumerators spent approximately 40 minutes per field collecting soil samples. Enumerator training must be thorough and with a serious tone regarding the careful implementation of soil sampling and handling. Soil samples are quite fragile in that the testing results can be biased if the samples are not handled properly. Contamination of soil samples (as a result of inadequately cleaned equipment, for example) or failure to dry the soils within 5-10 days of collection can lead to misrepresented results. Training should include a theoretical overview of the motivation for the study (field teams will need to explain this to respondents), review of key concepts and terminology, review of the enumerator manual (which can be based off this document), and group and individual practicals. Training for the LASER study was 16 days total, 3 days of which were dedicated to soil training, plus a day of field practice with the entire survey (including soil sample collection). In LASER, training of soil activities was conducted by an expert from ICRAF to ensure best practices were utilized. While hands- on training 13

14 was the most effective, it was imperative that the strategy, terminology, and importance of each step were reviewed in a classroom setting prior to equipment distribution and outdoor demonstration. All equipment must be procured prior to enumerator training, as they should be used in field practice. The list of equipment is provided below: Basic tools required for field sample collection (see Figure 3) Cost estimates for key equipment are provided in Annex Table 1. GPS device (a) Soil auger marked at 20 cm and 50 cm from tip (b) o Auger heads are constructed differently for different types of soil. Sandy soils require a different auger head than clay soils. Distribute materials to field teams accordingly. Measuring tape (c) Plastic sample bags (d): composite sample field (0-20 cm) = 2 bags; cm = 2 bags; 4 bags total per parcel/field Buckets (e) Mixing trowel / spade (f) A board, tray or plastic sheet for coning and quartering. A clean plastic bag can also suffice. Labels or barcodes Permanent marker pens Stationary like pencils and a field notebook. Water bottle filled with water (with hose or punch a small hole in the cap) (g) 14

15 Figure 3. Some of the tools required for field data collection 2.2 COLLECTING FIELD SAMPLES Soil sampling on agricultural fields requires that enumerators enter and use an auger in the field. Thus, there is a risk for a small area of crop damage in the area of sampling. 4 If possible, try to collect soil samples without causing much damage to the crop by timing the sample collection appropriately. Stable soil properties will not be affected by the time of sample collection. For some crops, such as maize, the sampling is possible earlier in the agricultural season because samples can be taken from between plants. For other crops such as teff, it may be necessary to wait until the harvest is completed in order to avoid crop damage. It is also advisable to avoid soil sampling when the soil is too wet as it will be difficult and may require a lot of time to collect and dry. 4 Originally, the intention was to conduct bulk density analysis as well as the conventional and spectral soil testing as part of LASER. However, the bulk density test requires a metal plate (approx. 40 cm x 40 cm) to be laid in the field. During fieldwork training it was determined that the bulk density plate caused undue damage to the crop and therefore was excluded from the study. 15

16 2.2.1 PLOT LAYOUT As mentioned above, the location of the soil sample collection points are derived from the AfSIS LDSF 1000 m 2 design (Figure 4), placed over the center of the crop field. The AfSIS LDSF design serves as an initial template for soil sampling but may need to be modified to accommodate agricultural fields that are too small or irregularly shaped to contain a 1000 m 2 soil sample plot (see Fig. 5). The protocol implemented in LASER, and that which is recommended in future studies, is illustrated below: Step 1: Define the field center point by taking halfway from the longest side and halfway from the shortest side of the field (Figure 4). Record the GPS coordinates of the plot center, or save the waypoints in the GPS unit with the appropriate label. Step 2: Using a measuring tape, measure out the distance (12.2 m) from the center point (point 1; Figure 4) directly uphill (where applicable) to the second point (point 2). Crop field Figure 4. Sample plot layout in a crop field, with four points. The distance between the center point and the other three points is 12.2 m which represents a 1000 m 2 sampling plot. 16

