Crops and their environments are highly complex systems with a multitude of variables that change from location to location to different degrees and on various time scales. Due to this complexity, practices optimized for a research station might not be so successful when transferred to another location.
Though the new location may appear similar to the research station, there may be an undefined key limitation or combination of minor but different limitations that constrain production.
In many cases, carrying out a small-scale trial, actually at the new location, will lead to an optimal local farming practice more rapidly than trying additional subtreatments at the research station.
This chapter is concerned with the practicalities of on-farm research. It identifies when on-farm research is needed, the benefits and traps of the approach and the types of trials that are likely to be appropriate. It and associated “Explore On-Farm” chapters are intended to provide general concepts, basic designs of trials, methods and sample data plus ways of interpreting that data to improve farm management. An aim is to stimulate ideas. The trials will need to be adapted significantly to match the requirements of the location.
When practices recommended by the research station or central authorities are being followed, but yields remain low;
when a location has special attributes that do not fit the general pattern for the area, the location may be on a hill, on a steep-sided valley, on more rocky ground than the norm, perhaps with wetter or more saline soil;
when a farm or group of farms is less productive than neighbouring farms despite best efforts by farmers.
Trials may vary but they have the common feature that they test whether changing something in a system alters the variable of interest, and by how much. An example might be checking whether adding a fertilizer to a crop at a particular time increases grain yield.
Good trials have these elements:
a question and, hopefully, an answer;
a hypothesis, which is your expected outcome of the trial and why you collectively think that way;
treatments, that is, variations to the normal procedure;
a control, which is what the farmer normally does - this control must always be included to compare or check against any treatment;
a design, that is, where treatments are positioned in the field and in relation to each other- the design is very important and may be the difference between obtaining an answer or not;
a method, that is, the steps taken to get the answer to the question; and
measurements and recording of effects by counting, linear or volume assessment, or weighing - these give the trial objectivity and show how big the effects are.
Recording the weather during the study is also very important.
Weather changes every season and may affect how well the treatment works. For example, the effect of added fertilizer will be very different depending on the timing and amount of rainfall. Having weather information helps to put the trial into the context of other years.
A trial that is not carried out properly is often worthless. It may seem to give an answer though the answer may in fact be quite wrong. If forward planning or changed farming practice is based on that wrong answer, the farmer could lose income on future crops. No trial is preferable to a bad trial.
Some benefits of trials:
the trial may show how to increase economic returns by a changed practice that increases yield. The change could be quite minor and cost little;
the trial may show how to reduce costs by reducing the use of fertilizer, machinery, labour or water without loss in yield;
in the long-term there may be indirect benefits of any changed technology that improves sustainability. An example might be a lime application which raises pH in the topsoil this season, thereby releasing previously bound nutrients, improving nitrogen fixation, flocculating clays and increasing earthworm populations over time leading to better incorporation of surface organic matter, improved aeration, better drainage, and finally, higher yield;
unexpected positive things may be discovered about the farm because the farmer has been thinking about it differently and observing it more closely and critically;
there is close interaction between the researcher and the farming community as they work together. The farmer learns research methods and concepts from the researcher and the researcher learns more about the local cropping systems and about limitations other than the technological ones;
neighbours will be interested in the trial, will doubtless give their opinions, may even join in and will be the first to adopt then adapt the changed technology if it works;
communal trials build up a feeling of community and benefits spread beyond the prime aims.
Some costs of trials:
the trial land will be out of normal use during the trial, so normal production may not be achieved, including the trial as part of the farm’s normal cropping pattern and keeping the trial small and manageable can minimize this cost;
the yield from the trial could be appreciably lower than normal yield, even perhaps a total loss, can the farm afford that risk?
extra inputs will be a cost, items like fertilizers, labour, machinery and perhaps most importantly, time;
there may be a personal and social cost of possible failure on the farmer, the family and friends. Expectations of the study should be moderate, not exaggerated. If anything, it is better to undersell expectations.
Small and simple is best
Do not plan too big an experiment, discuss demands on the farms’ resources and time as well as your own, aim to complete the study in full. Be realistic;
have a well-defined central question that can be clearly understood by collaborators and answered by your approach;
if a factor other than the main one appears to be strongly influencing the answer, then formulate with the farmer a subsidiary question by adding subtreatments, however;
do not add so many subtreatments that the farmer is lost in the complexity;
do not use fancy statistics to confuse yourself and collaborating farmers. Do not be misled by averages. Examine the detail in the data to assess the trends and the significance of any ‘outliers’.
