Developing generic tools for characterizing agricultural systems for climate and global change studies (IMPACTlite – phase 2) Mariana C. Rufino 1 , Carlos Quiros 1 , Moussa Boureima 2 , Solomon Desta 3 , Sabine Douxchamps 4,1 , Mario Herrero 1 , Jusper Kiplimo 1 , Dikite Lamissa 5 , Joash Mango 6 , Abdoulaye S. Moussa 7 , Jesse Naab 8 , Yacine Ndour 9 , George Sayula 10 , Silvia Silvestri 1 , Dhiraj Singh 11 , Nils Teufel 11 , Ibrahim Wanyama 12 1International 2Institut National de la Recherche Agronomique du Niger (INRAN), Niamey, Niger 3Managing risk for improved livelihood (MARIL), Addis Ababa, Ethiopia 4International 5Institut 6World Livestock Research Institute (ILRI), Nairobi, Kenya Water Management Institute (IWMI), Ouagadougou, Burkina Faso d’Economie Rurale (IER), Bamako, Mali Agroforestry Centre (ICRAF), Kisumu, Kenya 7Program on Climate Change, Agriculture and Food Security (CCAFS), Bamako, Mali 8Savanna Agricultural Research Institute, Wa, Upper West Region, Ghana 9Institut 10Selian Sénégalais de Recherches Agricoles, Dakar, Senegal Agricultural Research Institute (SARI), Arusha, Tanzania 11International Livestock Research Institute, New Delhi, India 12International Institute for Tropical Agriculture (IITA), Kampala, Uganda Report of Activities 2012 Submitted by ILRI to the CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS) 31st January 2013 1 Contents 1 Introduction .......................................................................................................................................... 3 2 Methods ................................................................................................................................................ 3 2.1 3 Village lists and village census in the Indo-Gangetic plains (IGP) ................................................. 4 Identification of production systems and village selection .................................................................. 5 3.1 Indo-Gangetic plains (IGP) ............................................................................................................ 5 3.2 East Africa (EA) .............................................................................................................................. 9 3.3 West Africa (EA) .......................................................................................................................... 13 4 Status of data across sites ................................................................................................................... 18 5 Data analysis ....................................................................................................................................... 18 5.1 Calculation of performance and livelihood indicators ................................................................ 18 5.2 Income calculations .................................................................................................................... 19 5.3 Crop diversity and activity diversity ............................................................................................ 20 5.4 Food security and food self-sufficiency ...................................................................................... 21 5.5 Household diet diversity score.................................................................................................... 21 5.6 Household domestic asset index ................................................................................................ 22 6 Data management and databases ...................................................................................................... 22 7 Future steps ........................................................................................................................................ 36 8 References .......................................................................................................................................... 37 9 Annex .................................................................................................................................................. 38 9.1 Village locations in research grids in IGP .................................................................................... 38 9.2 Village locations in research grids in Senegal ............................................................................. 42 2 1 Introduction Standard methods are needed to collect data to evaluate the performance of agricultural systems. Standardization allows comparisons across systems, and potentially the extrapolation of recommendations to similar development domains. The Integrated Modelling Platform for Mixed Animal Crop Systems (IMPACT, Herrero et al., 2007) is a data collection protocol and computer software tool designed to gather minimum datasets in smallholder crop-livestock systems. The protocol collects information ranging from household composition to crop and livestock production to household food consumption and household assets. Although IMPACTS’s datasets are detailed to conduct a wide range of crop-livestock systems analysis, the data collection on the field proved to be time and resources demanding. The protocol works in monthly time steps i.e., most of the data is collected per month, thus it takes considerable time to complete an interview. The printed forms comprise over 30 different templates that resemble the screens in the software, and although this could speed up data entry, its use on the field for data collection often tends to be inoperative. In 2011 CCAFS commissioned to ILRI the task of assessing the possibilities for simplifying IMPACT to carry out a characterization across the 15 CCAFS benchmark sites (Table 1). The objective of this project called ‘IMPACTlite’ was to modify IMPACT to be able to collect household-level data detailed enough to capture within-site variability on key performance and livelihood indicators that could be used for a range of analysis including the modelling of impact of adaptation and mitigation strategies on livelihoods, food security and the environment. A team composed of agronomists, economists, computer, environmental, and social scientists modified and tested the new tool in a number of sites. The changes implemented to the tool are reported in Quiros et al. (2011). In this report we describe the steps undertaken in the implementation of the surveys using IMPACTlite in the IGPs, the progress achieved, and next steps planned for 2013. 2 Methods The surveys were implemented across the 15 CCFAS sites (Table 1) starting from February until December 2012. The details of the procedures for the implementation can be found in the IMPACTlite Training Manual (Rufino et al. 2012). In brief, the steps were: 1) Gathering secondary data for each research grid, 2) Definition of agricultural production systems, 3) Villages selection, 4) Generating village information: creating a list of households, 5) Selecting households from a village list, 6) Replacing selected households, and 7) Implementing the survey. The IMPACTlite team at ILRI gathered secondary information for each of the research grids, with assistance from the CCAFS Science officers. Most of this information was gathered during the implementation of the Household Baseline surveys. We enclose to this report the training manual of IMPACTlite and a sample questionnaire used in Vaishali, India. 3 Table 1: CCAFS baseline sites and sampling frames description Country Site Site ID Research grid Research grid coordinates Kenya Nyando KE01 Katuko Odeyo (35.068, 0.269)(35.068, 0.361) (34.979, 0.361)(34.979, 0.269) Kenya Makueni KE02 Katheka-Kai (37.326, -1.581)(37.378, -1.657) (37.298, -1.702)(37.244, -1.624) Uganda Albertine Rift UG01 Hoima (31.546, 1.445)(31.546, 1.535) (31.457, 1.535)(31.457, 1.445) Uganda Kagera Basin UG02 Rakai (31.394, -0.921)(31.481, -0.621) (31.484, -0.713)(31.394, -0.713) Tanzania Usambara TZ01 Lushoto (38.714, -4.850)(38.301, -4.850) (38.301, -4.790)(38.417, -4.790) Ethiopia Borana ET01 Yabare (38.278, 4.975)(38.549, 4.975) (38.549, 4.704)(38.278, 4.704) Burkina Faso Yatenga BF01 Tougou (13.554, -2.113)(13.554, -2.391) (13.828, -2.391)(13.828, -2.391) Ghana Lawra-Jirapa GH01 Lawra (10.455, -2.624)(10.455, -2.911) (10.735, -2.624)(10.735, -2.911) Mali Segou MA01 Cinzana (13.228, -5.613)(13.228, -5.912) (13.509, -5.911)(13.509, -5.613) Niger Kollo NI01 Fakara (13.379, 2.826)(13.654, 2.826) (13.654, 2.547)(13.379, 2.547) Senegal Kaffrine SE01 Kaffrine (13.968, -15.407)(14.242, -15.407) (14.242, -15.686)(13.968, -15.686) India Bihar INXX Vaishali (25.803, 85.296)(25.803, 85.392) (25.713, 85.296)(25.713, 85.392) India Haryana IN15 Karnal (29.842, 76.888)(29.842, 76.993) (29.750, 76.888)(29.750, 76.993)1 Bangladesh Khulna BA03 Bagerhat (22.552, 89.812)(22.552, 89.912) (22.461, 89.812)(22.461, 89.912) Nepal Rupandehi NE03 Rupandehi (83.401, 27.500)(83.500, 27.500) (83.401, 27.589)(83.401, 27.589)2 1 Two squares were excluded as urban: (29.8090, 76.9190)(29.8090, 76.9380) (29.7920, 76.9190)(29.7920, 76.9380) and (29.8424, 76.9229)(29.8424, 76.9414) (29.8286, 76.9229)(29.8286, 76.9414) 2 One square was excluded as urban: (27.5270, 83.4372)(27.5270, 83.4712) (27.5000, 83.4372)( 27.5000, 83.4372) 2.1 Village lists and village census in the Indo-Gangetic plains (IGP) The first step in preparing the household survey in the four research grids was to establish a list of all villages within these grids and to collect some basic village-level information. Unfortunately, secondary data was not available in sufficient detail or quality. Also, the previous base-line survey did not provide village lists with geographic identification. Therefore, a member of the survey team set out initially to compile a list of all villages, record gps co-ordinates and fill short (1-page) questionnaire form on village characteristics: Village