17 Step 3: Mark points 3 and 4 at 120 and 240 degrees from the up- point (point 2), respectively, and 12.2 meters from the center point. The angles can be measured using a compass or estimated once the field team develops experience in plot layout. The distance should be corrected when plots fall on steep terrain (slope > 10 degrees). Use the following formula to calculate the distance from the center point to the other point (for slope > 10 degrees): horizontal distance Slope distance = cos (slope) Take the cosine values from the backside of your clinometer if you have one or you may prepare and print a cosine table to carry in the field. You can also get cosine values from the Internet if you have Internet connectivity in the field. For instance, for a field with 15 degrees slope the slope distance will be 12.2 m/0.9659= 12.6 m. In many contexts, agricultural fields can be quite small and/or irregularly shaped. In these circumstances, the AfSIS LDSF design (12.2m radius) may not fit within the field boundaries. If the field is too small, the point should be placed 2 meters in from the boundary in order to limit border/edge effects of other land uses (e.g. roads) on the crop field. It may also be the case that fields are irregularly shaped and cannot accommodate the LDSF design. Figure 5 provides examples of alternative sample collection points aimed at achieving soil samples representative of the full field. 17

18 c b a d (a) (b) (c) (d) Figure 5. Crop fields will have different shapes and sizes that may require the sample collection layout to be modified. Examples: (a) shape and size of the crop field allows a 1000 m 2 plot, take a composite sample from four points, (b) field crop less than 100 m 2, take sample from one point only, (c) shape of the crop field does not allow an LDSF type of layout, take a composite sample from 3 points, (d) one side of the field does not allow a full LDSF type of layout, leave a 2 m buffer from the boundary for point 2 (place it closer to center point) and take a composite sample from 4 points. 18

19 2.2.2 SOIL SAMPLING Soil samples are collected from two depth ranges (0-20 cm and cm). Every field selected for soil sampling is subject to the following: a) One composite topsoil sample collected at 0-20 cm. This sample is collected from 4 points within the field (points 1 4 in Figure 4). b) One composite subsoil sample collected at cm. This sample is collected from 4 points within the field (points 1 4 in Figure 4). The in- field sampling procedure is as follows: 1) Mark the auger at 20 and 50 cm (e.g. using a permanent marker). 2) Place the auger in the center sampling point and begin to auger straight down, using the same auger for all depths. If your augering becomes crooked, stop and start a new hole, otherwise you will get an inaccurate measurement of the depth. 3) Auger down to 20 cm depth and transfer all of the soil from the auger into the bucket designated for top soil. 4) The next sample is from cm. Continue augering further from the 0-20 cm (in the same hole) until you reach the 50 cm mark on the auger. Empty the soil from the auger into the separate bucket designated for subsoil. 5) Repeat step 3 for the remaining collection points on the field (points 2-4 on Figures 4 & 5). Place all top and soils collected into the bucket designated for each depth. Now you will have two buckets that contain composite samples from all four collections points for the top and sub soils. Mix each bucket with a trowel, cleaning the trowel between buckets. 6) If the auger was not able to reach the desired depth (due to rocks, for example), record the auger depth restriction (the maximum depth attainable). 7) Conduct the field soil texture analysis (see Figure 7 and 8). 19

20 8) Conduct the coning and quartering procedure for the topsoil and subsoil separately. The end result should be approximately 400 g of topsoil and 400 g of subsoil. Instructions for the procedure are found in the section below (Figure 6). 9) Put the samples in a bag and label the samples with the household identification, field identification, depth of the soil sample, and auger depth restriction (if applicable). If the soil is very dry it may be difficult to auger and collect all of the soil from the depth increment, in which case pre- wetting the soil before augering each increment may be helpful. If you wet the soil be sure to double bag the sample and do not place the sample label tag in the bag with the wet soil, as the tag will stick to the soil and make it difficult for processing in the lab. 20