Do not be fooled about outcomes
The farm is about to commit resources to a trial hopefully leading to a long-term benefit. There are some traps the farmer must also be wary of:
do not subconsciously design the study and select the information to give the expected answer, the expectations may be wrong;
when collecting and analysing information, have an open mind. One should look for little things that are unexpected, they may be the most important; the obvious things have probably been seen and addressed already;
the trial must be assessed in a detached way. Everything should be written down so that current conclusions can be reassessed later and new ones reached.
Use variation within and between fields
Many fields have good and less good parts and there are often gradients between the two. Explain to the farmer that if bits of the different parts can be included as subtrials without jeopardizing the main trial carried out on the dominant land type, the trial will probably provide much more information. However, exclude the areas that are very unrepresentative of the whole field, or unimportant.
Choosing which varieties to plant
In all trials include the variety most used on the farm or locally. If the trial is about comparing varieties, discuss with the farmer which varieties to include and why. Talk about their pest and disease resistance, lodging tolerance including propensity for tillering, fertilizer and water requirements, grain quality and yield, and likely crop duration from different planting dates.
How many plots, how big, where and when?
If practicable, use a strip design to answer the question posed by the trial.
This is the simplest possible form of design that farmers will readily understand. It is a rectangular strip a few metres wide, marked out within a normal farm crop. The treatments are applied in bands along the strip like the stripes across a scarf, and replicated as blocks at intervals.
This design is appropriate for simple trials like checking for fertilizer response of a crop whereas a research layout is better if there are many variables and subtreatments.
Strip plot positions (green) marked out in a farmer’s crop (yellow) and in a fallow field (grey). Controls (C) and treatments (T1-T4) are repeated in two blocks. Harvest areas are the blue-edged squares around the treatment letters.
Plot size - the edge effect
Plot size will depend on the question you are asking but larger is usually better. Explain to farmers why observations and measurements should only be on plants towards the central part of treatment plots, avoiding the outer rows that are subject to ‘edge effects’. Plants at the edges of experimental plots have more water, light, and nutrients available to them than the plants within the plots so they often yield better than the norm. These ‘edge effects’ are exacerbated if a clean cultivated border surrounds the plot, so consider surrounding the trial with a border of additional crop if it is not already located within a crop.
Examples of the two approaches are shown in the above diagram. Plot size, therefore, depends on balancing the need to eliminate edge effects against using bigger plots that may make the trial more costly and harder to maintain.
The control plot and how many plots are needed for a trial
There will always be a control plot (C in the diagram), which is the plot planted with the usual variety and with normal farm practice applied.
Additionally, there will be at least one treatment plot, the main variation to normal practice. Trials usually have several treatments that represent degrees of variation to normal practice.
A fertilizer study might have four treatments (T1-T4) of 0, 40, 80, 120 kg ha-1 plus a control of 60 kg ha-1 that represents normal practice. This would give a study of five plots. These would be T1, T2, C, T3 and T4 in the illustration.
Clarify for farmers why it is wise to have three replicates of all treatments, or more if resources permit. Replicates are copies of the same thing and when compared show how much the same thing varies. So in the fertilizer trial, the basic block would be repeated twice giving three blocks or replicates of five plots each, a total of 15 plots. Only two replicates, Blocks 1 and 2, are shown in the illustration for simplicity.
Depending on the complexity of the question the smallest study would be six plots (a control, one treatment and three replicates of both) and the largest would depend on available resources.
Laying out the plots
In a strip design of the above-mentioned fertilizer trial the treatments could be applied end to end along the strip and after a gap of a few metres, repeated in Block 2 and then again in Block 3.
Plots can be of any width within the strip as long as they are wide enough to permit the cutting of at least one harvest in their centre plus a wide enough border of plants around the harvest areas to avoid fertilizer interference from adjacent plots (wash on, root exploration, etc.). The harvest areas are shown as blue-edged squares in the illustration. Check the descriptions of specific trials for a guide to actual plot size and design.
If the question is more complex involving two or more interacting factors, like ‘What is the best variety by sowing date for the farm?’, requiring three varieties and three sowing dates, a conventional research design is generally required.
Location of plots
Locate the trial in a uniform average area that typifies the farm or field. Avoid areas that are likely to be useless for cropping or are in some way abnormal. If the field has small outstandingly good or substandard parts, also exclude these extreme types from the main trial. You could consider adding them as subtrials. Explain to the farmer that responses on the extreme parts will not apply to the farm in general, but could provide useful secondary information about the farm.