identification, population size, land and irrigation resources, ranking of livestock and cropping activities by importance to village. For this a 4 motor-bike was hired and the grid area thoroughly combed. During this exploratory phase initial results were passed back to the office and village locations were compared to Google Earth images. Visible settlements missing in the initial lists were discussed and added were appropriate. An important issue at this stage was the definition of "village". In India, the oldest and bottom-level administrative unit is the "revenue village". As the name implies this has been established for tax reasons and is therefore more linked to land than to population. Therefore, some revenue villages don't contain any settlements while others contain several which don't form any social unit. In addition, settlement patterns differ greatly throughout the Indo-Gangetic Plains. While in areas with relatively late colonization, as in Haryana and Punjab, villages are comparatively large and well defined, other areas which had had high population densities for centuries, as in Bihar, settlement patterns are much more dispersed. Thus, we developed an own approach to defining a settlement as a village: Wherever a settlement has an own distinctive name it is regarded as a village. However, if a settlement is only regarded as part of a larger village without an own name it is regarded as a hamlet (tola). Often such hamlets are home to a certain community and are known only by the community name. In Nepal, local administrative units have been consistently organized into village development councils (VDC). Each VDC consists of 9 wards, representing small hamlets. For sample selection purposes within this study, a VDC is a useful unit. It is small enough to enable its inhabitants to know of one another and function as a social unit. In Bangladesh settlement patterns resembled the Bihar situation and we applied the same procedure by defining "villages" according to naming practices. It took about two weeks to establish a complete village list with the accompanying basic data at each research site. Google Earth images of all four research grids with village locations are included in the annex (section 8). 3 Identification of production systems and village selection 3.1 Indo-Gangetic plains (IGP) The identification of production systems within research grids is based on the village census data (see 2.1). Respondents (one group per village) were asked to rank crops (including aquaculture) by their importance for the three main cropping seasons as well as livestock (no seasonal differentiation). Here, only count values (number of villages reporting specific rank or rank aggregate) are reported. These results were used to decide on whether to define one or several clearly demarcated production systems within the respective grids. In case of two or more production systems, villages are classified according to production systems and households are sampled separately within each production system class. This ensures a statistically useful sample of households for each production system. Dominant land use The first indication of dominant crop production system (including aquaculture) is whether a specific crop (or crop combination) is mentioned as rank1 in more than one season. Here, paddywheat is defined as assigning rank1 to paddy in kharif (rainy season) or summer and to wheat in rabi (winter). All other specific land-uses are defined as dominant when showing rank1 in more 5 than one season. Villages where none of the four specific cropping systems dominate are classified as mixed-cropping. Table 2: Dominant land use system in the IGPs [village counts] Vaishali Rupandehi Karnal Bagerhat 71 52 20 0 Paddy 0 0 0 29 Mixed cropping 4 1 1 5 Vegetables 7 0 1 1 Aquaculture 0 0 0 7 Paddy-wheat Vaishali, Rupandehi and Karnal are clearly dominated by paddy-wheat. Thus, only one production systems is recognized in these research grids. In Bagerhat, paddy dominates most villages. However, about a sixth of all villages see aquaculture as most important land use activity. Six out of these seven villages are located in the northern part of the grid. However, paddy dominated and mixed villages are also located there. In addition, paddy is of considerable importance also in the aquaculture dominated villages: 4 out these 7 villages include paddy 3 times within the nine ranking questions on crop production (rank 1-3 over three seasons), 4 villages mention paddy twice and 1 village once. On the other hand, aquaculture also frequently appears as important land use in the paddy dominated villages. Therefore, also within the Bagerhat research grid only one production system (paddy-aquaculture) is recognized. Livestock In Vaishali, Rupandehi and Karnal livestock is dominated by dairy production. In Vaishali and Rupandehi all three ranks are common to most villages in the grid; Vaishali: 1st dairy, 2nd small ruminants, 3rd poultry; Rupandehi: 1st dairy, 2nd small ruminants, 3rd draft animals. However, in Karnal the most common combination (1st dairy, 2nd small ruminants, 3rd draft animals) accounts for only about 30% of villages in the grid. In Bagerhat it appears that both dairy and poultry production are similarly important with small ruminants coming third. 6 Table 3: Livestock ranks in the IGPs Rank1 Rank2 .00 draft animals dairy animals dairy animals small ruminants pigs poultry other small ruminants poultry dairy animals draft animals dairy animals small ruminants Rank3 .00 .00 small ruminants pigs poultry draft animals .00 draft animals pigs poultry fish draft animals small ruminants small ruminants draft animals draft animals poultry dairy animals small ruminants draft animals dairy animals Vaishali 3 1 5 0 1 0 6 14 0 50 1 0 0 1 0 0 0 0 0 0 0 Rupandehi 0 0 0 0 0 0 0 33 2 6 0 0 0 1 0 8 3 0 0 0 0 Karnal 1 3 3 1 3 1 0 6 0 0 0 1 1 1 1 0 0 0 0 0 0 Bagerhat 0 0 0 0 0 0 0 0 0 4 0 0 0 19 0 0 0 1 15 1 2 Conclusion on production systems In all four research grids only one production system was found to be important. In Vaishali, Rupandehi and Karnal this is paddy-wheat with dairy animals while in Bagerhat this is paddyaquaculture with dairy and poultry. Village selection After having determined that only one production system was to be considered within each research grid, the agreed sample of 20 villages was randomly selected from each grid. For this, a random number was assigned to each village. Villages located on the grid border were excluded. Similarly, the 10% smallest villages (by number of households) and the 10% largest villages were excluded in order to avoid having the sample influenced by extreme cases. A very small village of less than 20 households for example would often consist of only a certain type of households while very large villages of perhaps more than 1000 households would already have characteristics of a small town. In Karnal all identified villages were included as only 19 villages were finally considered. Household census and household selection After having selected the 20 sample villages in each research grid a household census was conducted in these villages. For this a suitable village person was identified (e.g. school leaver, student). This person then listed all households within the own village and collected basic characterizing data for each household: Household identification (household head name, father's name, village name, hamlet name), household head (age, gender), household size, land (ownership 7 and cultivation) and livestock. In general the completion of the household census took 5 to 7 days for each village. The village enumerator was paid by listed household after checking for data quality. Subsequent to the completion of the household census lists, the data were entered by members of the survey team and additional data entry staff. The total number of households in the 20 villages selected in each grid varied considerably by site: In Vaishali the household census includes 7953 households, in Rupandehi 2251 households, in Karnal 7270 households and in Bagerhat the household census lists 6250 households. Because the decision to move the research site in the Upper Gangetic Plains from Sangrur (Punjab) to Karnal (Haryana) was taken late, a household census was also performed in Sangrur. Here 6807 households are listed. The differences in village size between the village census and the household census data are not negligible. Partly, this may be due to differences in including or excluding hamlets. However, they also underline the inherent level of accuracy of village level data collection. Following the implementation of the household census, the household sample for the household survey was selected. For this, households not cultivating any land and not keeping any livestock (large or small ruminants) were excluded as they don't have any direct involvement in agriculture. Then, each household was assigned a random number and ranked by this number. The first 200 households were selected for the survey. The second were identified as replacement households in cases households were not available during the survey period. Villages were not considered during the household selection process as it is not assumed that "village" will be an important factor during the analysis process, which would justify balancing the number of households per village. Also, travel distances within the research grid are very limited, so that it is not necessary to ensure a minimum number of selected households per villages in order to simplify survey logistics. Implementation of the survey The initial training of the survey team for India was conducted in Patna (Bihar) during the first week of July 2012. The team consists of two supervisors, four enumerators and two data entry operators. The two supervisors shared collecting preparatory data (identifying villages, village census, household census) and leading the survey teams during the household survey. The same survey team conducted the household survey in Vaishali (08/07/2012 to 05/08/2012), Rupandehi (09/09/2012 to 08/10/2012) and Karnal (13/10/2012 to 10/11/2012), as Hindi is sufficiently spoken in all these sites. Data entry was started in parallel to the actual survey at the survey site in order to discover data issues as soon as possible and to speed up the overall data collection phase. For Bagerhat, where only Bengali is spoken, a new survey team was established with the help of a local NGO (SMKK). Their training was conducted by Dhiraj Singh at Bagerhat. Madhuresh, one of the survey supervisors, was also able to join the team for the final training and for the survey to ensure consistency and data quality. The two data entry operators also travelled to Bagerhat to improve data quality by entering data during the survey process. Currently, the survey is on-going. It was completed the first week of January 2013, cleaned dataset available by end of January. 