21 CONING AND QUARTERING PROCEDURE Normally, soil sampling procedures lead to collection of more soil than necessary for analysis. To get a representative sample of about 400 g, a standardized and consistent procedure must be followed. For this aim, it is necessary to use the method of coning and quartering described below. Continue the coning and quartering technique on the topsoil and subsoil samples to obtain a representative 400 g subsample of soil for further soil processing and laboratory analysis. Step 1: Place the sample on a strong, clean plastic sheet or similar material (Figure 6b). Step 2: Thoroughly mix the soil sample and spread the samples into a conical pile. Step 3: Further mix the soil by circumventing the cone symmetrically, repeatedly taking a spatula- full of soil from the base and transferring the soil to the apex of the cone. Step 4: Ensure the spatula is large enough to reach to centre of the cone. Circumvent the cone twice. Step 5: Flatten the cone to a height of about 1 cm (Figure 6b). Step 6: Use a flat spatula or ruler, divide the pile into quarters with two perpendicular lines (Figure 6b). Step 7: Select one pair of opposite quarters as the sample to be retained. Step 8: If the sample is still too large then repeat the procedure from the beginning. Step 9: Take the representative subsample (about 400 g) and place it in a plastic bag. Label the bag and place an extra label tag inside the bag. 21

22 (a) (b) Figure 6. (a) soil sample collection, (b) soil sub- sampling via coning and quartering. FIELD TEXTURE ANALYSIS Soil texture is the amount of sand, silt and clay in the soil and is important for determining many soil properties including aggregation, structure, and water and air movement through the soil. This section provides instruction on determining if the soil texture is smooth, gritty, or neither smooth nor gritty. Refer to the steps in Figure 7 and the flow chart in Figure 8 for determining the soil texture by feel. 22

23 Texture by feel on soil samples (a) Step 1: Starting with the top soil sample, moisten a handful of soil using water from the water bottle until the soil has a putty- like consistency, but free water does not escape when ball is squeezed. Step 2: Shape the soil into a ball. If the ball retains its shape, move to step 3. (b) Step 3: Using your thumb and forefinger, form a ribbon with the soil by smearing the ribbon with your thumb. Observe if the soil is shiny or dull. Step 4: Measure the length of the ribbon when it breaks and record it in the questionnaire. The ribbon should be measured in mm. (c) Step 5: Classify the texture according to the flow chart (Figure 8) and report in the questionnaire whether it is smooth, gritty, or neither. Figure 7. Soil texture by feel steps 23

24 24 Figure 8: Soil texture by feel flow chart. Soil%Texture%By%Feel%Flow%Chart Place%approximately%two%teaspoons%of soil%in%your%palm.%%add%a%few%drops%of water%and%kneed%soil%to%break%down all%the%aggregates%%soil%is%at%proper consistency%when%it%feels%plastic%and moldable,%like%moist%putty. Add%dry%soil%to soak%up%water Start Does%the%soil remain%in%a%ball when%squeezed? Is%the%soil%too dry? Is%the%soil%too wet? No Yes No Sand No Place%ball%of%soil%between%thumb%and%forefinger,%gently%pushing%the%soil with%your%thumb,%squeezing%it%upward%into%a%ribbon.%%form%a%ribbon%of uniform%thickness%and%width.%%allow%the%ribbon%to%emerge%and%extend over%forefinger,%breaking%from%its%own%weight.%%does%the%soil%form%a ribbon? Yes Yes Loamy Sand No Does%soil%make%a weak%ribbon%<%1" long%before%it breaks? Does%soil%make%a medium%ribbon 1M2"%long%before%it breaks? Does%soil%make%a strong%ribbon%>%2" long%before%it breaks? No Yes No Does%soil feel%very gritty? Neither gritty%nor smooth? Does%soil feel%very smooth? Sandy Loam Loam Silt Loam Yes Yes Yes No No Does%soil feel%very gritty? Neither gritty%nor smooth? Does%soil feel%very smooth? Sandy Clay Loam Clay Loam Silty Clay Loam Yes Yes Yes No No Does%soil feel%very gritty? Neither gritty%nor smooth? Does%soil feel%very smooth? Sandy Clay Clay Silty Clay Yes Yes Yes No No Excessively*wet*a*small*pinch*of*soil*in*your*palm*and*rub*it*with*your*forefinger. %%CLAY % S A N D HI LO HI