H.M. RAWSON
The trial plots should preferably be within the normal crop. Mark the locations and dimensions of each plot clearly right from the outset and label the treatments. White sticks work well.
When should plots be prepared?
The trial plots should be cultivated at the same time and in the same way as the normal crop and as part of the normal crop preparation.
There are two exceptions. If the trial is about cultivation methods obviously there will be differences. If the sowing date is a variable, plots for each sowing date must be cultivated and otherwise prepared separately, each at a similar period in advance of its sowing date. This complicates the study because of the need to get machinery to small plots past other plots and through the normal crop.
Observing the crop as it grows
Seedling emergence
Variable results in trials start with variable plant stands due to uneven cultivation and problems during sowing.
These problems are unavoidable in some cases, but if they are noted or measured, their influence on the final data can be allowed for.
Ten days to three weeks after planting, when all seedlings are emerged, lay a metre-long measuring stick next to the row and count the number of seedlings along its length. Do this for one sample in every plot in every block. Note down all the numbers with their plot identifier and determine how homogeneous is the density among plots. Is one treatment or one block poorer than others? If the seed has been broadcast, put down four sticks of one metre to form a metre square or quadrat and count seedlings inside the square. A larger or smaller area can be enclosed but its size should be known and the same area enclosed for each plot. If preferred, the metre square approach can also be used in crops that are precision sown in rows by machine.
Flowering
Flowering is a most critical time for the crop, both in terms of the crop’s sensitivity to the environment and because flowering marks the beginning of the grain filling phase.
For each plot make a record of the date of flowering and of the weather at that time. A useful, though less important and less precise date, is that of crop maturity but, if possible, note it down too.
Harvesting
The common question in trials is, “How will the treatment affect final yield?” So final yield must be measured accurately and in an unbiased fashion.
Discuss some basic rules with the farmers:
do not leave plants in the field after they are ripe while waiting for the next treatment to be ready. Without doubt an unexpected event will occur and the ripe treatment will be lost to birds, mice, rain, hail or straying cattle;
use the same method for harvesting all plots. Do not harvest some by machine and others by hand unless there is an emergency and machinery is unavailable at harvest time;
if a small combined harvester is available, first trim away plot edge rows or borders and then harvest the remainder, plot by plot, collecting the grain (and trash) of each plot separately. The area harvested should be identical for all plots and must be a known area. If it cannot be identical because of driver errors or other factors, then measure and write down the areas actually harvested;
if harvesting by hand, ignore edge rows. Do not select individual plants in plots. Use the measuring stick or metre square quadrat to indicate a metre of crop row or crop area to harvest from the central part in each plot. Avoid parts of centre rows that you know established poorly from earlier observations and notes. In general, do not include in the sample sections that have been positively or negatively affected by something other than the treatment;
cut off the plants at ground level but do not include any soil;
tie up and clearly mark the plot sample with its plot identification, block, treatment, and harvest date. Bag if possible so that parts are not lost;
keep the samples together in a vermin proof area;
a main source of error in assessing biological and grain yield is that the amount of moisture in samples may differ between treatments. Dry the samples together in the hot sun, a hot greenhouse, a plastic house or in an oven.
Measuring the results
· Weigh the-dried samples. Weigh in the bags if the bags are uniform weight and then subtract the bag weight from the total. Write the numbers down;
· if samples are hand harvested, thresh and weigh the grain. If they are machine harvested, just weigh. Write down the weight;
· if hand harvesting, calculate Harvest Index (HI) by dividing the weight of the threshed grain by the weight of the whole sample. The result will commonly be between 0.25 and 0.55. If machine harvesting, divide the weight of the threshed grain by the sum of the trash and grain weights;
· refer to the companion Explore-On-Farm chapter Constraints to cereal-based rainfed cropping in Mediterranean environments and methods to measure and minimize their effects for details of how to harvest the crop, calculate yield components and estimate water use and water use efficiency.
H.M. RAWSON
Recording the weather on the farm
If you are collaborating with literate farmers, and there is no weather station close by, think seriously about the possibility of recording temperature and rainfall on the farm.
Weekly maximum and minimum temperature over a year at a cold location showing frost-free months
Temperature should be recorded regularly to pinpoint frosts, heat waves and calculate mean temperatures. All are needed to interpret why a crop yielded well or poorly and matured early or late, and to plan changes in management for future years.
