Land quality and other indicators of sustainable development statistical data, quality control and problems of aggregation
AN OVERVIEW OF THE WORK IN THE STATISTICS DIVISION
Recently, work has started to develop land quality indicators (LQIs) to monitor changes which make an impact on sustainability of land resources in meeting human needs. This work takes into account current status of land resources along with a measure of extent of its utilization for meeting today’s need and impact on future requirements. These indicators help in monitoring the major land issues and answer policy-related questions. In the work on LQIs land issues have been grouped into three clusters viz. inappropriate land-use systems, land degradation and inadequacies in the policy environment for land users. Looking from a limited angle of agriculture (crop and animal husbandry) some of the indicators on LQIs deal with items such as trends in crop yields, crop nutrient uptake/fertilizer use, farm gate/market prices for inputs, visible soil erosion (area, degree and proportion of land affected), major land-use, proportion of farmed area with recognized title, ratio land/people and livestock/people, intensity index (permanent cropped area/total cultivable area), diversity index (number of species x area of land-use type)/total area. This work on indicator definition also serves the broader context of sustainable development.
The Statistics Division (ESS) is responsible for collecting, analysing and disseminating statistics on food and agriculture, including data on agricultural inputs which are relevant to the environmental area. ESS has so far not made any effort to compile or collect LQIs but this paper briefly lists the current status of statistical data (called LQIs data hereafter) for compiling them, indicating its reliability and efforts being made to improve its coverage in future.
The agricultural database maintained by ESS contains time-series data (starting 1961) for over 210 countries and 1500 items on the production and trade of crops and live products, agricultural machinery, fertilizer, pesticides and insecticides (trade only), land use and population. These data series contain national aggregates/averages and are maintained in the form of supply/utilization accounts (SUAs) for most commodities. A SUA consists of elements that are basic or essential such as production, imports, exports and domestic use, and those that are supporting or supplementary such as animals slaughtered, harvested area or seeding rate. The basic equation of a SUA equates supply with utilization (demand) for a given period.
It has long been recognized that it is no longer meaningful to deal separately with individual statistical series, such as those for production and trade, etc. The statistical framework of SUAs has been developed with the aim of providing reliable data series for various uses. To derive these data series ESS performs three main activities, namely, collection of country data through questionnaires sent to member countries and from publications and reports, selection of the data collected after careful analysis and scrutiny and filling-in gaps when necessary. These data series are published in yearbooks and otherad hoc publications and also disseminated through FAO’s computerized databank World Agricultural Information Centre (WAICENT).
WAICENT is a corporate database composed of two linked data systems: FAOSTAT, which contains statistical information, and FAOINFO, which contains textual information. FAOSTAT contain a constantly updated collection of time-series data on demography, agriculture, fisheries and forestry from 210 countries and territories, as well as data on trade flows, food aid, development assistance and the results of the World Agricultural Census and household budget and food consumption surveys. The Centre has been created to provide clients – including governments, research institutions, universities and private users – with fast, economical access to FAO’s vast library of information on agriculture, fisheries, forestry, nutrition and rural development. As it becomes fully operational, WAICENT will allow users to retrieve information – initially in English, French and Spanish – from a wide variety of media, including floppy disks, CD-ROM and mainframe tapes as well as via “on-line” access through computer networks and telephone lines. FAO is promoting wide access to WAICENT through its new “Computerized Information Series” which collects WAICENT products on floppy disks or CD-ROM. The series includes a PC dissemination module of WAICENT which contains statistics collected since 1961 on population, land use, production, trade, food balances, forest products and food aid. Fisheries data will soon be added. The series also includes a digitized Soil Map of the World in seven volumes and 63 diskettes, the results of decades of study on the soil situation of each country.
LAND DATA
The importance of land data lies in its use as an inventory variable and an object for monitoring. A complete land-use database should also serve as an alert for emerging land-use resource problems. It should include detailed information on land uses/cover along with losses and gains between sectors, e.g., between forestry and agriculture, between agriculture and manufacturing, etc. This aspect of intersectoral relationships and connections has not been handled adequately in policy studies and efforts are needed to address them. The value of land-use data is influenced by its timeliness, coverage, accuracy and structure. The data must fit policy needs and be at appropriate disaggregated levels to make them useful. This recognizes the need for suitably geo-referenced data at the national and sub-national levels for developmental work.