25 SOIL SAMPLE LABELLING Labelling is critical. In order to ensure that results from soil analysis are linked to the correct household and field, several pieces of information must be captured. The unique household identifier, the parcel and/or field identifier, and the depth of the soil sample must be listed on the sample bag (with a duplicate label place inside the bag in case of damage to the outer label and for use by the laboratory). The labelling scheme using in LASER is illustrated in Table 1 below. Table 1. Examples of Soil Sample Labels used in LASER Labelling for soil samples Field No samples Remark HHHH PP FF 0-20 cm 1 HHHH = Household; PP = Parcel; FF = Field HHHH PP FF cm 1 Total 2 For example, for a field with Household ID 3108, parcel ID 03, and field ID 02, you would have the following labels: cm (for the top soil) cm (for the sub soil) In case of soil depth restrictions, indicate the actual depth in the label. For example, the label for a site where a soil depth restriction occurred at 35 cm depth would be the following: cm 25

26 Sample IDs should be legibly recorded with a permanent marker on the outside of the plastic bags. A paper label containing the same information (written in pen or pencil) should also be placed inside the bags. When possible, the use of barcoded labels is highly encouraged, especially for surveys administered via Computer Assisted Personal Interviewing (CAPI) 5. Pre- ordered barcodes with a serialized numbers printed on them can greatly reduce data entry problems. A duplicate of every barcode should be ordered such that one barcode is placed on the outside bag of the soil samples and the second is placed inside the bag (without the backing peeled so it is not sticky). 6 The duplicate bag can then be easily used by the laboratory when they take only a subsample of the soil. Ensure that the type of barcode you choose can be read by the CAPI tablets. SOIL SAMPLE SUMMARY The process of collecting soil samples from agricultural fields is a multi- stage process, but is not overly technical. Enumerators can be trained to effectively implement the modified AfSIS LDSF sample layout, collect soils via augering, sub- sample the collected subsoils via coning and quartering, test the texture of the soils, and properly label the samples. A summary of the steps in soil sample collection are found in Table 2. 5 Barcoded labels were used in a follow- up study to LASER conducted in Uganda and proved to ease fieldwork and reduce data entry problems related to labelling. 6 If using barcodes, auger depth restrictions can be captured in a separate question on the tablet. 26

27 Table 2. Major steps in soil sample collection Step Top soil Crop field Sub soil Step 1: Using the soil auger, collect top soil (0-20 cm) from the four points and place it in a bucket (Figure 6a). Step 2: Using the soil auger, collect sub soil (20-50 cm) from four points and place it in a bucket (Figure 6a). Step 3: Analyze the texture and measure the ribbon size for the top and sub soils according to the instruction in Figures 7 and 8. Step 4: Take a representative (about 400 g) subsample of the top and sub soils following the coning and quartering procedure and place them in separate plastic bags. Label the bags accordingly (as in Table 1 or otherwise determined by the project). Place an extra label inside the sample bag. Step 5: If there are auger depth restrictions, adjust the label accordingly. For example, if the auger will only dig to 15 cm, label the sample as HHID- parcel ID Field ID 0-15 cm (rather than 0-20 cm). (0-20 cm) * * * * (20-50 cm) * * * * PART III SOIL PROCESSING & LABORATORY ANALYSIS Part III describes soil processing and analysis. While much of the content is directly relevant to the laboratories where drying, processing and analysis take place, it is essential for survey practitioners to fully understand the process. 3.1 SOIL PROCESSING All soil samples should be transported to the laboratory where analysis will be conducted (whether it be a local, regional, or international lab) using a standard operating procedure. Prior to transportation of the samples to these laboratories, especially if distant from the study site, samples must be dried and 27

