Chapter 9 Sampling, structuring, documenting and presenting information for action plans 133
Chapter 9 Sampling, structuring, documenting and
presenting information for action plans (with the collaboration of D. Louette, P. Mathur, P. Quek and I. Thormann)
9.0 Objectives 134
9.1 Obtaining a representative sample 134
9.1.1 Choosing a representative sampling size 134
9.1.2 Sampling strategy: random, systematic or stratified 135
9.1.3 Structuring multidisciplinary information for sampling – the number problem 135
9.1.4 Sampling over time 137
9.2 Collecting and structuring information to support in situ
conservation on-farm 137
9.2.1 Determining the amount and distribution of genetic diversity
maintained on-farm 140
9.2.2 Studying the processes used for maintaining and managing diversity 141
9.2.3 Determining the factors that influence farmers to maintain diversity 142
9.2.4 Determining who maintains diversity 143
9.3 Documentation for in situ conservation on-farm 143
9.3.1 Data recording and verification 144
9.3.2 Written records 144
9.3.3 Visual records 145
9.3.4 Triangulation 145
9.4 Returning information to the community 145
9.4.1 Community information systems 145
9.5 Using information for action plans for on-farm conservation 147 9.6 References 147
9.7 Recommended reading 148
A Training Guide for In Situ Conservation On-farm: Version 1 134
By the end of this module, readers should have an understanding of how to:
; Carry out representative sampling over space and time
; Structure information to answer key questions related to in situ conservation on-farm
; Document in situ conservation efforts and return information to the community
; Present information for conservation and development action plans.
9.1 Obtaining a representative sample
Appropriate sampling is important to ensure that adequate representation of the situation or site is reflected in the information collected. Regardless of the approach chosen, some part of the diversity in human, environmental and genetic diversity factors will not be sampled. Sampling strategies must consider resource constraints along with scientific needs.
9.1.1 Choosing a representative sampling size
How many subsamples of households, plots, and plants are necessary to have a representative sample of the site or population in question? The objective is to determine the smallest number of samples to adequately characterize the region in question. Sampling size will depend on the amount of variation among samples. A larger sample size will give more information on the variation between samples than will a smaller sample. Thus, the more homogenous the population, be it in terms of household characteristics, field soil types or variety populations, the less the need will be for larger sample sets. This concept or trend in the decrease in new information as more and more, or larger and larger, samples are included is often used by plant ecologists in a technique called the species-area curve to
determine the minimum size and number of samples that will be representative of a population. A more general form of the relationship is known as the law of diminishing returns.
A species-area curve is created by sampling the area in a system of nested plots (larger and larger plots arranged on the ground such that each successive plot encompasses all of the previous plot) (Barbour et al. 1987). After a certain point, very few new species are found in each successive plot.
10Number of species
1 2 3 4 5 6 7 8 9 1011Plot area
Besides this nested method of determining sampling sizes, decisions are often made through consulting existing information (e.g. scientific literature) regarding the population.
Chapter 9 Sampling, structuring, documenting and presenting information for action plans 135
For example, social scientists commonly sample 5–10% of the households in a village when
they know there is significant variation in household characteristics. For the IPGRI Global project on in situ conservation on-farm, population geneticists have agreed that a minimum of 30 individuals per population of a variety should be sampled (Jarvis and Hodgkin 1998; see Chapter 5, this Training Guide). In addition, under the "Recommended reading" section of this chapter we list some key references from different disciplines that can guide the reader on determining appropriate sampling size.
9.1.2 Sampling strategy: random, systematic or stratified
A key decision to make is whether samples should be taken in a random or systematic
manner (Causton 1988). In random sampling every point in the field site has an equal chance to be sampled. In systematic sampling, transects or grids are used to take samples at regular intervals. This has the advantage of avoiding the over-sampling of “uninteresting”
areas at the expense of “interesting” ones. However, consideration is required when selecting
a sampling approach, because sampling can have a profound impact on study results; poor sampling choices can lead to results that may be considered invalid. Potential biases, particularly when choosing non-random samples, require special attention.
The primary purpose of random sampling is to detect and assess correlation between distributions of one factor in relation to another factor. One might want to know how the distribution of genetic diversity relates to certain social, economic, biological, environmental and other factors in the community. Does wealth status relate to any level of genetic diversity in the community? Do farmers with more fragmented land contain more diversity? These questions require a random sample of households across the landscape. Random samples of farmers' fields throughout the site also may be appropriate for gaining an idea of the range and diversity of the main abiotic and biotic factors affecting crop diversity.
Although it is the most statistically robust method, the placement of random samples is particularly time consuming. Regular or systematic placement of samples is the simplest method to carry out, but may be less appropriate for statistical analysis. A compromise method is stratified random sampling. In this method, the area to be sampled is first
systematically subdivided into relatively homogenous sections, and then randomly placed samples are taken in each subsection (see Barbour et al. 1987; Kershaw and Looney 1985;
Gauch 1982; Greig-Smith 1983). Stratified random sampling is often used once an area has already been stratified into different areas. Criteria are used to divide the study site/population into various strata, from each of which a certain number of samples are drawn randomly; criteria for stratification may be socioeconomic, agronomic or environmental in nature.