8 3.2 East Africa (EA) Nyando and Makueni (Kenya) The process of identification of production systems in Nyando and Wote involved: i) analysis of satellite high-resolution images, ii) transect drives through the grid, iii) group interviews, iv) individual interviews to key-experts, v) household interviews. Three production systems were identified based on land cover, production orientation and land use intensity. Verification was conducted on the ground through key stakeholders which involved officers from the Ministry of agriculture, Chiefs, opinion leaders, local farmers and own observation. The three dominant production systems identified after verification on the ground were for Nyando: a) Crop-livestock with free-grazing local breeds (maize and sorghum based), b) Sugarcanemaize with cross-bred cattle, c) Dairy-perennials-maize based (tea, Napier grass). The three systems were delineated spatially. A list of villages was developed for each system, villages falling in the buffer or transition zones were filtered out. To cover variability within the larger production system eight villages were randomly selected. The other two systems were randomly assigned six villages each. Within each village ten households were randomly selected from the household lists for the survey. In Wote, the team identified two main production systems: i) crop-livestock mixed with local sheep, ii) crop-livestock mixed with dairy. The village definition used is that described in the training manual. The village list was built with the help of the village elders who are in charge of various villages of which they know the boundaries. The household list was developed by using the list provided by the village elder as a starting point but verification was done by going through the village from one household to the other by numbering them, we found out additional household within some villages. Based on the three systems identified the areas where they fall were marked. Most of the systems were falling within sub-location of which the boundaries are well known by the chief who are governing the areas, we used the same boundaries. It come out that the first system covered a larger area than the rest. A list of villages was developed in each system and villages falling in the buffer zones were eliminated from the list as they were impossible to distinguish actual system on them. Due to the wide coverage of system one it was randomly assigned 8 villages in order to cover the heterogeneity within the system. The other two systems were randomly assigned six villages each. Within each village ten household was randomly picked for the survey. The same process was used in Makueni site. In Kenya, the training of enumerators and testing of questionnaires were carried out together by ILRI and ICRAF staff. Carlos Quiros and Mariana Rufino conducted the first training and testings for IMPACTlite in Nyando in March 2012. Joash Mango (ICRAF) and Ianetta Mutie (ILRI) conducted trainings in Wote in August 2012. Local partners implemented the data collection in Wote. 9 Lushoto (Tanzania) Definition of production systems in Lushoto was supported by the use of a satellite image of the grid. All villages in the grid were identified. After this exercise, the team conducted an inventory of farming activities at each village engaging the village authorities and extension staff. The farming activities inventory was based on the history and current farming systems. Activities considered were food crops, cash crops, livestock keeping, agroforestry and horticultural crops. Finally, and on the basis of the activities and their frequencies the grid was classified into three production systems namely: i) Production system 3 includes the uplands experiencing three rain seasons namely; short, intermediate (muluwati in kisambaa), and the long rains. This production system is dominated by maize and beans, fruits trees, some vegetables and some perennials (coffee, tea and woodlots). Majority of the householders keep crosses of exotic and local zebu cattle and some goats. ii) Production system 2 is characterized by maize and beans with cassava and some horticultural crops. This householders practice zero grazing with crosses of exotic and local cattle breeds. iii) Production system 1 is different from 2 and 3 in that householders only keep indigenous species of livestock. The village list was built using the original list from the CCAFS Baseline survey block during the first CCAFS HHS baseline survey which basically was met the program protocol to identify, the villages, the households and lastly the 20 randomly households. In this case, the villages were listed from the same grid but grouped according to the existing production systems within the block. The household list was built through the village household list prepared by the village authorities for the old (7) villages and verified by the CCAFS before conducting the HHBS. In new villages the same procedure was used. What was new here is that we requested the old village’s authority to update their household list and from there we followed the same procedure of randomizing them and came up with households to be respondents. Process followed to identify production systems: The team went through each village and consulted the village government and extension agents and we were satisfied with the information and of course, the team especially the enumerators and the site coordinator had enough experience on the farming systems in the district as well as in the CCAFS grid. Determining the boundaries on the ground was more difficult but we based on the government boundaries or village boundaries to establish those three dominant production systems in the block of 10 km x 10 km. Training of enumerators and testing of the questionnaires was conducted by Joash Mango from ICRAF. The survey was finalized by the end of July 2012, and the data was delivered by September 2012. Borana (Ethiopia) The Borana household survey was conducted in 20 village clusters in Yabello and Arero woreda for 20 consecutive days; starting from September 18 – October 7, 2012. It involved five enumerators and one field supervisor. The four pastoralist associations (PAS) out of five (PAS) inside the 30 x 30 10 blocks were selected and only one village namely Fuldawa from Arero, which is located at the extreme corner of the block and inaccessible was deliberately excluded. Each cluster was chosen based on shared community enclosure which is owned, managed and utilized exclusively by the household in the cluster. The sampling frame was constructed using a total of 1150 households’ list obtained from 20 clusters of the 4 Pastoral Associations (PAs). The sampling frame included 397 households from Denbela Saden, 387 from Dikale, 269 from Alona, and 97 from Gada PAs. Table 4: List of study clusters in the Borana grid Wereda PA Clusters Clusters’ name Yabello Denbelasadden 7 Tadech Denbela; Doyo Duba; Guyo Kuyo; Tuke Halake; Bule Arero; Boru Goliso; Dureti Karata and Dida Adi) Yabello Dikale 5 Jirmo Dida; Guyo Jateni; Gelgelo Bule; Elman Simpere and Elman Eiya Arero Alona 6 Debeso Jateni; Wako Elema; Jateni Molu; Kotola Soka; Doyo Guyo; Gutu Kela Arero Gada 2 Konso Dida; Halake Goyo 4 20 The settlements pattern and the clusters in Arero were more scattered over a vast area than the settlements in Yabello, which are much more consolidated in specific areas. The households in the sample clusters were listed using key informants, elders, local development agent and PA leaders. Ten households per village were selected randomly. A random access table was used to select households from the list. Only few replacements were made in 25% of the selected 20 clusters. This was done based on satisfactory reasons and confirmations for absence of the selected respondents from the locality. Key informants from among the herders, elders, PA leaders and government pastoral development and landuse and administration experts were used to define the production system prevailing in the study block. A consensus was reached by most of the key informants that there is only one production system that is a pastoral production system; hence the block was categorized under pastoral system. This is in line with the results of the household baseline survey of CCAFS carried to characterize the block. Two hundred households 10 from each cluster were interviewed under a single production system. All the croplands were small size, opportunistic, and fragmented subsistence type virtually with no use of inputs at all. The purpose of the survey was well introduced to cluster leaders before listing households, sampling and data collection. Once confidentiality issue ensured respondents became fully 11 cooperative