28 processed. If laboratory analysis will be conducted out of the country, such as at a regional laboratory, it is recommended that soil processing be conducted in- country prior to shipment of the samples. This will help to avoid changes in soil properties from storage of wet samples and reduce the costs of shipping a large volume of soils. Soil laboratories found in national or regional agricultural research institutes and academic institutions located near the study area could be used for soil processing. In LASER, the soil laboratory facilities and technicians from Awassa Agricultural Research Center (for Western Arsi Zone), Ambo University (for Wellega zone), and Yabello Pastoral and Dryland Agricultural Research Center (for Borena Zone) were utilized. In the case of LASER, samples were dried, sieved, and weighed prior to the shipment of samples to ICRAF for analysis. Note that the drying of samples is critical and must be completed within a week of collection in order to minimize any decomposition of organic material. Soil collected from an agricultural field is a mix of gravel (fraction of the soil with a diameter of > 2 mm), sand (fraction between 0.05 mm and 2.0 mm), silt (fraction between mm and 0.05 mm) and clay (fraction smaller than mm or 2 μm). Since the contribution of gravel fraction to the dynamics of soil nutrient cycles and supply to plants is minimal, they are excluded from the analysis and reporting of plant available nutrients. During the preparation of soil samples, this fraction should not be crushed and pulverized, because then this fraction is considered part of the fine soil and will results in low values due to dilution effect. However, aggregates of >2 mm size formed by fine particles have to be crushed in order to pass through a 2 mm sieve and included in the analysis. Normally, soil sampling procedures lead to collection of more than required amount of soil and to get a representative sample of the necessary volume, there is a need to follow standardized and consistent procedures. For this aim it is necessary to use the method of coning and quartering to reach the required volume of soil. 28

29 3.1.1 EQUIPMENT & MATERIALS FOR SOIL PROCESSING LAB Soil processing does not require technical equipment. It requires only the items listed below and a clean working environment with space for drying the soil samples. Drying trays Wooden rolling pins Sieve 2mm Wooden Rolling pin Plastic sheet Markers Brown paper bags (size 5 recommended) Weighing / balance scale Plastic zip- lock bags Particulate respirator nitrile gloves (eg. N95 of 3M brand) Damp cloth Forced air drying oven (optional) LAB PROCEDURES Soil processing involves sample reception, drying, crush and sieving, and sub- sampling of fine particles, all while ensuring the health and safety of the lab technicians. Each of these processes is discussed below. SOIL SAMPLE RECEPTION Layout the samples received in order of labeling and check against the master list of samples received from the field. If barcoded labels are used, the lab can simply scan the inventory. Make detailed notes in a laboratory record book of any labeling discrepancies or problems due to damaged sample bags or lost samples. 29

30 DRYING Spread the soil out as a thin layer into shallow trays or plastic or paper sheets. It is important to ensure that no material from a sample is lost or discarded (Figure 9). Air dry the samples in shade. Drying can also be done in a large room, custom- made solar dryer, or a forced- air oven at 40 C. Drying time will depend on the condition of the samples and ambient conditions, but the samples must be thoroughly dried. Break up clods as far as possible to aid drying. Take care to avoid crushing gravel- sized particles. Great care should be taken at all stages to ensure sample labels remain with the samples. Exercise care to avoid contamination from dust, plaster or other potential contaminants during drying, as soils are subject to trace element analysis. Figure 9. Sample drying CRUSHING AND SIEVING Spread the sample onto a plastic sheet on a solid table (Figure 10). Using a wooden rolling pin, crush the sample to pass through a 2 mm. While crushing, remove any plant materials (e.g. roots) and any possible pieces of gravel (making sure they are gravel and not soil aggregates) and place in a separate pile (the coarse fraction). Pass the crushed sample through the 2 mm sieve. DO NOT use the sieve as a grinder; i.e. do not rub or mash the soil on the sieve, but shake the sieve gently to allow the soil to pass through. If a large amount of soil needs to be sieved, it is easier to do it in small batches rather than all at one time. 30