Preferably the farmer should use a maximum and minimum thermometer hung in a slat-sided box painted white or in the shade.
Explain how it is used, especially the need to reset it after reading. It is best to read and reset it at least once a week, preferably on the same day each week, and all the year round. The data in the figure were collected using such a thermometer.
The amount of rainfall and its seasonal distribution will be an indicator of both potential yield for the farm and the variability of potential between years. It will also enable the farmer to work out water use efficiency.
If available, use a standard rain gauge to collect rain. Otherwise use a large straight-sided can and measure by dipping a ruler into the water. If rainfall is very low, the water should be tipped into another much smaller can, calibrated to the large can, for measurement.
H.M. RAWSON
Once a week measure then discard accumulated rainfall, preferably on the same day each week. Locate the measuring devices in the trial field away from buildings or overhanging trees.
Using long-term weather information
Timing of frost
If long-term weather data are available either from the farm or from a local meteorological station, the farmer should note down the beginning and end of the frost-free period for the area and check whether it changes from year to year.
Frost can have a devastating effect on yield particularly if it occurs around flowering. On the example figure from a high latitude site, the frost-free period marked as a yellow zone is less than four months.
Amount of water
Ask collaborating farmers to assess from long-term weather data, if available, how much water is available in the average season, and how much rainfall varies among years. Explain that a crop needs a certain amount of water during its growth to reach its potential. If that water is not available when needed, yield will be lower than potential. A shorter duration variety might be more appropriate for the area as that will require less water but it may also have lower potential yield.
Writing it all down
Keeping accurate and frequent records is vital
Records must be collected and set out in a structured way following patterns determined largely before the trial begins. Structured data and worksheets lead to clear and structured thinking. Haphazard trials with haphazard observations and recording are generally worthless. Discuss and design the worksheets with collaborating farmers before the trial starts. Provide them with hard copies of the final design in a workbook produced for the specific trial. Encourage them to use the workbook for all data and notes.
Discuss some rules for recording data:
write things down as they occur. Memories play tricks;
do not write important information on scraps of paper. Someone will throw them away. Back of the hand is OK as a temporary measure;
dedicate one workbook or worksheet to one trial. Using a diary that mixes experimental data with appointments and shopping lists is messy and can be confusing;
enter information collected on different dates in sequential date order. Avoid mixing up dates in the workbook;
always label data with its plot identifier and date. What seems obvious today will be confusing next month;
organize data in tables (with rows and columns) so comparisons between replicates or treatments can be made by eye;
avoid long, single columns of data such as those used in adding machines that print on paper rolls;
where possible, keep a standard design for tables. See the chapter on Optimizing variety x sowing date for the farm for examples;
when possible, enter the data as it is collected straight into tables so that they can be easily summarized as totals or means at the bottoms of columns and the ends of rows. Rewriting data copied from scraps of paper, sample bags or distributed notes in a workbook can be very time consuming and prone to mistakes.
Start your discussions about the farm(s) itself by trying to define with farmers the extent of the available growing season.
This will be an introduction to thinking about the interplay of the environment and any varieties used. Explain to him/her that knowing the growing season is important for using any cropping area to its full, sustainable potential, particularly in a rainfed environment.
The critical factors are: how long is the local environment warm enough, moist enough and sunny enough to start and support the growth of the crop; and to what extent can different varieties and management approaches alter the length of the season.
If available, long-term weather records from the local meteorological agency could well be useful for calculating the average starting and finishing dates for the area and to illustrate the discussion.
Limits to growing season length
Discuss with collaborating farmers what you all see as the main limitations to the growing season. Its length is likely to be limited by a cold season such as winter and by the start and end of the rainfall period.
It can also be affected by one-off catastrophic weather events, like frosts, hailstorms, rainstorms, gales, and hot dry winds. There are other catastrophic risks like locust plagues and some diseases that are more likely to occur at specific times in the year and may limit the crop season. If such events occur every year they should be avoided by having the crops at stages that are not damaged during these events.
In some perennially warm, damp areas, both starting and finishing dates may be largely under the farmers’ control, but in all areas farmers can have a degree of control over season length either through farm management or choice of crop.
Debate with farmers the following questions relating to the growing season
Could local crops yield better if sown earlier or later? Why?
Are crops maturing fully or are they drying off before their grain is completely filled?
Are local temperatures too high or too low at any stage for any of the crops? When?
Is there sufficient water available at the right times for the crops? When is it limiting?
Is it possible to modify soil structure to enable crops to use the full potential growing season?