and expressed their willingness to participate in the survey. Time was taken to describe the objectives of the research and respondents’ willingness, feelings and reactions were assessed before questionnaires were filled out. This helped out the trustworthiness of information obtained. No one has shown unwillingness to participate. There were few replacements made because of satisfactory reason for the absence of the sample households from the area during the survey. Randomly selected replacements are made as per the procedures. Five enumerators and 1 field supervisor were involved to cover one cluster in a day. The training of enumerators and testing of the questionnaires was conducted by Mariana Rufino and Solomon Desta in August 2012. Trained enumerators were used to fill out the questionnaires throughout the survey. Each enumerator filled 40 questionnaires in twenty days. The questionnaire typically lasted in average between 2.30 – 3.00 hours with each respondent. Enumerators are supervised during the data collection using available time frame. The mobile phone helped further to communicate and troubleshoot problems encountered by the enumerators at times when the supervisor is engaged in GPS recording and other activities. Each day the questionnaires filled were checked for completeness, clarity and consistency and discussed with enumerators each morning before meeting respondents for data collection. A minimum of 45 – 60 minutes per day is spent each morning with enumerators before survey began. Major challenges encountered: i) inaccessibility of some of the clusters due to poor access roads, ii) the extended unusual short rain blocked road access to some of the cluster villages, iii) villages especially in Arero were scattered over a wider area and that made travel distances between households cumbersome for taking GPS coordinates. The survey was successfully finished in October 2012, and the data was delivered in November 2012. Hoima and Rakai (Uganda) To determine production systems in the two grids in Uganda pre-determined GPS coordinates were used to demarcate the 10 by 10 kilometer area by overlaying the coordinates on shape file of Uganda parishes in ArcView GIS 3.3 software. After identification of the parishes that lie within two grids, we approached district officers at the respective agricultural production departments to brief them on the research we were to carry out and for possible important contacts. There was information at the district level about the farming system (crops and livestock produced) but too general. More precise information regarding farming systems at the village level was obtained from parish leaders and village local council leaders. In addition, transect walks across the grids were carried out to complement on information obtained from local and village leaders. The Hoima grid comprises four parishes: Buraru, Bulindi, Kibugubya in Hoima district and Kahembe in Masindi district. The householders from the three parishes of Hoima district are mainly cultivating maize, beans, cassava, bananas and a few keep local indigenous livestock. Ten villages were randomly selected from which one hundred farmers were randomly chosen, ten from each village. In Kahembe Parish (Masindi) maize, beans, banana, cassava, sugarcane and few local livestock are the major farming activities. This parish is in close proximity to the Kinyara sugar 12 factory so sugarcane production is taking root in the area. In this parish there were 7 villages from which 100 farmers were randomly sampled from the village list. These two situations in the Hoima block were first treated as two different production systems with sugarcane production being the differentiating factor. One hundred farmers were interviewed in each of the systems however upon randomly selection of farmers in the system which has sugarcane only around 20% of the farmers interviewed had sugarcane. Therefore, the Hoima grid was treated as a single production system since sugarcane farmers in the parish were not enough to qualify as a different production system. In the Rakai grid, six parishes fall within the grid: Kiyovu, Kasensero, Bitabago, Buyamba, Butiti and Byakabanda. Across all the villages, there were two major crop components, the perennial crops (banana and coffee) and annuals (mainly maize, beans, cassava, groundnuts and sweet potatoes). Farmers in this grid also keep local livestock like cattle, goats, poultry although in small numbers. Therefore, one production system was identified in this grid: Coffee-banana with annuals and few local livestock. A total of twenty eight villages were randomly selected from which 200 farmers were randomly selected from the village lists. The training of enumerators and testing of questionnaires took place in March 2012, and conducted by Mariana Rufino, Carlos Quiros, Silvia Silvestri (ILRI) and Joash Mango (ICRAF). This was the second training of enumerators of the whole project. The purpose of having such a large team was to agree on the contents of the training so that we could split responsibilities for other sites. Josh mango came back to Uganda to support the start of the survey in June 2012. Surveys were completed by September 2012, and the data delivered in December 2012. The local team at Uganda encountered difficulties with data entry, so finally data was entered at the Kisumu office of ICRAF under the supervision of Joash Mango. 3.3 West Africa (EA) Kaffrine (Senegal) The list of villages for the Kaffrine grid was established. There were some difficulties as some names on the list from CCAFS did not correspond exactly to the names of villages they had. Two villages from the list of CCAFS were not encountered in the actual list of villages: Moula Ndiaga and Sare Lamou. The identification of the production systems in Kaffrine was based on a consultation with the key services active in the region (Agriculture, Water and Forests, Ecology, ISRA, ANCAR) and a series of documents: i) a soil map of Senegal, showing the areas with crop and livestock production, ii) an hydrologic map of the region of Kaffrine (“Direction de la Gestion et de la planification des Ressources en Eau” - DGPRE), iii) a map showing the forest resources (“Centre de Suivi Ecologique” - CSE), iv) a map of Kaffrine, v) the local development plans of Malem Hodar and of the rural community of Kahi, vi) a document from ISRA on the characterization and the typology of farms in the region of Kaffrine. At first, only one system had been identified by the local agricultural services: a cropping system with a biennial rotation groundnut/cereal. But some more discussion allowed identifying 3 13 production systems for the block (Fig. 5 Annex): a) Agriculture and forestry (31 villages), where a development of some agroforestry activities can be noticed with the proximity of classified forests, in addition to the cereal and groundnut cultivation (Zone 1); b) Mixed crop-livestock system (30 villages), in a more pastoral area (Zone 2); c) Crops and vegetable production (62 villages), where a NGO (World Vision) drilled wells (Zone 3). The list of villages for each production system was established. Villages in transitions zones between two systems were excluded from the lists (6, 4, and 11 villages for the production systems a, b and c, respectively). Villages were randomly selected for each production system for a total of 20 villages, with 7, 6, and 7 villages for the production system a, b and c, respectively. One village (Keur Sandao) of the production system b (mixed crop-livestock systems) has been replaced as it had less than 4 households: most of the villagers migrated because of floodings. This village was replaced by the village of Kahi, after a random selection. The villages selected are: a) Agriculture and forestry (7): Goria Mbande, Mbella Ouolof (Mbella Saloum), Ngalick, Khende, Diagle, Ndodji, Nianghene Ouolof; b) Mixed crop-livestock system (6):Bagana, Kanka, Kebe Keur Lahine, Korky Bambara, Moukhoume, Kahi; c) Crops and vegetable production (7): Loumene, Gainth Gouye, Gainth Peulhi, Ngatou Malick, Medina Ndiayene, Mbene Diouma, Ngidiba A phase of sensitizing was carried on between the 18 and the 27th of June 2012, to meet the chiefs of the villages and the population and explain how would the survey would be organized. During this first visit, the actual lists of households were established for each village with the chiefs. Then, 10 households were randomly selected per village, for a total of 200 households. The training of enumerators and testing of the questionnaires took place in May 2012 and conducted by Sabine Douxchamps (IWMI-ILRI), Mariana Rufino and Yacine Ndour. The survey was finished by September 2012 and the data delivered in November 2012. Tougou (Burkina Faso) The list of villages from the CCAFS baseline survey was confirmed. The site of Tougou, in the Northern Region of Burkina Faso, comprises four rural communities (Namissiguima, Ouahigouya, Barga, and Titao) and 51 villages. Together with the technical services (chefs ZAT – Zone d’Appui Technique) of the communities of Namissiguima, Ouahigouya and Barga, and the Province Director of the Lorum, three production systems were identified: i) mixed crop-livestock system (25 villages). There can be a dominance of either crop or livestock. This system occupies most of the area; ii) mixed crop-livestock system + vegetables (20 villages). Vegetable cropping during the dry season if some water is available; iii) mixed crop-livestock system + agroforestry (3 villages). There is no clear separation between the systems. The whole area is under mixed crop-livestock system, and then the two other systems are scattered here and there, depending if there is a source of water during the rainy season (Dam of Tougou, wells, or dugouts) or a forest. Vegetable growing 14 and