Revision as of 11:32, 24 September 2006; view current revision
←Older revision | Newer revision→
1.1. General introduction to crop forecasting and its methods.
Jan Jansonius, René Gommes, Peter Hoefsloot Crop Forecasting Methods
Various disciplines make a contribution to crop forecasting and all over the world different methods are used. Several methods of crop forecasting are briefly reviewed here, but a detailed treatment is beyond the scope of this manual. In practice, the choice of a specific methodology or mix of methodologies depends also on cost and available budget.
Forecasting and monitoring
We should mention the links between forecasting and monitoring. Traditionally, monitoring is implemented by direct observation of the stage and condition of the organisms being monitored (type 1), or by observing the environmental conditions that are conducive to the development of organisms (type 2). The second type allies mostly to pests and diseases. Surprisingly, type 1 monitoring, which involves the direct observation of organisms, is often more expensive than type 2 because of elevated labour costs.
Micro and Macro level Crop Forecasting
It is important to note that operational crop forecasting is done for different spatial scales. At the lowest end, the “micro-scale”, we have the field or the farm. Data are usually available with good accuracy at that scale, for instance the breed or the variety are known, and so are the yield and the environmental conditions: soil type, soil depth, rate of application of inputs. The micro-scale is the scale of on-farm decision making by individuals, irrigation plant managers, etc. The macro-scale is the scale of the region, which is why forecasting for a district, or a province is usually referred to as “regional” forecasting. Regional forecasts are at the scale of agricultural statistics. Regional forecasts are relevant for a completely different category of users, including national food security managers, market planners and traders, etc. At the macro-scale, many variables are not known and others are meaningless, such as local soil water holding capacity. Needless to say, the real world covers the spectrum from macro- to micro-scales, but the two extremes are very well defined in terms of customers and methods. Several applications are at an intermediate scale. They would include, for instance, certain types of crop insurances, the “livelihood analysis” that is now applied in many food security monitoring systems, fire monitoring systems, etc.
Forecasting techniques in general
There are a variety of generic forecasting methods, of which most can somehow be applied to agrometeorological forecasting as well. According to Armstrong, “judgement pervades all aspects of forecasting”, which is close to the following definition of crop yield forecasting: “the art of identifying the factors that determine the spatial and inter-annual variability of crop yields”. In fact, given the same set of input data, different experts frequently come up with rather different forecasts of which, however, some are demonstrably better than others, hence the use of the word “art”. There appears to be no standard classification of forecasting methods. Roughly speaking, forecasting methods can be subdivided into various categories according to the relative share of judgement, statistics, models and data used in the process. Armstrong identifies 3 groups of methods:
- Judgemental, based on stakeholders’ intentions or on the forecasters' or other experts’ opinions or intentions. Some applications of this approach exist in agrometeorological forecasting, especially when other variables such as economic variables play a part (for instance the “Delphi expert forecasting method” for coffee, Moricochi et al., 1995);
- Statistical, including univariate (or extrapolation), multivariate (statistical “models”) and theory-based methods. This is the category where most agrometeorological forecasting belongs;
- Intermediate types include expert systems, basically a variant of extrapolation with some admixture of expert opinion, and analogies, which Armstrong places between expert opinions and extrapolation models.
Examples of judgmental forecasting systems
Quantitative crop reporting by Extension Workers
Many countries still apply a system whereby each Agric. Extension Worker actually produces area and production figures for all crops in his area, simply based on his judgement. Such data are added up to the next administrative level and finally aggregated into provincial and national crop estimates. Quantitative crop reporting by Extension Workers is not recommended, as the figures will normally be unreliable.
On the other hand, experienced agricultural field staff are normally well qualified to make useful reports on crop conditions in qualitative terms and to relate the current situation to last years crop or to “normal” conditions. They can also highlight the specific critical factors affecting the crop.
Damage assessment
Natural calamities such as devastating crop disease, pest attack (locust, army worm), severe drought, floods, typhoons may strike standing crops and cause partial or total loss of crop production. It will be necessary to assess the damage rapidly. The crop area in ha. that suffered total crop failure or partial crop losses and the number of villages and population affected will need to be determined. The affected area can in most cases also be identified from satellite imagery or aerial photography and it may also be possible to assess the degree of damage using remote sensing methods.
A sample of farmers affected by the calamity may be interviewed or the damage to their fields assessed by an eye estimate of the investigators. The size of the sample will probably be small, it may be by reasoned choice, not necessarily a random sample.