National data on land use available in the ESS data bank include the following categories covering:
¤ total area (i.e., area including area under inland water bodies);
¤ land area (i.e., area excluding area under inland water bodies);
¤ arable land;
¤ land under permanent crops;
¤ permanent meadows and pastures;
¤ forest and woodland;
¤ other land (includes built-on areas, roads, barren land, etc.).
These data are published regularly for continents and countries in the FAO Production Yearbook for selected years. Similar data at sub-national level are available for a few countries for selected years and are generally related to data collected in agricultural censuses. ESS also has data on irrigated land, including land irrigated by controlled flooding. FAO compiles these data from questionnaires forwarded to countries and national statistical publications and an array of other sources such as project reports, including studies available from other FAO Divisions and economic journals. This database is not always up-to-date for several reasons. In many countries there is no established statistical system to generate such data. In some cases the primary sources go back more than twenty years. It is also clear that no single source can provide all of the data required to study land-use patterns which introduces the additional problem of consistency.
The quality of international data depends first on the quality of data collected by national institutions and secondly on success in compiling data collected from different sources into one framework for international use. The only source of data at international level is the data collected by the countries themselves. The Agricultural Census or an agricultural survey is often the basic source for such data collected by the countries. These data relate to arable and permanent crop land while other classes such as marginal land and eroded land which are of great environmental importance are obviously ignored (or greatly underestimated). Obviously, data on forest and water bodies are often missed. Even in those countries where a system for collection of agricultural statistics exist, there is still a need to improve the system and the reliability of the data.
Unlike land cover data, almost all land-use data are collected on a sample basis. These sample surveys are designed to provide estimates of known accuracy and reliability for the area sampled. It is desirable to have a full cadastral survey but very few countries are in a position to find resources to prepare one. In such situations, one is left with the Agricultural Census (which is a periodical operation in many countries) and surveys. These censuses/surveys cover agricultural holdings rather than the complete land area.
PROBLEMS IN COLLECTION OF DATA
Assembling and tabulating this enormous mass of data in internationally comparable form presents many problems arising from differences found in countries’ data as regards to concepts, definitions, coverage and classifications. From the early sixties to the present day, particular attention has been given to these problems at various international and regional meetings, seminars and training workshops/national demonstration centres such as those promoted by ESS in collaboration with UN Economic Commissions, the Inter American Statistical Institute, the Conference of European Statisticians, the FAO Statistics Advisory Committee of Experts 1. Some of the common problem which are faced in collection, compilation and presentation of data on land use are given below:
1 For illustration, mention may be made of the Conference of European Statisticians (Geneva, 1995) where detailed discussions were held on Handbook of Concepts and Definitions used in International Collection of Food and Agriculture Statistics, and the “National Demonstration Centre” on supply/utilization accounts (SAUs) organized by FAO (Harare, one week in November 1995) to train national statisticians.¤ Non-reporting: In a given year, generally about two-thirds of the countries provide some data on land use.
¤ Incomplete coverage: Data on fallow areas, pasture, forest and shifting cultivation are very rarely available. Shifting cultivation presents considerable environmental problems (and many nations today face environmental damage after centuries of this activity) yet this is one category where data are not available for various reasons.
¤ Concepts and definitions: There is no universally-accepted standard or definition for some of the classifications. Two important issues in this connection are as follows:
a. Definitions of the categories of land used by various countries are sometimes different from those given by FAO for certain items. The best example of this is that most countries take arable land as the land which is potentially cultivable, whereas the FAO definition refers to land under temporary type crops, meadow and pasture. This problem is so widespread that it may be necessary to modify the FAO terminology. This is more so because one cannot arrive at an estimate of arable land by some easy method such as adding areas sown or harvested because of multiple sowing and harvesting and inter-cropping with areas double counted.b. Definitions used by reporting countries vary considerably and items classified under the same category often relate to differing kinds of land. For example, once we arrive at a definition of pasture and forest land we are confronted with the concept of wooded land. Wooded land is applied by some countries to refer to what statisticians prefer to call wood land. In most developed countries animals graze in these areas. Statistically, the areas where animals graze are classified as pasture, while those involved in resource assessment classify them (on the basis of satellite imagery) as wooded land. Similarly there are also problems about classification of area under some plantation crops, e.g., if rubber plantations should be classified as wooded land or woodland or forest.
¤ Problem of aggregation: In view of non-reporting and under-coverage it becomes difficult to get world and regional totals. However some efforts are made to prepare such aggregates using approximations and projections. This exercise seriously affects the reliability of the data between various continents.
