IIASA

Home

About

Foreword

Introduction

Research

Arguments

In-depth Analyses

All Data

     Tables

     Charts / Figures

     Thematic Maps

FAQ

Summary

Conclusions

Resources

Bibliography

Web Links

Index

Other

Feedback

Thanks

Help

Presentation

2nd Revision

Introduction

 
In-depth Analyses
The Agro-Ecological Zones (AEZ) methodology.
The crop cultivation potential is certainly one of the most important factors for China's food security. It describes the upper limit for the production of crops under given agro-climatic and soil conditions on a specific level of agricultural technology. Various methods have been used to calculate this upper limit (see for instance: Luyten / Qinghua / de Vries, 1996).
A most detailed and mature methodology is the so-called agro-ecological zones (AEZ) approach, which was originally developed by FAO and IIASA with support from UNFPA in the early 1980s (FAO / IIASA / UNFPA 1982) and was then repeatedly improved in several global and national studies (FAO / IIASA, 1993).
The most recent version of a global AEZ analysis is currently under development in a collaborative project by IIASA and FAO (see: Fischer / van Velthuizen  / Nachtergaele, 1999). The following discussion is based on preliminary results from the LUC AEZ study, which investigates the cultivation potential of China, Mongolia, and the Former Soviet Union on the basis of recently updated soil, terrain, and climate databases.
blank_3.gif (810 bytes) blank_3.gif (810 bytes) blank_3.gif (810 bytes) blank_3.gif (810 bytes)
The AEZ model algorithm
To understand the basic idea of the AEZ approach let us imagine a hypothetical farmer who has the task to evaluate the suitability of China's various land areas for crop production. He would use a whole range of criteria to assess each unit of land, such as the quality of the soil, the local climate conditions, and the possibilities of using different types of agricultural input (fertilizers, pesticides, machinery). The farmer would also consider various kinds of crops, because a particular area might be very suitable for one particular crop (for instance rice), while only moderately suitable for any other. The AEZ algorithm proceeds in the same way. However, the model systematically tests the growth requirements of 154 major crop types (including 83 types of grains) against a very detailed set of agro-climatic and soil conditions. For China the model operates on a 5 by 5 kilometer grid; so the total grid matrix has 810 by 970 cells, of which some 374,814 grid cells cover the Mainland of China. Water bodies are automatically excluded. In each of these land-related grid cells the AEZ model performs the following (principal) steps:
numb1.jpg (5531 bytes) The algorithm first evaluates the climate conditions. Obviously, crop cultivation is only possible, if temperature and precipitation are within a certain range. From an agronomic point of view the key concept is the "potential evapotranspiration". Plants need a constant supply of water for their metabolism, in which they loose moisture due to evapotranspiration - especially during the growth period. In rain-fed agriculture the moisture supply to plants depends on the precipitation and the water-holding capacity of the soil. Some soils (such as Andosols or Chernozems) can store water much better than others. A given amount of rainfall might be sufficient for a particular crop production on a soil with high water-holding capacity, while it might be insufficient, when the soil lets the water seep away or evaporate. To take into account these principal differences between soils the AEZ algorithm evaluates the climate conditions for 6 soil classes of water-holding capacity - from a water-holding capacity of 150 mm to15 mm, depending on soil characteristics and depth. For each soil class the algorithm calculates a soil moisture balance under local climate conditions. So far, we have only mentioned temperature and precipitation as climate parameters, but the AEZ algorithm actually uses a much more detailed set of climate indicators, which include (a) monthly precipitation, (b) minimum / maximum temperature, (c) relative humidity, (d) sunshine fraction, and (e) wind speed. For each grid cell the algorithm uses these climate parameters to calculate a crop-specific potential reference evapotranspiration. In that procedure the thermal and water requirements during the typical growing periods of all crops are matched against the actual temperature and precipitation profile of the current grid cell. Basically, in this first step the algorithm determines how much water the various crop plants would need under the climate conditions of a particular grid cell. These crop water requirements vary with croptype, soil class, and climate parameters.
numb2.jpg (5887 bytes) The algorithm uses the results from step one to estimate location-specific potential biomass and yields for each crop. For this calculation the program applies a crop model and the Length of Growing Period (LGP) concept. The algorithm simulates a series of growth cycles on a daily basis for each crop - with starting-days covering a complete 365-day period from January to December. In other words, the model tries to match crop-specific growth cycles into the Length of Growing Period of a particular grid cell, which is determined by its climate conditions. For each growth cycle the algorithm calculates a crop-specific photosynthesis response in a two-step procedure:
bluearr_r.gif (847 bytes) First, the thermal and radiation conditions of each grid cell (that is temperature profile, day length, cloudiness, and sunshine duration) are compared with the thermal and radiation requirements of the crops during their growth cycles. For this calculation the crops are grouped into four classes (so called adaptability groups), because in some crops photosynthesis is more sensitive to changes in thermal and radiation conditions than in others. Crops are also adapted to different temperature ranges.
bluearr_r.gif (847 bytes) Second, the actual moisture supply in the particular grid cell is compared with the water requirements of the crop. The AEZ model essentially calculates - on a day-by-day basis - a crop-specific soil moisture balance. If the water supply is less than the water demand of a particular crop, empirical yield-loss factors are applied, which reduce the potential biomass yield. On the other hand, the water balance also shows the crop-specific irrigation demand for each grid cell.
The purpose of this two-step procedure is to determine the starting-date for a growth cycle that produces maximum potential yield. Under some climate conditions it is obviously better for a particular crop to start cultivation early in the year, while under other conditions it might be better to wait for a rainy period. This procedure is repeated for both rain-fed and irrigated conditions and for three levels of agricultural inputs. To quantify potential yields, the program basically uses three characteristics: (a) the so-called Maximum Leaf Area Index (LAI), which is the ratio of leaf area as compared to the crop cultivation area. (b) The model also uses a so-called Harvest Index, which is the proportion of the primary produce (e.g. grain) to total biomass. (c) And the crop adaptability group defines the relationship between maximum rate of photosynthesis and the day-time temperature. The first two measures vary with the type of crop and level of input.
numb3.jpg (6042 bytes) So far, the algorithm has only dealt with hypothetical yields; in this step the model tries to determine the level of attainable production per grid cell. For that purpose the model applies three types of constraints: (a) agro-climatic constraints, (b) agro-edaphic (soil) constraints, and (c) terrain constraints:
  bluearr_r.gif (847 bytes) Agro-climatic constraints: In addition to the temperature profile and the other climatic factors, which the model already takes into account in step 1, two climate-related aspects affecting crop management are taken into account here: (a) workability constraints; and (b) wetness-related constraints. For instance, if a particular grid cell has a very high level of soil moisture (as determined by step one), harvest operations with machinery can become difficult, so that a high-input level cultivation may become impossible. On very dry and hard soils, on the other hand, ploughing is more difficult, which also limits the workability. The procedure also takes into account that humid conditions typically reduce yields through more frequent pests and crop diseases.
  bluearr_r.gif (847 bytes) Agro-edaphic constraints: These deal with soil-chemical and -physical constraints to crop production (in addition to the soil-moisture factors, which the model includes in the calculations of step 2). The AEZ model uses an agro-edaphic suitability classification, which was developed by FAO and other organizations and provides additional information on soil types, soil texture and soil phase. This soil rating scheme defines the suitability of each soil unit for each individual crop at defined levels of inputs and management circumstances. For instance, some soils may be very stony, others may have chemical problems, which will reduce the attainable crop production.
  bluearr_r.gif (847 bytes) Terrain constraints: These are limitations of crop production due to landform characteristics. For instance, soils on steep slopes are much harder to cultivate than soils in flood plains. They are also more susceptible to erosion and, consequently, fertility loss. For each grid cell the AEZ model takes into account a number of these constraints (which are specified by a terrain slope suitability classification) to further reduce the attainable grain production  if necessary. 
numb4.jpg (5658 bytes) In the calculation above the algorithm has only considered one crop per year. Obviously, this would be unrealistic in many grid cells, because often multiple crops can be grown in one season. To take into account the possibility for multicropping the AEZ model assigns each grid cell to one of 10 cropping zones for both irrigated and non-irrigated conditions - from single cropping zones (for cryophilic crops) to triple cropping zones (for thermophilic crops). Several climatic parameters are used to determine the cropping zone for a particular grid cell, such as the LGP, the days with minimum temperature above 5 degrees, the accumulated temperature during the growing period, and others.
Now everything is prepared for the actual selection of an optimal crop for each grid cell (or a sequence of up to three optimal crops in the case of multicropping): (a) The model has calculated the potential yields of all 83 grains under the specific climatic, soil and landform conditions of a particular grid cell; and (b) the algorithm has assigned each grid cell to a cropping zone. Now the algorithm has to select those grains among the 83, which maximize production in that particular grid cell. In the case of multicropping, the algorithm has to combine up to three grains to find the maximum potential yield.