agroforestry would then represent an additional source of diversification of the basic croplivestock system. All key informants agreed on this. There was never only one production system in a village, but always a mixture. The dominant production system in a village (i.e. more than about 60% of the household practice it) defined to which production system the village would be assigned. An important non-agricultural activity in the region is gold washing: the 3 villages concerned by this activity were discarded before the random selection. Seven villages were randomly selected by production system, except for the mixed system with agroforestry where only 3 villages were available, and were all selected. i) mixed crop-livestock system: Rapougouma, Longa, Sillia, Salla Foulbe, Todiam, Hargo, Poukouma ramssa, Ramdolla peul ii) mixed crop-livestock system + vegetables: Dinguiri, Karma, Lemnogo mossi, Sabouna, Rikou, You iii) mixed crop-livestock system + agroforestry: Tougou, Solgom, Bagayalgo The lists of households per villages were established with the technical services and the Villagers Development Council (“Conseil Villageois de Developpement”) of each village. Then, 10 households were randomly selected from the list for each village of the mixed crop-livestock system and the mixed crop-livestock system + vegetables production systems, and 20 households were selected per village for the mixed crop-livestock system + agroforestry system, so in total 200 households. The training of enumerators was conducted in June 2012 by Sabine Douxchamps. The survey ended September 2012 and the data was delivered in November 2012. Cinzana (Mali) Samples villages were selected in the rural communes of Cinzana and Katiéna. Criteria used for villages’ selection were: i) villages within the block of 30x30 km2, ii) permission from the village authorities/elders to conduct the survey, iii) representative villages (size and inhabitants) among the villages of the block, iv) ease of access (road). The selection of villages was done with extension services and IER team led by Lamissa Diakité. Following the introduction on the CCAFS programme, the activities/projects conducted over the past two years at the Segou site, and presentation of the objectives of the survey, 20 villages were selected from the total list of 46. There is about 12 805 households and 56 744 inhabitants in the block. The block is homogenous (from a physical and socioeconomic perspectives), and therefore very little difference between villages and households. To reach 200 households among the 20 villages, samples of 10 households were selected in each of the village. In each village, households were selected from the census list available at the village level. The census list was updated beforehand. A random sampling approach was used whereby all the households listed were put together in a hat and 10 households were selected randomly. 15 The identification of the production system was done through grey literature and consultation of key informants such as extension services, rural development projects operating in the area, villages’ elders and local decentralized public officers (agriculture, environment, livestock and fisheries). From this process, only one production system (agro-pastoral, with agriculture as the main component and extensive livestock production as the second component) was identified and agreed upon by all the stakeholders. The training was conducted by Abdoulaye Moussa from the CCAFS West Africa office in June 2012. The survey was finished by October 2012 and the data delivered in December 2012. Fakara (Niger) The Niger household survey was conducted in 46 villages in 2012, for 15 consecutive days; starting from 11th of August 2012 to 25th of August 2012. In order to identify the production systems the site coordinator has carried out an expert consultation on site. The expert that has been consulted has been working in the research area for about 20 years, and has been collaborating with different research institutes. The following two production systems have been identified: 1. Subsistence crops and livestock farming with local species. Specifically this production systems presents: Subsistence crops such as: millet, sorghum, cowpea, sesame, maize, peanut, okra and Livestock farming with local species: oxen, goat, sheep, chicken, guinea fowl. 2. Subsistence crops, market gardening and livestock farming with local species. Specifically this production system presents: Subsistence crops such as: millet, sorghum, cowpea, sesame, maize, peanut, okra, Market gardening: cabbage, potatoes, salad, onion, tomato, gourd, courgette, carrot, sweet potatoes, cassava, Cultivation of Moringa, Livestock farming with local species: oxen, goat, sheep, chicken, guinea fowl. The market gardening is practices during the dry season. The Moringa is cultivated the entire year and both leaves and grains are commercialized. Most of the market gardening products and Moringa are sold. The identification of the list of villages has been based on the map ‘CCAFS_hbs_Fakara_Niger /Icrisat GIS lab : Novembre 2010’. In this map are localized the 46 CCFAS villages/sites in Niger. In each of the two clusters of villages established on the basis of the two production systems, a random selection has been done to select 10 villages, 5 for each of the production system. The selected villages are reported in bold in table 5. In the table 5 is represented the distribution of the villages according with the production system. Table 5 - Distribution of the villages according with the production system. In bold are indicated the villages selected for the survey. Production systems Villages with a specific production system Subsistence crops and livestock farming with local species 1. Baboussay 2. Bari Touri 16 Subsistence crops, market gardening and livestock farming with local species 3. Touliel 4. Darey 5. Dolohi 6. Balal Sagui 7. Fandou Béri 8. Fetokadie 9. Fandobong 10. Kida Bazagaize 11. Koma Koukou 12. Touloua Kouarey 13. Bagoua 14. Tigo Tegui (Abamate) 15. Gorou Yena 16. Tondi Kiboro 17. Sama Dey 18. Boundou Warou 19. Ko Kaina 20. Boula Darey 21. Kampa Zarma 22. Kampa Peul 23. Kida Tafa Kouara 24. Karbanga 25. Katanga 1. Dantiandou Tegui 2. Maourey Koira Zeno 3. Maourey Tokobinkani 4. Sabou Dey 5. Youloua 6. Bankadey 7. Gasseyda 8. Gao Bangou 9. Guileyni 10. Zoroney 11. Dantiandou 12. Bokossay 13. Falanga 14. Dey Tegui 15. Tigo Zeno 16. Korto 17. Yerima Dey 18. Gorou 19. Banizoumbou 20. Boundou 21. Kalassi For each village the list of the households has been compiled. The choice of the households to interview has been done during an assembly in each one of the selected villages. All the social categories have been taken into account when sampling the villages. Two hundred households, 100 from each production system were interviewed. Five enumerators and 1 field supervisor were involved. Trained enumerators were used to fill out the questionnaires throughout the survey. Each enumerator filled 40 questionnaires in 15 days. The questionnaire typically lasted in average 1 hour and 30 minutes with each respondent. The survey activity did not present any particular problem of implementation. This may have been due to the 17 fact that the chiefs of the villages together with opinion leaders have been informed before the beginning of the survey and were fully involved in all steps. The training was conducted by Silvia Silvestri (ILRI) and Abdoulaye Moussa in June 2012. The survey was finished by November 2012, and until 1 February 2013 there are some issues with data delivery. 4 Status of data across sites Because of planning and budgetary constraints, the start of the implementation of the surveys varied across the sites and regions (see Table 4). For the IGPs, the surveys started later than at the other regions, partly because there was no consensus on which were the CCAFS benchmark sites to be surveyed. At the moment, there still one site where the survey has not finished. We expect that the dataset will ready by the end of January 2013. Table 4: Current status of the implementation of the surveys across sites Production system identified CCAFS site 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Nyando Hoima Rakai Lushoto Wote Borana Lawra Tougou Cinzana Kaffrine Fakara Vaishali Karnal Rupandehi Bagerhat KE UG UG TZ KE ET GH BF MA SE NI IN IN NE BA Feb-12 Jun-12 Jun-12 Jul-12 Jul-12 Aug-12 Jun-12 Jun-12 Jun-12 Jun-12 Jun-12 Jul-12 Oct-12 Sep-12 Dec-12 Survey started Mar-12 Jun-12 Jun-12 Jul-12 Aug-12 Sep-12 Oct-12 Aug-12 Aug-12 Aug-12 Aug-12 Aug-12 Oct-12 Sep-12 Dec-12 Survey finished Data entered Data delivered May-12 Aug-12 Sep-12 Aug-12 Sep-12 Oct-12 Nov-12 Sep-12 Nov-12 Sep-12 Nov-12 Aug-12 Nov-12 Oct-12 Jan-13 May-12 Oct-12 Oct-12 Sep-12 Oct-12 Oct-12 Feb-13 Oct-12 Oct-12 Oct-12 Oct-12 Sep-12 Nov-12 Oct-12 Jan-13 Jun-12 Dec-12 Nov-12 Oct-12 Oct-12 Nov-12 Feb-13 Nov-12 Dec-12 Nov-12 Jan-13 Sep-12 Dec-12 Dec-12 Jan-13 Contact person IMPACTlite team facilitator Joash Mango Ibrahim Wanyama Ibrahim Wanyama George Sayula Ianetta Mutie Solomon Desta Jesse Naab Leopold Some Diakité Lamissa Yacine Ndour Moussa Boureima Dhiraj Singh Dhiraj Singh Dhiraj Singh Dhiraj Singh Joash Mango Joash Mango Joash Mango Joash Mango Joash Mango Mariana Rufino Abdoulaye Moussa Sabine Douxchamps Abdoulaye Moussa Sabine Douxchamps Silvia Silvestri Nils Teufel Nils Teufel Nils Teufel Nils Teufel 5 Data analysis 5.1 Calculation of performance and livelihood indicators Once the data was processed, cleaned and documented, we generated scripts to calculate a number of performance and livelihood indicators (Table 5). These indicators are appended to the database and will be used for within and across site comparisons, and for modeling studies. 