¤ Outputs of the data: Up until recently FAO data have been published in Production Yearbooks and other reports. However, the audience we are addressing today has wider interest in environmental matters. There is also an interest in receiving data electronically and in formats that can run easily on much more sophisticated software. Today users are keen to calculate indices, ratios and indicators and switch between graphical packages and picture images which can be transmitted and down-loaded from site to site. However to accomplish this task it is necessary to make the data more objective and complete.
ENVIRONMENTAL AND SUSTAINABILITY INDICATORS: NEW DATA NEEDS
ESS is also charged with the responsibilities on aspects of environmental statistics, particularly in the context of sustainable development. The Division has in recent years participated in the development of appropriate indicators and identification of relevant variables. Although the existing data available in ESS can be used to construct the indicators, in a recent exercise in the Division it has been identified that there are two important classes of variables, viz., those which are directly related to risk and those which are related to potential risk. Most of the ESS data available today are indicative of potential risk, i.e., they indicate the potential for environmental damage but not whether and where it actually occurs.
ESS is planning to bring out a manual on “Agricultural-environmental statistics and indicators for sustainable development, a guide for national statistical offices in the area of agriculture and the environment”. ESS has a major role here for development efforts in preparing guidelines for collection, compilation and interpretation of data as well as in imparting training to national staff. It has also been clearly identified that there are areas of new data needs which call for additional efforts. Considerable efforts have been made to expand the coverage of countries on input statistics and attempts are being made to get full coverage of data on pesticide use and use of other inputs. New areas of interest include items such as: productivity, surface water and groundwater quality, land use and soil quality/soil management, use of agri-chemicals, agricultural biodiversity (genetic and species diversity, wildlife and natural habitats and diversity), pollution and waste management, climate change, water use (irrigation) and quality, agricultural extension costs, management of fertilizer and pesticide.
The UN Department for Policy Coordination and Sustainable Development project on development of methodologies for calculating sustainability indicators as a follow-up to Agenda 21 (chapter 40) has led to the preparation of methodology sheets for indicators dealing with use of agricultural pesticides, use of fertilizers, arable land per caput, irrigated land as percent of arable land, cost of extension services provided and investment in agricultural research.
Potential sources for this kind of information are the Agricultural Census and ad hoc/regular surveys. The Division has considered this aspect while issuing guidelines to countries for conducting the census. It has also been recognized that detailed information required by policy-makers cannot be obtained through census programmes and it is necessary to restructure other sources for collection of data (Narain, 1995).
LQI: SOME OPERATIONAL ASPECTS OF DATA NEEDS
Before going to the efforts made by ESS for improving the future scenario and having listed the various aspects of LQIs data, it is necessary to mention some of the operational aspects of collection of data required for compilation of LQI. Collection and compilation of data pass through three stages. The most important in the series is the last in the sequence of creation of the data bank, i.e., the end use of the data. While making any recommendations for data collection to member countries it must kept in mind who are the data users, what are the data needs and how these data would be used.
The second stage in the sequence is the formulation of concepts and definitions of the attribute required to be measured. This question indirectly centres around the size of data collection operation, human resources and equipment required for collection of the data. The matter could be understood more clearly by looking at some of the new items recommended for inclusion in the World Census of Agriculture Programme 2000 presented in the next section. ESS has faced considerable problems while including some of the essential items on land and soil in the programme. The issue under consideration was whether the investigators who are going to collect data would be able to measure the attribute.
The third stage is the presentation of the information/data. Here, apart from the format and media of presenting data, it is also important to take into account who is the decision maker concerned with making policies for taking corrective measures. Thus it has been thought appropriate by ESS to limit its efforts to the field of agriculture in discussing the indicators. Needs of forestry and fisheries may be taken up separately to draw the direct attention of the concerned departments.