31 Place whatever remains on the sieve back onto the plastic sheet and crush again gently. Then pass again through the 2 mm sieve. Make sure that all soil materials are crushed. Transfer anything that now remains on the sieve into the coarse fraction pile. The whole sample should be processed. You will remain with two fractions: Figure 10. Sample crushing and sieving o The coarse fraction (> 2 mm), which cannot pass through the sieve. Discard the course fractions. o The soil fines (< 2 mm), which have passed through the sieve. Clean off the plastic sheet with a damp cloth to remove soil dust, so as to prevent contamination from one sample to another. SUB- SAMPLING OF FINE FRACTIONS If the weight of the soil fines is more than 400 g, subsample the soil fines using coning and quartering to give about g of fine soil. Continue the coning and quartering technique on all samples to obtain a representative 20 g subsample of soil fines for shipping to ICRAF Soil- Plant Spectral Diagnostic Lab in Nairobi or other laboratory for specialized spectral analyses. Place the 20 g subsample in a zip- lock polythene bag labeled properly. All of these 20 g samples will be delivered to the lab conducting the spectral analyses. Place the remaining +/- 350 g sample of dry soil fines into a strong size 5 brown paper bag, labeled properly, and store it at ambient temperature for possible use in future. Note, this is where having a duplicate barcode or other label (with the same identification information) comes in to play. A select subset of the 350 g soil fine samples will be tested with conventional analysis. This is typically in the range of 10-25% of all samples. The exact samples will be identified following completion of the spectral analysis on all samples, thereby gaining an understanding of the 31

32 variation of soil properties present in the samples. The selected 350 g reference samples will be shipped to ICRAF Soil- Plant Spectral Diagnostic Lab in Nairobi or to regional lab for reference analyses. Soil fines samples not shipped should be stored at the regional laboratory in case they are needed for future analyses. HEALTH AND SAFETY Wear nitrile gloves to reduce the incidence of skin contact with potentially contaminated soil and to reduce the risk of cross- contamination. Wear a respirator that covers the mouth and nose to filter out harmful dust particles. Inhaling such particles irritates the nostrils and sinuses and can lead to lung diseases. Refer to the site- specific Health and Safety Plan for other safety concerns and applicable personal protective equipment. 3.2 LABORATORY ANALYSIS Soils samples can be analysed for chemical (e.g. carbon and nitrogen) and physical properties (e.g. texture) at the soil- plant spectral lab of the World Agroforestry Centre (Figure 11) or other laboratories with the necessary equipment and expertise. All soil samples should be scanned for mid- infrared soil spectroscopy (IR) and 10-25% from the total sample for reference analysis which is used to calibrate and validate MIR- spectral prediction for the remaining samples (75-90%). (a) (b) Figure 11. (a) MIR spectrometer, (b) a team of three staff from the LSMS project of the World Bank and the Central Statistical Agency (CSA) of Ethiopia attended a three day (19-21 February 2014) soil infrared spectroscopy exposure training at ICRAF s Soil- Plant Spectral Diagnostics Laboratory in Nairobi. 32

33 Infrared spectroscopy is now routinely used for analyses of a wide range of materials in laboratory and process control applications in agriculture, food and feed technology, geology and biomedicine (Shepherd and Walsh, 2007). The mid infrared (MIR, μm) wavelength region was investigated for non- destructive analyses of soils and can potentially be usefully applied to predict a number of important soil properties, including: soil colour, mineral composition, organic matter and water content (hydration, hygroscopic, and free pore water), iron form and amount, carbonates, soluble salts, and aggregate and particle size distribution (Shepherd and Walsh, 2004). Importantly, these properties also largely determine the capacity of soils to perform various production, environmental and engineering functions. IR enables soil- sampling density (samples per unit area) to be greatly increased with little increase in analytical costs. At ICRAF labs, the reference samples are analysed for soil organic carbon concentration and nitrogen using thermal oxidation method (Skjemstad and Baldock, 2008). To avoid the influence of inorganic carbon (carbonate) soil organic carbon is determined on acidified samples, i.e. fumigated with hydrochloric acid to remove inorganic carbon (Harris et al., 2001). Soil texture is analysed using laser diffraction particle size analysis method. For LASER, two objective measures were employed by ICRAF laboratories. Conventional soil analysis (CSA), which included traditional wet chemistry methods for soil nutrient extraction and some basic soil physical analyses, was conducted on 10% of samples (n=361). Conventional analysis, while often regarded as the gold standard in soil analysis, is expensive and destructive in nature. Spectral soil analysis, the second set of tests conducted under the LASER study, is significantly less expensive and non- destructive, allowing for multiple tests over time. This included the following tests: mid- infrared diffuse reflectance spectroscopy (MIR), laser diffraction particle size distribution analysis (LDPSA), x- ray methods for soil mineralogy (XRD), and total element analysis (TXRF). MIR and LDPSA spectral tests were conducted on all top- and sub- soil samples (n=3611), while the x- ray tests, XRD and TXRF, were conducted on the same 10% in which conventional testing was executed. For details on the predictive power of the mid- infrared spectroscopy on multiple soil properties, see Aynekulu et al. (forthcoming). Results from the conventional and spectral soil testing are presented to the end user in the form of a dataset, complete with all identification variables (which are necessary to merge the data with the household survey) and soil quality indicators included in the testing. Whenever possible, it is strongly encouraged to make the data publicly available, as it will increase the visibility of the data by making it available to researchers, students, and policy makers. However, household surveys require a degree of confidentiality. Exact coordinates of households, agricultural fields, or soil collection 33