18 Table 5: List of indicators calculated using variables extracted from the household survey database. Variable name Description Incomeland Income from Leasing land out ($) Expense land Cost of leasing land out ($) Cropincome Gross income from sale of crops ($) Croplabour Cost of hiring labour ($) Cropinputs Cost of crop inputs used on farm (e.g seeds) ($) Liveincome Income from sale of livestock products ($) Liverumsold Income from sale of livestock species ($) Livefeed Cost of livestock feed used ($) Liveinputs Cost of livestock inputs used (e.g. supplements) ($) Livelabour Cost of labour for livestock ($) Otherincome Off-farm income (e.g remittances) ($) Otherexpense Household expenses (e.g. school fees) ($) Totalarea Total area of farm (ha) Totalplots Total area of plots (ha) Totalcrops Total area of crops (ha) Totalhhsize Number of household members Totalcattle Total cattle (no’s per household) Totalgoats Total goat (no’s per household) Totalsheep Total sheep (no’s per household) Totalfamcrplabour Family labour used on crops (Mandays per season) Totalfamlvlabour Family labour used on livestock keeping (Mandays per season) 5.2 Income calculations Total net income, cash income, non-cash income and off-farm income for the household were calculated using revenues from livestock, crops, value of consumed food products and as shown in equations (1), (2), (3) and (4). Total net income (1) where: is total annual income for household i Lsale is annual income from livestock sales Crsale is annual income from crop sales VP is the annual monetary value of consumed farm produce Lcost are the annual direct costs of livestock production CrCost are the annual direct costs of crop production 19 Cash income (2) where: is the annual cash income for household Lsale is the annual income derived from livestock sales Crsale is the annual income from crop sales Non-cash income (3) where is the annual non-cash income for household Off-farm income (4) was the sum of the cash earned from all j off-farm activities the household members are engaged in: ∑ (4) where Off_inci is the annual off-farm income of household i Off_farm incomej is the revenue from each j off-farm activities Poverty line at household level (5) was calculated by considering household size, an income of USD 1.25 per capita per day, and a conversion rate of 1USD= 95 Kenyan shillings. (5) where PovLinei is the poverty line expressed as annual income in Kenyan shillings for household i HHsizei is the number of members of household i 5.3 Crop diversity and activity diversity Crop diversity (6) was the maximum number of crops grown by the households in a site. (6) where CropDivk is the crop diversity for site k 20 Max NumCropsi is the maximum number of crop grown by the i households at site k Activity diversity (7) is the maximum number of farm and non-farm activities households are engaged in a site. (7) where ActDivk is the activity diversity for site k Max NumActi is the maximum number of activities in which the i households are engaged at site k 5.4 Food security and food self-sufficiency Energy availability was calculated for each household based on production data and food consumption. Households reported food items produced on-farm and those purchased on a weekly basis, indicating seasonal differences. With this information we calculated a food security ratio (FSR) as shown in equation (8) and a food self-sufficiency ratio (FSSR) as shown in equation (9) to reflect the reliance on farm production and purchases to meet energy needs, calculated using World Health Organisation standards. FSR includes total energy in available food (purchased and on-farm produce) divided by total energy requirements for the household family. FSSR is total energy in on-farm produce divided by total energy requirements for the household family. ∑ (∑ ∑ (∑ ) ) (8) (9) where: FSRi is the food security ratio for household i FSSRi is the food self-sufficiency ratio for household i QtyCm is the quantity of food item m produced on-farm that is available for consumption (kg or litre) QtyPm is the quantity of food item m purchased that is consumed (kg or litre) Em is the energy content of food item m (MJ kg-1 or litre) Kj is the energy requirements in MJ per capita for j member n is the number of members in household j 5.5 Household diet diversity score The HDDS is the sum of all food groups consumed by the household divided by the total possible food types. The types of food considered were categorized to 8 main food groups: Main staples, vegetables, fruits, pulses, meat and fish, milk, oil and sugar (Table 6). 21 Table 6: Food groups used to describe diet items in the analysis of food consumption Types of foods A. Groups Main Staples C. Millet, sorghum, maize, rice, wheat or other local grains e.g. ugali, bread, rice Potatoes, yams, cassava or foods made from roots or tubers Vegetables D. Fruits Fruits E. Beans, peas, lentils or nuts Pulses F. G. H. I. J. Red meat, pork, lamb, goat, rabbit, wild game Poultry including chicken and duck Eggs Fresh or dried fish or shell fish Milk, cheese, yoghurt or other milk products Meat and Fish K. Oils and fats Oil L. Sugar,sweets,honey Sugar B. Vegetables Milk Source: World Food Programs vulnerability assessment mapping (WFP, 2008) If the household consumed a food type, then it had a score of 1, otherwise zero. ∑ 5.6 Household domestic asset index The asset index analysis is adapted from analyses recommended for all Bill and Melinda Gates funded projects. It is calculated for all movable assets. Each of the assets is assigned a weight (w) and then adjusted for age (Agricultural Development Outcome indicators, 2010). ∑ ∑ where, W= weight of the ith item of asset g N=number of asset g owned by household A=age adjustment to weight G= number of assets owned by household 6 Data management and databases 6.1 Data management 22 The data management process relied on a software system called CSPro for entering and editing the survey data. Data entry was performed at each site by one member of the team. On completion of data entry each site submitted a DAT file to ILRI headquarters in Nairobi for processing, transforming and storing the data into a standard MySQL database. The DAT passed through a series of automated processes generating error log files describing diverse problems with the data. Each tier of processing was followed by constant interaction between ILRI staff and the site team to resolve the problems. The result is 15 consistent and standard1 databases storing detail information of 3000 households. The following diagram shows the data management process. 6.2 Database schema The data for each site is stored in a MySQL database in 22 tables with 27 supporting lookup tables. The following images show a graphical representation of the database separated in three main sections: a) Crops, containing the generic information of the household plus data on crop, plots, management activities and crop production; b) Livestock, containing information about livestock numbers, management activities and production and; c) Other, containing information about other sources of income and expenses plus household consumption patterns. 1 Surveys in the IGP region are almost standard with the rest except for the household consumption forms. 23 Crops section 24 Livestock section 25 Other expenses/income and consumption sections 6.3 Data dictionary and supporting documents 6.3.1 Data dictionary A data dictionary describes the different tables and fields in the database. This information can be used to produce queries or reports. List of tables Table actninputs cropprods croprprods cropsinv farmassets hhconsoff hhconsper hhconsprods hhsize landsplots landusage livestock Description Activities and inputs Production of main crops Residue production of main crops Inventory of crops, trees and aquaculture products Farm and domestic assets Household consumption off-farm products Household consumption periods Household consumption on-farm products Household Size Subplots Land usage Non ruminant livestock species Lookup table No No No No No No No No No No No No 26 lkpactactoin lkpasset lkpcountry lkpcrpprdprod lkpcrprsprcode lkpcrpsrpgrz lkpexpcode lkpfeedcode lkpfeedprod lkpfeedseason lkpfeedsrc lkpgender lkphhcoffperiod lkphhconprod lkphhcontime lkpincome lkpincperiod lkplandcrops lkplanguage lkpliveid lkplvactactoin lkplvprodprd lkpplotcrop lkpplotown lkpprodsys lkprumbreed lkprumcat lvstact lvstactinputs lvstfeed lvstprods mainsurveyinfo otherexpen otherincome plotinputs rumentext ruminants Activity / input Assets Country Product Residue Grazing Expense Feed Feed product (if source is on-farm) Season Feed source Gender of household head Frequency period Product Timing (Good / Bad) Income Period Taking most of the land Language used in interview Livestock Activity / input Product Crop / aquaculture species Land ownership Production System Breed Category Livestock activities Livestock inputs Livestock feeding Livestock production Main Survey Information Other expenses Other income Crop activity inputs Ruminant enter and exit Ruminant livestock species Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No No No No No No No No No No List of columns in each table Table actninputs actninputs actninputs actninputs actninputs actninputs Column actactoin acthlcost actlabf actlabh actlabm actlabnotes Description Activity / input Hired labour cost (Ksh / hour) Labour (Female) Labour (Hired) Labour (Male) Notes on labour Data type varchar(4) decimal(7,2) int(2) int(2) int(2) varchar(30) Primary Key Yes No No No No No 27 actninputs actninputs actninputs actninputs actninputs actnotes actothcosts actplot actsplot acttdays actninputs actninputs actninputs actninputs actninputs actninputs cropprods cropprods cropprods cropprods cropprods cropprods cropprods cropprods cropprods cropprods cropprods cropprods cropprods cropprods croprprods croprprods croprprods croprprods croprprods