Is it possible to grow one crop species after another, perhaps using shorter duration varieties, to make more productive use of the potential growing season?
These questions are intended to lead to the conclusion that two main factors determining the potential growing season are the environment and how well the plant material chosen matches the environment.
A third all-important factor combining the first two together is the farmer’s management of the crop. Activities must be accurately timed around the changes occurring in the environment and in the crop.
Some issues associated with timing follow:
Is the land ready for planting at the start of the season?
Is the land too dry or too wet to allow the crop to be planted at the optimal time? Can a change in tillage or planting methods overcome the problem?
Are previous crop residues causing delays? Can they be better managed in advance, removed or incorporated or used for mulch?
Was the previous crop harvested in time to allow planting of the current crop at the optimum date or should a shorter duration variety have been used?
Was the previous crop harvested early to take advantage of the best market price? Would the current crop yield be more if planted early?
Are activities during crop growth timely?
Are fertilizers and irrigation, where available, being applied at the most beneficial time for the crop and are they used efficiently?
Are pests, diseases and weeds being controlled at the right time and effectively? Could the outcome be achieved by using less effort or fewer chemicals in smaller amounts?
Is machinery available when needed?
Finding out which, and to what extent, these environmental, crop variety and management factors have an impact on the best use of the growing season is the first reason for doing on-farm trials.
The second reason for trials is to test ideas to overcome the constraints to production that have been identified.
Debating these questions should have improved your knowledge and the farmers’ awareness of local limitations.
Hopefully in your discussions you will have identified the key constraints to productivity on the farm and have debated ways that the constraints might be overcome. These possible solutions could become the treatments in a trial.
Design the trial with collaborating farmers including these solutions as treatments. It may take several meetings before the trial is finalized to the satisfaction of all parties but it can be an exciting and profitable time.
Throughout, encourage the free-flow of ideas. Do not dominate or be elitist in these interactions.
When you talk with collaborating farmers about the trial results do it in such a way that the numbers discussed are paralleled with a mental picture of the crop. Stick initially to raw data and simple averages and trends.
Get the farmers to think about why a particular plot has done better or worse. Debate its state of weediness, whether it was wetter, whether a carcass of an animal was left there a few years ago and so on. Use the farmer’s knowledge and notes in the workbook to involve him/her in the numbers and guide him/her to do the interpretation. The farmer’s confidence and enthusiasm will grow rapidly. This will build up a much more reliable picture of what the trial means than a statistical analysis. If appropriate, use that analysis yourself to substantiate the farmer’s interpretation and if you decide to publish the results of the study.
Follow-up these discussions about interpreting the data taking into consideration what these interpretations mean to future practices on the farm. Always link the data back to practicalities. As the discussions proceed, get the literate farmers to note down the conclusions, but make your own notes. Even rough notes are very useful for jogging the memory and can save a lot of thinking time later when the memories of the discussions have dimmed.
Send a copy of your conclusions to the farmer as soon as you have tidied them up. Do this soon while details of the study are still fresh.
Consider following up the study with an on-farm field day to involve other local farmers in the results. Encourage them to air their views on the weaknesses and strengths of the project and to suggest ideas for future collaborative studies.
Some researchers are afraid that results from on-farm trials may not be acceptable to scientific journals because of statistics considerations. However the chapters provide examples of simple but statistically satisfactory trials.
Ashby, J.A., Braun, A.R., Gracia, T., Guerrero, M.P., Hernández, L.A., Quirós, C.A. & Roa, J.I. 2000. Investing in Farmers as Researchers: Experience with local agricultural research committees in Latin America. CIAT, Cali, Colombia
FAO. 1999. Las escuelas de Campo para Agricultores (ECA): Un proceso de extension grupal basado en métodos de educación no formal para adultos, by K. Gallagher. FAO Global IPM Facility, Rome.
FAO. 2000. Guidelines and reference material on integrated soil and nutrient management and conservation for farmer field schools. 164 pp. Information Resources, FAO Land and Water Development Division (also available at www.fao.org).
FAO. 2000. Guidelines for on-farm plant nutrition and soil management trials and demostrations. 71 pp. FAO Land and Water Development Division (also available at www.fao.org).
FAO. 2001. Guidelines for the Qualitative Assessment of Land Resources and Degradation. FAO Land and Water Development Division (also available at www.fao.org).
FARM Programme. 1998. Farmer field school on integrated soil management, Facilitator’s Manual. FAO Land and Water Development Division (also available at www.fao.org).