34 points must not be released (unless all respondents agree otherwise). Dataverse ( or other server could be used to archive datasets, depending on the security requirements. PART IV. RESULTS FROM LASER Key results from the LASER study are reported below. First, summary statistics of the objective soil testing results are reported followed by a summary of the subjective farmer assessment. Then, a brief comparison of the subjective farmer assessments and the objective lab results is made. For greater detail regarding results from the LASER study, refer to Aynekulu et al. (forthcoming). 4.1 LABORATORY RESULTS: SOIL PROPERTIES IN THE STUDY AREA Variation in soil organic carbon (SOC) and nitrogen were observed across the three agro- ecological zones (Figure 12). The highest SOC concentration was measured in West Arsi (3.8%), followed by that in East Wellega (3.4%) and Borena (2.3%) areas. East Wellega has more acidic soils which could be suitable for maize production (FAO, 19083). Figure 12 illustrates the distribution of carbon levels and ph, two key soil properties. Significant differences of several variables are observed between the top- and sub- soil samples. The difference in means and corresponding t- test significance levels for key properties are reported in Table 3. Levels of all presented properties are significantly different between top- and sub- soil at the 1% level, with the exception of sand percentage, which is significant only at the 10% level. 34

35 (a) (b) (c) (d) Figure 12. (a,b) West Arsi has the highest SOC content while (c, d) soils from Borena area were more acidic than East Wellega and West Arsi. 35

36 Table 3. Objective Soil Results, by Depth!Soil!properties Top!Soil!(0>20!cm) Sub!Soil!(20>50!cm) Difference!in! Mean SD Mean SD means Physical %!Sand * %!Clay >2.3*** %!Silt *** Chemical ph *** Macronutrients: Total!Carbon!(%) *** Total!Nitrogen!(%) *** Exchangeable!Calcium!(mg!kg^>1)! +! *** Potassium!(mg!kg^>1) *** Exchangeable!Magnesium!(mg!kg^> *** Micronutrients: Iron!(mg!kg^>1) *** Zinc!(mg!kg^>1) *** Exchangeable!Manganese!(mg!kg^> *** +!Extracted!with!Mehlich!3!method *!Extracted!with!wet!method ***!p<0.01,!**!p<0.05,!*!p<0.1 Note:!Data!limited!to!plots!with!both!top!and!subsoil!samples!(n=1599) 4.2 SUBJECTIVE FARMER ASSESSMENT Prior to the collection of physical soil samples, a series of subjective plot- level questions was administered to the self- identified best informed household member on each plot. These questions ranged from a categorical coded- response what is the soil quality of your crop field? to questions on soil color, texture, and type (clay, sand, loam, etc.). An excerpt from the household survey is available in Annex II. It is worth noting that the subjective questions were administered at the dwelling, not upon direct respondent observation of the soils, as the study was aimed at assessing farmer knowledge for larger- scale surveys that may not allow for visitation of each plot. While subjective assessments of soil quality are both cost- and time- efficient, the quality of results is questionable. Summary statistics of the subjective questions included in the LASER study, found in Table 4, reveal little discrimination by respondents. 7 Slightly more variation is observed in the more specific subjective questions, such as texture or color, as opposed to overall soil quality. Figure 13 suggests that 7 The sample is limited to plots in which a topsoil sample was tested in the laboratory. Due to mislabeling of soil samples and/or transportation between the field and the laboratories, 120 plots with subjective measurements do not have matching objective measurements. These observations have been dropped. 36