croprprods croprprods croprprods croprprods croprprods croprprods croprprods croprprods croprprods croprprods croprprods croprprods acttdaysf acttdaysh actthrs actthrsf actthrsh HH_ID crpprdcrop crpprdhqty crpprdlqty crpprdnotes crpprdoqty crpprdplot crpprdpqty crpprdprod crpprdsplot crpprdspri crpprdsqty crpprdwho HH_ID record_id crprspplot crprsprcode crprspsplot crpsrpamtg crpsrpamtl crpsrpcqty crpsrpcrop crpsrpgrz crpsrpleft crpsrplqty crpsrponot crpsrpoqty crpsrpspri crpsrpsqty crpsrpwho HH_ID record_id cropsinv cropsinv cropsinv cropsinv cropsinv hhcons HH_ID invnotes labour land Notes Oher costs Plot Subplot Male frequency (Days / month) Female frequency (Days / month) Hired frequency (Days / month) Male time (Hours / day) Female time (Hours / day) Hired time (Hours / day) household id Crop Consumption (kg) Livestock feeding (kg) Notes Other use (kg) Plot Production (kg) Product Subplot Sale price (Ksh / kg) Sales (kg) Who controls the income household id Record ID Plot Residue Subplot Amout grazed (kg) Amout left (kg) Collected (kg) Crop Grazing Left on field Livestock feeding (kg) Notes Other use (kg) Sale price (Khs / kg) Sale (kg) Who controls the income household id Record ID Most important for household consumption household id Notes Taking most of the labour Taking most of the land varchar(30) varchar(20) int(2) int(2) decimal(5,2) No No Yes Yes No decimal(5,2) decimal(5,2) decimal(5,2) decimal(5,2) decimal(5,2) varchar(28) int(4) decimal(7,2) decimal(7,2) varchar(30) decimal(7,2) int(2) decimal(7,2) int(4) int(2) decimal(7,2) decimal(7,2) int(4) varchar(28) int(3) int(2) int(4) int(2) decimal(7,2) decimal(7,2) decimal(7,2) int(4) int(4) int(4) decimal(7,2) varchar(30) decimal(7,2) decimal(7,2) decimal(7,2) int(4) varchar(28) int(3) No No No No No Yes No No No No No No No No No No No No Yes Yes No No No No No No No No No No No No No No No Yes Yes varchar(4) varchar(28) varchar(30) varchar(4) varchar(4) No Yes No No No 28 cropsinv cropsinv farmassets farmassets farmassets farmassets farmassets farmassets farmassets farmassets farmassets farmassets record_id sales asset35years asset3years assetfemale assetjoint assetmale assetmore3 assetnotes assetnown asset_cod HH_ID hhconsoff hhconsoff hhconsoff hhconsoff hhconsoff hhconsoff hhconsoff hhconsoff hhconsoff hhcofffreq hhcoffitem hhcoffperiod hhcoffprice hhcoffqtyp hhcofftime hhconnotet hhconperioe hhconqtz hhconsoff hhconsoff hhconsoff hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconsper hhconweel HH_ID record_id bparp bpaug bpdec bpfeb bpjan bpjul bpjun bpmar bpmay bpnov bpoct bpsep gpapr gpaug gpdec gpfeb gpjan gpjul gpjun gpmar gpmay gpnov gpoct Record ID Most important for sales 3 to 7 years < 3 years Number owned by female Number owned jointly Number owned by male > 3 years Notes Total number owned Asset code household id Frequery of purchase (times / period) Item Frequency period Purchase price (Ksh / kg) Quanity purchased (kg) Timing (Good / Bad) Notes Consumption period (months) Quantity consumed (kg) How often in a week (days / week) household id Record ID April August December February January July June March May November October September April August December February January July June March May November October int(3) varchar(4) int(5) int(5) int(5) int(5) int(5) int(5) varchar(30) int(5) int(4) varchar(28) Yes No No No No No No No No No Yes Yes decimal(7,2) varchar(30) int(4) decimal(9,2) decimal(7,2) int(4) varchar(30) decimal(7,2) decimal(7,2) No No No No No No No No No decimal(7,2) varchar(28) int(3) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) int(1) No Yes Yes No No No No No No No No No No No No No No No No No No No No No No No 29 hhconsper hhconsper hhconsprods hhconsprods hhconsprods hhconsprods hhconsprods hhconsprods hhconsprods hhconsprods gpsep HH_ID hhconitem hhconnotes hhconperiod hhconplot hhconprod hhconqty hhconsplot hhcontime hhconsprods hhconsprods hhconsprods hhsize hhsize hhsize hhsize hhsize hhsize hhsize hhsize hhsize hhsize hhsize hhsize hhsize hhsize landsplots landsplots landsplots landsplots landsplots landsplots landsplots landsplots landsplots landsplots landusage landusage landusage landusage landusage landusage landusage landusage livestock livestock livestock hhconweek HH_ID record_id hhgender hhnotes hhoffwdesc hhsea1off hhsea1on hhsea2off hhsea2on hhsea3off hhsea3on HH_ID persid poshh record_id yearb HH_ID plotcrop plotcropname plotend plotinter plotnotes plotsparea plotstart plotsubplt plot_id HH_ID plotarea plotcost plotgender plothhd plotnotes plotown plot_id HH_ID liveafemales liveamales September household id Crop, livestock or aquaculture Notes Consumption period (months) Plot Product Quantity consumed (kg) Subplot Timing (Good / Bad) How often in a week (days / week) household id Record ID Gender Notes Off-farm work description Long rains off-farm Long rains on-farm Short rains off-farm Short rains on-farm Dry spell off-farm Dry spell on-farm household id Person ID Position in Household Record ID Year of birth Household id Crop code Name of variety End month Inter-cropping Notes Area of subplot Start month Subplot id Plot id household id Area (acres) Cost / return (Ksh if leased) Gender ownership Distance to HH (metres) Notes Land ownership Description of plot_id household id Adult females Adult males int(1) varchar(28) varchar(4) varchar(30) decimal(5,2) int(2) varchar(4) decimal(7,2) int(2) int(4) No Yes No No No No No No No No int(5) varchar(28) int(3) int(4) varchar(20) varchar(30) int(1) int(1) int(1) int(1) int(1) int(1) varchar(28) varchar(4) varchar(30) int(3) int(4) varchar(28) int(4) varchar(15) int(4) int(1) varchar(20) decimal(7,2) int(4) int(2) int(3) varchar(28) decimal(7,2) decimal(7,2) int(4) decimal(7,2) varchar(20) int(4) int(3) varchar(28) int(5) int(5) No Yes Yes No No No No No No No No No Yes No No Yes No Yes Yes No No No No No No Yes Yes Yes No No No No No No Yes Yes No No 30 livestock livestock livestock livestock livestock livestock livestock lkpactactoin lkpactactoin lkpasset lkpasset lkpcountry lkpcountry lkpcrpprdprod lkpcrpprdprod lkpcrprsprcode lkpcrprsprcode lkpcrpsrpgrz lkpcrpsrpgrz lkpexpcode lkpexpcode lkpfeedcode lkpfeedcode lkpfeedprod lkpfeedprod lkpfeedseason lkpfeedseason lkpfeedsrc lkpfeedsrc lkpgender lkpgender lkphhcoffperiod lkphhcoffperiod lkphhconprod lkphhconprod lkphhcontime lkphhcontime lkpincome lkpincome lkpincperiod lkpincperiod lkplandcrops lkplandcrops lkplanguage lkplanguage lkpliveid lkpliveid lkplvactactoin livegender liveid livenotes livenum liveyfemales liveymales record_id lkpactactoin_cod lkpactactoin_des asset_cod asset_des lkpcountry_cod lkpcountry_des lkpcrpprdprod_cod lkpcrpprdprod_des lkpcrprsprcode_cod lkpcrprsprcode_des lkpcrpsrpgrz_cod lkpcrpsrpgrz_des lkpexpcode_cod lkpexpcode_des lkpfeedcode_cod lkpfeedcode_des lkpfeedprod_cod lkpfeedprod_des lkpfeedseason_cod lkpfeedseason_des lkpfeedsrc_cod lkpfeedsrc_des lkpgender_cod lkpgender_des lkphhcoffperiod_cod lkphhcoffperiod_des lkphhconprod_cod lkphhconprod_des lkphhcontime_cod lkphhcontime_des lkpincome_cod lkpincome_des lkpincperiod_cod lkpincperiod_des lkplandcrops_cod lkplandcrops_des lkplanguage_cod lkplanguage_des lkpliveid_cod lkpliveid_des lkplvactactoin_cod Gender ownership Livestock Notes Total number of animals Young females Young males Record ID Lookup field code Lookup field description Description of asset_cod Description of asset_des Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code int(4) int(4) varchar(30) int(5) int(5) int(5) int(3) varchar(4) varchar(120) int(4) varchar(120) int(4) varchar(120) int(4) varchar(120) int(4) varchar(120) int(4) varchar(120) int(4) varchar(120) varchar(4) varchar(120) varchar(4) varchar(120) int(4) varchar(120) int(4) varchar(120) int(4) varchar(120) int(4) varchar(120) varchar(4) varchar(120) int(4) varchar(120) int(4) varchar(120) int(4) varchar(120) varchar(4) varchar(120) int(4) varchar(120) int(4) varchar(120) varchar(4) No No No No No No Yes Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes 31 lkplvactactoin lkplvprodprd lkplvprodprd lkpplotcrop lkpplotcrop lkpplotown lkpplotown lkpprodsys lkpprodsys lkprumbreed lkprumbreed lkprumcat lkprumcat lvstact lvstact lvstact lkplvactactoin_des lkplvprodprd_cod lkplvprodprd_des lkpplotcrop_cod lkpplotcrop_des lkpplotown_cod lkpplotown_des lkpprodsys_cod lkpprodsys_des lkprumbreed_cod lkprumbreed_des lkprumcat_cod lkprumcat_des HH_ID lvactactoin lvactdmnt lvstact lvstact lvstact lvstact lvstact lvstact lvstact lvstact lvstact lvstact lvstact lvstact lvstact lvstact lvstactinputs lvstactinputs lvstactinputs lvstactinputs lvstactinputs lvstactinputs lvstfeed lvstfeed lvstfeed lvactdmntf lvactdmnth lvacthhday lvacthhdayf lvacthhdayh lvacthl lvactid lvactnotes lvactothcosts lvactownlf lvactownlm lvactprice lvactsea record_id HH_ID lvactactoin lvactinamnt lvactinprice lvactnotes record_id feedamt feedcode feedliveid lvstfeed lvstfeed lvstfeed lvstfeed lvstfeed feedmkval feednotes feedothcost feedperiod feedplt lvstfeed lvstfeed feedppaid feedprod Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description Lookup field code Lookup field description household id Activity / input Male frequency (days / month) Female frequency (days / month) Hired frequency (day / month) Male time (hours / day) Female time (hours / day) Hired time (hours / day) Hired labour (No. people) Livestock / livestock category Notes Other costs Own labour (No. Females) Own labour (No. Males) Cost of hired labour (Ksh / day) Season Record ID Description of HH_ID Input code Input amount Input price Notes Record number Quantity (kg) Feed Livestock / livestock category Estimated market value (Ksh / kg if in-farm) Notes Other costs (Ksh) Feeding period (months) Plot (if source is in-farm) Price paid (Ksh / kg if purchased) Feed product (if source is