Crop assessment mission
In many countries it is an established practice to organize once or twice-yearly a crop assessment missions by a team of experts in various fields. The purpose is to produce an up-to-date picture of crop conditions and all related information. Rough and ready methods are used. The team travels through extensive cropping areas for direct observation of crop conditions. Every opportunity will be taken to interview farmers. Meetings will be held with staff of MOA, Local Government, other Departments or Services, Agricultural Research Institutions, NGO’s, Traders and Processors representatives.
This approach has often proven itself to be very effective in mobilizing locally available information and identifying important aspects and new developments that otherwise would have gone undiscovered. The reliability of such assessments will improve, if the same persons participate in the mission several years in a row, allowing them to compare their current observations with the situation seen during previous years.
Statistical and model-based forecasting methods
Rainfall monitoring
Rainfall is one of the most critical factors to be monitored, especially in conditions of rainfed agriculture, but indirectly also for irrigated crop production (water availability). Many crops are highly sensitive to water stress, for example paddy during the whole growing cycle and maize during flowering stage.
Monitoring rainfall distribution in rainfed crop production areas will provide “early warning” indications rather than a quantitative forecast of crop production. Nevertheless it plays an important role in forecasting in conjunction with other methods.
In most countries it is the Meteorological Service that operates a network of stations at national level, the major stations often being located at airports and in the bigger towns. MOA often operates its own network of cheap raingauges with a better representation of important cropping areas. It has to be seen if in each of the Crop Zones, which are the basis of the crop reporting system, rainfall records are available from at least two and preferably more stations. If not, additional raingauges should be installed.
It is recommended that direct decentralised use is allowed to be made of rainfall data recorded in the Provinces and Districts. This is in addition to the regular system of meteorological data collection and publication by the National Meteorological Service, which may result in delays in data availability. In many countries the Met. Service claims a monopoly on all meteorological and climatic data and its interpretation and will only give access to its database against payment.
Officers in charge of crop forecasting may simply record weekly or 10-daily rainfall records for each available station and compare with normal values, possibly also with estimated crop water requirements. This information will be used to detect dry spells and potential damage from excessive rainfall.
Weather models
More sophisticated agro-meteorological monitoring is done with the help of models. India and various other countries have developed models based on weather data. It involves identification of the most significant weather parameters during different stages of crop growth, which are then used in a regression model with or without trend.
These models require a long series of weather data and reliable crop yield data for calibration of the model. The data available for this purpose are often insufficient. A practical problem is also scarcity of current data. While in most countries there is a large number of rainfall stations, the number of climatic stations that record also other variables such as temperature, sunshine hours and relative humidity is much lower. The data of one station is then used for large areas, while there may be in fact considerable variability in weather conditions.
In China a 30-year series of meteorological data has been compiled in six regions. The meteorological variables most relevant for each crop stage were studied by region, by crop and by month, taking into account the growing cycle of each crop. On this basis, the weather conditions in the current year and those in the previous years are weighed and clustered so as to identify the year in which the weather conditions were most similar to the current year.
In Japan it was found that the yield of paddy is highly correlated with weather conditions. A weather index representing temperature, rainfall, sunshine hours and wind speed has been used successfully to forecast paddy yield.
Most models do not respond well to extreme conditions in weather. They work best when the weather conditions are within the range of the data that were used for fitting the models.
The FAO Water Balance Model
This model forms the core of this manual and has been used for many years, especially by countries relying on rain-fed agriculture where inadequate availability of water to the crop is the main constraint. The model calculates a running water requirement satisfaction index (WRSI) during the growing season of the crop. The final value of this index is correlated with crop yield level. The correlation will be high, if all other factors affecting crop yield are fairly constant from year to year. Historical data on rainfall and crop yield are required, to establish a regression relationship between Index and yield. Unfortunately such data is in practice often not available.
Constraints of the model:
- A fairly dense network of raingauges and reliable records is required, because of high spatial variability in rain distribution in many countries. However, satellite based data are improving every year and could partly substitute ground data.
- The model is not reliable in case of extreme weather conditions
- The effects of excessive rainfall can not easily be accommodated
- Water holding capacity of the soil is a key variable in the model, for which there is often insufficient information and the variability may be high from place to place
- The input in the model should in principle be effective rainfall rather than the recorded rainfall. Therefore runoff as a proportion of total rainfall should be taken into consideration. Runoff depends on average slope of farmland, soil conditions and farming practices and again there is little data available.
To summarize, the Water Balance method provides a very useful “early warning” indicator for yield reduction due to water stress. However, in most circumstances, it can not be expected to produce a reliable quantitative yield forecast. The method will be used together with other indicators.