RECENT EFFORTS
ESS, having analysed the need for more reliable data to meet the more recent demands on land use and land cover and other areas connected with natural resources and environment, has taken action in two directions: first, to improve upon the methods for collecting new data, and second, using the existing data provided by remote sensing techniques, to improve the available data on land cover wherever it is feasible. In the case of the former, mention may be made of issuing guidance for collection of data under the World Census of Agriculture 2000 and development of the area frame technique for undertaking sample surveys. These efforts are briefly discussed below.
a. World Census of Agriculture 2000
The World Census of Agriculture 2000 Programme, like previous census programmes, assists countries by providing definitions, concepts, standards and guidelines for censuses in the decade 1996-2005 in order to generate a database of internationally comparable figures. This Census, like the earlier one in the series, collects a large amount of information to get a few key indicators to measure the change in the status of agricultural activity. These indicators include area under cultivation and other global land use, area under compact forest type tree cover, human-land and livestock-land ratio, pasture and grazing land, etc. A change in these indicators can point to whether current agricultural practices are sustainable.
Technology development requires more and better information on agro-ecological and socio-economic factors and on use of land and natural resources. Keeping this in view and “land theory of value”, which puts physical characteristics for use of land at the centre of economic analysis, this programme for the first time recognizes the need for collecting data on environment related issues. The present census has included, in addition to the information on size of holdings and type of irrigation facilities available, the information on distribution of agricultural land by agro-climatic/agro- ecological zones 1. It has been recommended that these data may be collected together with a minimum set of data on quality of land. This information would provide:
1 Even in countries where no work has been done on documentation of agro-ecological zones, the census can collect the following information on soil quality parameters for each holding: soil type (sand, loam, light, clay, heavy clay); soil colour (black, grey, yellow); soil depth (< 30 cm, 30-90 cm, > 90 cm); soil salinity (nil, moderate, high); surface drainage (good, moderate, poor); rate of percolation (high, moderate, low); soil degradation (light, moderate, strong, extreme); relative area (of the holding) of degradation (6%, 6-10%, 11 – 25 %. 26-30% and > 30%)¤ a framework for determining which kinds of government policy and research would be needed for agriculture on different types of land;
¤ data for developing sustainability indicators at the global, national and sub-national levels;
¤ information for establishing targets and thresholds for safe and sustainable agricultural practices; and
¤ data for an information system to analyse environmental, social and economic data in an integrated manner.
The Census Programme are also recommends augmenting existing information on agricultural practices by including more information related to area irrigated, area affected by salty soil or high water table, area with irrigation potential. As regards shifting cultivation, in addition to the question “whether shifting cultivation practised” the programme suggests gathering information on “year current parcel cleared for cultivation”. Similarly the programme has, for the first time, suggested that information on “amount of inorganic fertilizer applied per crop”, “frequency of pesticide applications per crop” and “crops with high yielding/traditional varieties of seeds” also be collected for analytical uses.
However, the agricultural census is a worldwide operation and the framework, concepts and definitions are designed keeping in view the level of literacy/knowledge of the respondents in various countries. While this allows meaningful data to be collected for policy-makers, the preparation of frames for undertaking in-depth studies and making international comparisons, there are necessary compromises made that impose some limits on the usefulness of the data for comparison purposes.
Some of the other indicators in the suggested list, like productivity, use of agri-chemicals, water management and climatic changes can only be collected by sample surveys because agricultural activity is seasonal in nature (at least in most of the developing countries where this activity is being done in the traditional manner as a major means of livelihood). The agricultural census is a one time operation in a number of countries and may be conducted during an abnormal year 2, and derived statistics can be deceptive. Thus data from agricultural censuses could be used only for limited purposes.
2 Since preparations for undertaking an agricultural census are to be done much in advance, generally it is not possible to ensure that the reference year should be a normal one.
b. Area Sample Surveys
One of the perennial problems in data collection in most developing countries is the non-availability of a suitable sampling frame. Construction of the frame in the traditional manner is a time consuming and costly affair. Often, by the time a frame is prepared it is outdated. This feature seriously affects the quality of data collected. To circumvent this problem the FAO Statistics Division has proposed to publish a technical manual on the subject. The first volume of the manual entitled “Multiple frame agricultural surveys – current surveys based on area and list sampling methods” is a contribution to the statistical design, organization and implementation of large-scale agricultural sample surveys based on current information. The manual intends to introduce the subject in a straightforward and practical manner, maintaining, at the same time, its statistical rigor. Procedures are presented from the viewpoint of the considerations and steps needed to initiate and ensure the maintenance of an agricultural statistical data collection programme where experience with area frame and multiple frame sampling methods is lacking.