numb5.jpg (1417 bytes) The selection of the "best" grain for a particular grid cell can be described as the task of matching the requirements of the grains with the characteristics of a particular grid cell in such a way that a maximum grain production can be achieved. For this, all crops are grouped into adaptability classes. Some crops have long growing periods (more than 120 days) others have short ones (less than 120 days). Some grains are typically sown before the winter (such as winter wheat), others are adapted to hot temperatures. The AEZ algorithm uses a scheme of 8 generic crop groups to specify the typical growth requirements of the crops. Each crop type belongs to one of these 8 groups.
In the case of a single cropping zone the selection of the best grain is easy. The algorithm compares the grid cell characteristics with the requirements of all 83 grains. Among those grains that fit, the algorithm selects the one that produces the highest potential production. In the case of multicropping the selection is more difficult. Of course, the algorithm cannot test all possible combinations of the 83 available crops - this would multiply the time needed for the calculations. Instead only those grains are tested as a second or third grain, that have the highest yield in each of the 8 adaptability classes. A number of rules are applied to guide this selection process (for details see: Fischer / van Velthuizen / Nachtergaele, 1999). They have been designed to make sure that the algorithm uses typical crop sequences in cultivation cycles. For instance, in the typical double-cropping areas around Shanghai, the algorithm would select a long-cycle rice or maize crop as the most productive summer crop, and winter wheat or barley (depending on which is more productive) as the winter crop - if the combination of both grains match the Length of Growing Period of the particular grid cell. In a triple-cropping zone either three short-cycle crops or two long-cycle crops are permitted.
numb6.jpg (1419 bytes) With the calculations above the AEZ algorithm can now compare the potential yields of the selected crop in a particular grid cell with the overall maximum potential yield of that crop in all other grid cells of China. For instance, the algorithm might find that in a particular grid cell the selected wheat has a potential yield of 9 tons per hectare. This potential yield is now compared with the overall maximum yield for wheat in China (which might be 10 tons per hectare). Now the algorithm "knows" that in this grid cell one can potentially produce 9 / 10 = 90% of the maximum yield, which would be equivalent to a grid cell that is "very suitable" for wheat production. In other words, the potential yields of the primary crop are used to classify each grid cell into one of the following suitability classes:
very suitable = yields are equivalent to 80% or more of the overall maximum yield,
suitable = yields between 60% and 80%,
moderately suitable = yields between 40% and 60%,
marginally suitable = yields between 20% and 40%,
not suitable = yields between 0% and 20%.
numb7.jpg (1256 bytes) As a result the AEZ model provides the suitability class for each individual grid cell and the potential maximal yield estimates by level of input. These are maximal crop yields, which could be expected, if the land would be cultivated under given climate and soil conditions. However, not all suitable land can be cultivated. Some arable land must be set aside for settlements and for the water and transportation infrastructure. China has a large network of open irrigation canals, which take up a considerable amount of land suitable for cultivation. Some of the potentially suitable land is also needed for mining and industrial production sites. Finally, some land with cultivation potential should not be used for agriculture, because it is still covered by valuable ecosystems, such as natural forests or wetlands. How much of that land can be reserved for nature depends on the increase in overall food demand and on agricultural productivity.
In a final step, the AEZ methodology takes into account non-agricultural land use within arable areas. First, the digital land-cover map of the Chinese Institute for Remote Sensing Applications (IRSA) is used to determine the size of the "usable" land area. The "usable" land area is the total land area, minus water bodies and (almost) unused land, such as rocky mountains, deserts, forests, and grassland. This actually usable land is used as the denominator for calculating the percentages of land areas that are used for settlements, infrastructure, and mining (which are taken from province-level statistics). Second, this province-specific percentage of infrastructure and settlements is multiplied by 1.33 to account for future urban and infrastructure expansion. Third, the potential arable land area from the AEZ model is reduced according to this province-specific correction factor.
Basically, this procedure reduces the number of grid cells that are suitable for crop cultivation by a province-specific correction factor for infrastructure.
As a result of this seven-step algorithm the AEZ model produces a database with some 375,000 records that specify for each grid cell the maximal crop-specific yields that can be expected under given agro-climatic conditions. For each grid cell the AEZ model gives the distribution of suitability classes, from "very suitable" to "not suitable".
blank_3.gif (810 bytes) blank_3.gif (810 bytes) blank_3.gif (810 bytes) blank_3.gif (810 bytes)
Discussion