37 farmers use soil color and texture as key indicators of soil quality. Dark and fine textured soils were categorized as good soils while red and course textured soils were often categorized as poor soils. This result is in agreement with the findings of Karltun et al. (2013) who found that crop yield, indicator plants, soil softness, and soil colour were useful indicators that farmers use to judge soil fertility in Ethiopia. Table 4. Subjective Farmer Assessment East% Wellega Borena West%Arsi Total N % Soil%Quality Good % Fair % Poor % Soil%Color* Black % Red % White/Light % Yellow % Soil%Type Sandy % Clay % Mixture%of%Sand/Clay % Other % Soil%Texture Very%Fine % Fine % Between%Coarse%and%Fine % Coarse % Very%Coarse % *%Categories%"White/Light"%and%"Yellow"%combined%for%analysis %Categories%"Very%Fine"%and%"Fine"%were%combined%for%analysis,%as%were%"Coarse"%and%"Very%Coarse" 37

38 (a) (b) Farmer Identified Soil Texture & Quality Farmer Identified Soil Color & Quality Percent Percent FINE BETWEEN COARSE AND FINE COARSE BLACK RED WHITE/LIGHT SR Good Quality SR Poor Quality SR Good Quality SR Poor Quality Figure 13. Farmers use (a) soil texture and (b) color as key indicators of soil quality. 4.3 COMPARISON OF OBJECTIVE AND SUBJECTIVE MEASURES If farmers are able to assess the fertility of their soils with reasonable accuracy, the benefits of soil sampling in household surveys may not outweigh the costs (depending on study objectives). However, results of the LASER study suggest that subjective farmer assessment of soils are not sufficiently correlated with laboratory results, the latter of which is taken to be the more accurate and specific measure. Carbon content is often considered to be the best single indicator of soil quality. Higher levels of organic carbon indicate good soil fertility and more optimal soil structure. Carbon is also highly correlated with other key properties such as total nitrogen. Do farmer assessments of overall soil quality reflect carbon levels? Descriptive analysis reveals little relation between soil organic carbon (%) levels and the respondent s assessment of the soil as poor, fair, or good. T- test results showed no significant difference in total carbon between good and fair soils, between good and poor soils, and between fair and poor soils (all comparisons with p > 0.1). To better illustrate the distribution of carbon levels across self- reported soil quality categories, Figure 14 includes a scatter plot relating carbon, ph, and self- reported quality category (Figure 14a) and box plots of carbon levels disaggregated by self- reported quality category (Figure 14b). The scatterplot reveals that the soils reported as poor are not concentrated in areas with low carbon levels, but rather seemingly randomly distributed. Disaggregation by plot manager sex yields even less insight. Neither male nor female manager assessments of overall soil quality discriminate by carbon level. Similarly, when disaggregating by 38

39 (a) (b) Figure 14. Self- reported overall soil quality estimates show little correlation to objective measures, presented here in terms of soil organic carbon and ph levels. manager literacy, there is no significant difference in carbon levels between subjective soil- quality categories. Refer to Aynekulu et al. (forthcoming) for more detailed analysis on respondent characteristics and the soil assessment. There does appear to be a relationship between farmer- reported soil type and percent sand, and between farmer- reported soil texture and percent sand, as the subjective analysis in Figure 13 suggested. Figure 15 plots the distribution of sand concentration in soils reported as fine, coarse, and between coarse and fine, with coarser soils expected to have a higher concentration of sand as opposed to silt and clay. The difference in sand concentration is significantly different than zero between all three categories, but the levels are in an unexpected direction as reportedly fine soils have 12.4% sand while sands reported between coarse and fine have 11.0% sand (reportedly coarse soil has 15.1% sand on average). Analysis over reported soil type (sandy, clay, or mixture of sand and clay) yields similar results. 39