on- varchar(120) int(4) varchar(120) int(4) varchar(120) int(4) varchar(120) varchar(4) varchar(120) int(4) varchar(120) int(4) varchar(120) varchar(28) varchar(4) decimal(7,2) No Yes No Yes No Yes No Yes No Yes No Yes No Yes No No decimal(7,2) decimal(7,2) decimal(7,2) decimal(7,2) decimal(7,2) int(5) varchar(20) varchar(30) varchar(30) int(5) int(5) decimal(7,2) int(4) int(3) varchar(28) varchar(4) decimal(7,2) decimal(7,2) varchar(30) int(3) decimal(7,2) varchar(4) varchar(20) No No No No No No No No No No No No No Yes Yes Yes No No No Yes No No No decimal(7,2) varchar(30) decimal(7,2) decimal(7,2) int(2) No No No No No decimal(7,2) varchar(4) No No 32 lvstfeed lvstfeed lvstfeed feedseason feedsplt feedsrc lvstfeed lvstfeed lvstfeed lvstprods lvstprods lvstprods lvstprods lvstprods lvstprods lvstprods lvstprods lvstprods lvstprods lvstprods lvstprods mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo mainsurveyinfo feedweek HH_ID record_id HH_ID lvprodhc lvprodid lvprodnotes lvprodot lvprodprd lvprodpri lvprodprod lvprodsale lvprodsea lvprodwho record_id country date day division elevation enumname gender hheadname HH_ID language latcard latdeg latmin location longcard longdeg longmin month prodsys province rankingnotes resp1gen resp1name resp2gen resp2name state sublocation supname farm) Season Subplot (if source is in-farm) Feed source How often in a week (days / week) household id Record ID household id Consumption (kg) Livestock / livestock category Notes Other (kg) Product Sale price (Ksh / kg) Production (kg) Sales (kg) Season Who controls the income Record ID Country Date Day Division Elevation Numerator's name Gender or household head Household head Household id Language used in interview Latitude cardinal Latitude degrees Latitude minutes Location Longitude cardinal Longitude degrees Longitude minutes Month Production system Province Crop ranking notes Gender 1st respondent Name of 1st respondent Gender 2nd respondent Name of 2nd respondent State Sub location Supervisor name int(4) int(2) int(4) No No No int(5) varchar(28) int(3) varchar(28) decimal(7,2) varchar(20) varchar(30) decimal(7,2) int(4) decimal(7,2) decimal(7,2) decimal(7,2) int(4) int(4) int(3) int(4) int(8) int(2) varchar(30) int(6) varchar(30) int(4) varchar(30) varchar(28) int(4) varchar(1) int(3) int(6) varchar(30) varchar(1) int(3) int(6) int(2) varchar(4) varchar(30) varchar(60) int(4) varchar(30) int(4) varchar(30) varchar(30) varchar(30) varchar(30) No Yes Yes Yes No No No No No No No No No No Yes No No No No No No No No Yes No No No No No No No No No No No No No No No No No No No 33 mainsurveyinfo mainsurveyinfo otherexpen otherexpen otherexpen otherexpen otherexpen otherexpen otherexpen otherexpen otherincome otherincome otherincome otherincome otherincome otherincome otherincome otherincome otherincome plotinputs plotinputs plotinputs plotinputs plotinputs plotinputs plotinputs plotinputs rumentext rumentext rumentext rumentext rumentext rumentext rumentext rumentext rumentext rumentext rumentext rumentext rumentext rumentext rumentext ruminants ruminants ruminants ruminants ruminants ruminants village year expamnt expcode expnotes expperiod expsea exptimes HH_ID record_id HH_ID incamnt incgender incnotes income incperiod incsea inctimes record_id actinamount actincost actnotes actplot actsplot act_id HH_ID input_id HH_ID rumborn rumcons rumdead rumgin rumgout ruminant_id rumnotes rumnumly rumothers rumprch rumpresly rumprice rumsold rumsoldp HH_ID rumage rumageto rumbreed rumcat rumgender Village Year Amount (Ksh / season) Expense Notes Period Season Times household id Record ID household id Amount (Ksh / season) Gender ownership Notes Income Period Season Times Record ID Input amount Input cost Notes Plot Subplot Activity Household id Input code household id Born Consumed Dead Gifts in Gifts out Ruminant ID Notes Total number last year Others Purchased Present last year? Price (Ksh / animal) Sold Price (ksh / animal) household id Age From Age To Breed Category Gender ownership varchar(30) int(4) decimal(9,2) int(4) varchar(30) int(4) int(4) int(5) varchar(28) int(3) varchar(28) decimal(9,2) int(4) varchar(30) int(4) int(4) int(4) int(5) int(3) decimal(7,2) decimal(7,2) varchar(30) int(2) int(2) varchar(4) varchar(28) varchar(4) varchar(28) int(5) decimal(7,2) int(5) int(5) int(5) int(2) varchar(30) int(5) decimal(7,2) int(5) int(4) decimal(7,2) int(5) decimal(7,2) varchar(28) int(5) int(5) int(4) int(4) int(4) No No No No No No No No Yes Yes Yes No No No No No No No Yes No No No Yes Yes Yes Yes Yes Yes No No No No No Yes No No No No No No No No Yes No No No No No 34 ruminants ruminants ruminants ruminants ruminants ruminants ruminants ruminants ruminants ruminants ruminants ruminants rumid rumlact rumlactlen rumoffnum rumpreg rumsex rumspe rumtnum rumtrac rumwea rumyldm rumyldp Ruminant ID No. Lactating Lactation length (months) Number of offsprings in a period No. Pregnant Sex Species Total number Used for traction Weaning age (months) Milk yield min (l / day) Milk yield peak (l / day) int(2) int(5) decimal(7,2) varchar(15) int(5) int(4) int(4) int(5) int(4) decimal(7,2) decimal(7,2) decimal(7,2) Yes No No No No No No No No No No No 6.2.2 Reference to paper survey Although the data dictionary explains every column in the database it is usually quite difficult to relate the data in a database with the paper survey. To minimize this Impact Lite provides a reference to paper survey. Each page of the printed survey indicates the table in the database where such data is stored plus the name of each column. Information on the related lookup tables is also provided. The following images are examples. 35 7 Future steps 7.1 Data standardization and documentation As soon as datasets are delivered they undergo a consistency check, a data cleaning process, and once this is finished we calculate the performance and livelihood indicators. This has been done for a number of CCAFS sites already. The datasets will be uploaded to the CCFAS site and properly documented latest by the end of 2013. 7.2 Production systems and farm typology Using a number of key indicators and variables collected we will build farm typologies for all 15 sites. The farm typologies will be used for within and across site comparisons, and for modeling studies to test interventions. Proposed methods: Select variables, production system, household characteristics, market access etc. Then classify (e.g. cluster analysis). Create class variable(s), share for comments. Classification results will probably not be final immediately; several class variables (for production system and farm type) will allow teams to test them for their theoretical soundness and for the usefulness of the results. The categorization into production systems done a priori will be verified because it is expected that that this definition helps to determine for which wider populations results are relevant (recommendation domains). There are various possibilities: several production systems in each site and several sites per production system. Proposed methods: Select variables (e.g. income share from dairy, cereal share of land, off-farm income, livestock income, cropping intensity etc.); classification (e.g. cluster analysis); check geographical distribution of household classes in research grids with coordinates; if they are contiguous this might indicate a production system. When selecting variables think of how they would compare to secondary data available for determining production systems (in order to find the recommendation domains). Size of classes (production systems) depends on the scale. If only one site is considered, smaller classes could be included than if the classification is performed over all three regions (15 sites). 7.3 Modelling The IMPACTlite team gathered at the beginning of 2013 to plan a number of studies based on the datasets collected. We envisage at least 4 analysis including 3 regional comparisons of performance and livelihood status, and one across site comparison. The team will gather information on plausible scenarios of change to perform exploratory modeling studies. In consultation with the regional CCAFS programmes, the IMPACTlite team would like to explore opportunities for collaboration with local partners in the development of local scenarios of change, and on training of local scientists in the use of simple modeling techniques. 36 8 References Herrero, M, E. González-Estrada, P.K. Thornton, C. Quiros, M.M. Waithaka, R. Ruiz, G. Hoogenboom, IMPACT: Generic household-level databases and diagnostics tools for integrated crop-livestock systems analysis. Agricultural Systems 92, 240-265. Quiros, C., M.C. Rufino, M. Herrero. 2011. Developing generic tools for characterizing agricultural systems for climate and global change studies. International Livestock Research Institute, Nairobi, Kenya. Report of Activities 2011 Submitted by ILRI to the CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS) 31st January 2011. M.C. Rufino, C. Quiros, N. Teufel, S. Douxchamps, S. Silvestri, J. Mango, A.S. Moussa, M. Herrero. 2012. Household Characterization Survey – IMPACTlite Training Manual. CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS) Working document, December 2012. 37 9 Annex 9.1 Village locations in research grids in IGP Figure 1: Vaishali research grid with village locations 38 Figure 2: Rupandehi grid with village locations and exclusion zone for Bhairawa town 39 Figure 3: Karnal grid with village locations and exclusion zones for Traoria and Nilokheri towns 40 Note: Marker colour indicates dominant production system (paddy-yellow, aquaculture-blue, vegetables-green, mixed-white); dependent hamlets are marked red with a "T". Figure 4: Bagerhat grid with village locations and their dominant production systems 41 9.2 Village locations in research grids in Senegal Fig. 5. Map of the production systems in Kaffrine 42