9.1.3 Structuring multidisciplinary information for sampling – the number
Stratifying to structure date becomes extremely important when more than one discipline is involved in a sampling procedure. For example: a community or village containing 1000 households is selected as a study site. Ten percent or 100 households of the village are sampled for social and economic variables. In each household, a farmer grows four target crops. For each crop, the farmer manages an average of three varieties grown in a particular field with certain soil and topographical characteristics. For each variety a minimum of 30 samples will be needed for an adequate study of the population. Thus, in order to compare household characteristics, plot characteristics and varieties, we end up collecting 30 samples × 3 varieties × 4 crops × 100 households or 36 000 samples for analysis. Of course this number of samples is unreasonable given time and resource constraints. How is this problem solved? One method is to return once again to the key questions that the study is trying to answer – what is the amount and distribution of genetic diversity maintained on-
A Training Guide for In Situ Conservation On-farm: Version 1 136
farm, by what processes is this diversity being maintained, by whom and why – to structure
the data collection.
Chapter 9 Sampling, structuring, documenting and presenting information for action plans 137
What has been useful in ongoing on-farm work is to first carry out a random sample of the households in a village to have an idea of the range of social, economic, cultural and farming management practices. At the same time, it will enable a preliminary idea of the total range of abiotic and biotic factors in the sites and the total range of genetic variation within the village. Where topography is an important factor, it may be useful to run a transect across north- and south-facing slopes and randomly sample households along this transect to have an idea of the entire range of agroecological factors for the site. Likewise, it may be useful to collect samples of all the landraces in the village for preliminary characterization to have an idea of the total range of variety diversity.
Once the range of (1) household and village social, economic and cultural characteristics (Chapter 2), farming management practices (Chapter 3), abiotic and biotic factors (Chapter 3), and genetic diversity of target crops (Chapters 4 and 5) is known, this information must be structured, or stratified to make comparisons among groups of different data types.
9.1.4 Sampling over time
On-farm conservation of crop diversity is influenced by agroecological, socioeconomic and genetic processes with different directions and rates of change over time. Some factors are stable, meaning that they do not change over time, at least in the short term, such as the soil, parent rock or climate. Rainfall may fluctuate annually, and market infrastructure may develop rapidly or gradually. It will be crucial for on-farm conservation efforts to study how changes in these factors over time might affect crop genetic diversity. For annual or biennial plants, sampling over time may be feasible within the lifetime of one research project, but this may not be possible for perennial species. For such long-lived species, spatial proxies for time will need to be used to understand how these different factors may influence genetic diversity over time.
When data are collected at different times, time-series analysis methods can be used to
examine relationships between variables over time (Kendall and Ord 1990). Time-series analysis is based on the idea that measurable variables that are observed continuously may be regarded as information signals. Sampling this signal at different intervals produces a discrete signal, or time series (1996-2000 Finney:
http://www.chaos.engr.utk.edu/CTSA.html). 1985). An autocorrelation coefficient is
determined as a measure of the similarities of measurements that are separated by a particular time interval, while a cross-correlation coefficient is used for detecting patterns of
variation between variables over time. For comparing frequencies of events over time, a common statistical tool is power-spectral analysis. Power-spectral analysis is used to
determine periodicities within the data by giving an indication of the different frequencies over time of variation, which account for most of the variability in the data. All these tools usually require numerous data points or sampling occasions. In some cases, it may be interesting to calculate transition probabilities among classes (e.g. identify land-use types in different aerial photographs of the study area and then use Markov modelling to simulate changes over time).
9.2 Collecting and structuring information to support in situ
In Chapters 2 through 6, we saw how different categories of information are needed to answer key questions important to in situ conservation on-farm. In Chapters 7 and 8 we
discussed creating a framework, selecting sites and sensitizing the community. Once these aspects are in place, partners in an on-farm conservation programme are in a position to collect and structure information that will support in situ conservation on-farm under the
A Training Guide for In Situ Conservation On-farm: Version 1 138
four major topics discussed throughout this guide:
1. The amount and distribution of genetic diversity being maintained on-farm. 2. The processes being used to maintain this diversity.
3. The social, economic, cultural and environmental factors influencing farmers to
maintain diversity on-farm.
4. The people maintaining this diversity in terms of gender, age, ethnic and social or
economic status in the community.
The information to study these topics comes from different levels and disciplines. The different sources and levels of information, i.e. the variety, the crop, the parcel or plot, the
household, the village or community, the landscape or region are summarized in the
diagram below. Information from one aspect may be useful to answer more than one question. What is important is that the information collected at the level of the household or farmer’s plot may not be the appropriate scale for analysis or for crop diversity conservation.