The AEZ approach uses a detailed, spatially explicit methodology that is closely linked to established agronomic concepts, such as the Length of Growing Period concept. It is also based on widely available soil, terrain, and climate databases. The modeling results for China certainly represent the most detailed estimate of the potentially arable land on the basis of currently available data sets. However, to avoid misinterpretation of the modeling results, it is important to be aware of the following restrictions:
Other crops
The assessment of China's potentially arable land is based on grain suitability. While the AEZ model can handle 154 different crops - including non-grain crops such as oilcrops or cotton - only the 83 grains were used in the China assessment. This restriction was mainly applied to reduce the time requirements for modeling. The consequence of this restriction is a more "conservative" estimate for the potentially arable land. There are a few crops which can be cultivated in areas where grain production is impossible. For instance some tropical crops, such as bananas or oilpalms, can be cultivated under climate and soil conditions, which would not be suitable for grain production. Therefore the model might slightly underestimate the areas that can be cultivated by restricting the crop range to grains.
Fallow requirements
We have used the estimates of potentially arable land to calculate China's maximal attainable grain production. These estimates are based on the typical grain yields for low, medium, and high input levels. While there is no problem with the yield estimates for the high and medium input level, the estimates for the low input level are certainly too optimistic. They are possible for a few years, but cannot be achieved in the long run. Sustainable agriculture at low input levels (that is without chemical fertilizers) needs frequent fallow periods so that the natural soil fertility can recover. Fallow requirements vary depending on the specific soil and agro-climatic conditions; but we must assume that a sustainable yield over long periods is between 10% (for high-input conditions) and 30% (for medium-input conditions) lower than the typical yield during the cultivation years. Fortunately, this restriction does not strongly affect the modeling results, because most of the arable land in China is suitable for high or medium input agriculture, where fallow restrictions are only minimal.
Current land cover and land use
The AEZ algorithm assesses the potential cultivation suitability of a particular land area, depending on its soil, terrain and climate conditions - without consideration of its current land cover or actual land use. This was misunderstood by some critics of previous AEZ models as a weakness of the method, while, in fact, it is one of its strengths. The objective of the method is to assess whether a particular land area could be used for cultivation, not to explain why the land is currently used in a particular way. The method might find a particular land area to be suitable for crop production, while, in fact, people have used it for a settlement or for recreation. This also applies for the natural land cover. The AEZ method might find potential cropland areas, which are currently covered by natural forests or wetlands. The model does not suggest that these areas should be used for agriculture, it only indicates whether they could be used for cultivation.
To get a realistic estimate for the maximum cropland area in China the potential areas have to be reduced by the land requirements for habitation, water and transportation infrastructure, recreation, and non-agricultural production. A correction procedure was outlined above. Some areas should also be set aside to preserve valuable ecosystems. The AEZ approach could be used to find areas for these non-agricultural land requirements outside of highly suitable cropland areas. It can also show that only about 10% of the area suitable for crop cultivation is currently covered by non-agricultural ecosystems, such as grasslands, forests, and wetlands.
Some readers might feel that a 10% reserve for "natural" ecosystems is not enough. However, the percentage refers to the land that is suitable for crop production. Some 80% of China's total land area is not suitable for crop production; so there is abundant space for "natural" ecosystems in China. Most of these largely unmanaged ecosystems are located in the central, northern, and western parts of the country.
blank_3.gif (810 bytes) blank_3.gif (810 bytes) blank_3.gif (810 bytes) blank_3.gif (810 bytes)
Some technical details of the AEZ model
The AEZ algorithm is implemented in a series of FORTRAN programs for DOS-compatible personal computers. However, the complexity of the computations and the huge volume of the spatial databases that are required prevent easy portability. For instance, one modeling cycle for China includes the calculations for 31 years of climate data (1958-1988) plus one run for the average climate - multiplied by 6 levels of water retention capacity of the soils, plus one run for irrigated land. Each run tests 83 grains against the agro-climatic conditions of some 374,000 grid cells. The whole procedure would take some 2,300 hours (or about 96 days) on a 450 MHz Pentium II PC. At IIASA, 6 PCs were used in parallel to run the China model, which reduced the time for one complete modeling exercise to about 16 days. The run also required several dozen MB of input data and produced some 12 GB of output files.
blank_3.gif (810 bytes) blank_3.gif (810 bytes) blank_3.gif (810 bytes) blank_3.gif (810 bytes)
blank_3.gif (810 bytes)blank_3.gif (810 bytes)
Revision 2.0 (First revision published in 1999)  - Copyright 2011 by Gerhard K. Heilig. All rights reserved. (First revision: Copyright 1999 by IIASA.)