The manual describes multipleframe sampling survey designs that combine an area sampling frame and complementary list frames of agricultural holdings. The list frames of agricultural holdings utilized are of two types:
¤ lists of holdings used for the estimation of agricultural items studied also through the area frame questionnaire. These are relatively short lists (easy to update and combine with the area frame) of special holdings that correspond, for a given item, to a significant percentage of the total estimate. For instance, a list of the agricultural holdings with the largest area for a certain crop or with the largest number of livestock heads;¤ list frames of holdings used for the estimation of items for which area frame estimates are not obtained. For instance, lists of holdings to estimate horticultural production.
The above-mentioned multiple frame statistical model might be the most practical way for a country to produce annual agricultural statistics: estimates of planted and harvested areas, areas intended for harvest, potential and actual yields and production of crops; livestock estimates, grain stocks, social and economic characteristics of agricultural holdings and of farming systems. The use of an area frame is necessary to geo-reference micro data, which are necessary for the development of meaningful environmental statistics.
The agricultural survey designs described in the manual represent an improvement over the more usual methods based exclusively on a list frame of holdings. In fact, apart from other considerations, the area frame methods generate more precise estimates of agricultural areas, a important item studied in all agricultural surveys. Multiple frame models utilize cartographic materials (e.g., satellite images, maps and aerial photography) and area measurement instruments for the construction of the frame and for data collection.
It should be noted in particular that satellite imagery has become available to a large number of countries only during the last few years. The availability of images as a tool for area frame construction greatly facilitates the application of multiple frame survey procedures. Furthermore, the accumulated experience of adapting area sampling frame models in many countries, and in general the increasing availability of computers for data processing provide an opportunity for fostering the use of multiple frame models for national statistical agricultural programmes.
However, the decision to base an agricultural survey on multipleframe sampling methods should carefully consider, as is the case with any statistical model, the local conditions and requirements. In fact, to be appropriate for a country or a region of a country, the statistical survey design and the implementation procedures should satisfy a number of conditions. In the first place, the sample design and the construction of the frames require trained staff and availability of suitable cartographic materials and equipment. Secondly, the survey implementation requires that certain agricultural characteristics such as proximity of the holder to the holding must be met.
The manual describes the initial steps and considerations required for the planning of a large-scale agricultural survey and refers to the problem of choosing an appropriate sampling design. The major part of the manual describes the construction of an area sampling frame using physical boundaries for strata and sample units. It covers the preparation of the complementary list frames, the replicated area sample selection, the organization and the survey data collection procedures, the multipleframe estimation methods and the calculation of sampling errors. It includes area frame construction procedures using different combinations of cartographic materials (satellite images, maps and aerial photographs), measuring instruments and equipment. It also covers the case of area sampling frame methods that use geometric (square or rectangular) sample units without physical boundaries.
The second volume of the manual to be published shortly, will illustrate a large number of national case studies on agricultural surveys based on multiple frame sampling methods prepared for FAO by experts in the field.
c. Remote Sensing
With the rapid development of space technology there is great hope of using remote sensing techniques as a means of improving land-use statistics. In particular, the use of remote sensing for agricultural statistics seems to have the greatest potential in countries where the statistical system is not good enough to produce reliable and timely data to meet the increasing needs of users.
It is extremely useful to classify land cover into main categories of land-use like forest areas and urban areas. It is a first line weapon in the monitoring of the environment (including drought degradation and fire damage to land) and in dealing with desertification, deforestation and pollution.
Although remote sensing by itself can provide only the broadest land-use information it does constitute a valuable enhancement to ground-based systems except for building sampling frames. ESS recently undertook a study of the land-use data in relation to remote sensing efforts in three countries. In these cases a key was developed (which is not unique) to make adjustment in data available from different sources. It is hoped that this method could be used for many countries in the African region. The classes which matched most closely our data are shown in Table 1.
The initial efforts have been a valuable experience and resulted in the recognition of an underestimation of arable land in one of the case studies. This exercise has shown that it is possible to use remote sensing and GIS techniques to build data for land-use classification even though the original data are land cover calibrations. The keys in each country study differed and what remains to be accomplished is the development of a standard key for widescale use especially for the Africa Land Cover Study (AFRICOVER).
TABLE 1. Various land-use or cover terms found in a comparative study in three contrasting countries
| Arable Land | Forest and woodland | ||
| agriculture in sloping areas | evergreen forest | ||
| intensive agriculture in flat areas | plantation forest | ||
| agriculture on sloping or steeplands | coniferous forest | ||
| receding rice fields (flood recession cropping) | deciduous forest | ||
| paddy rice fields | mixed forest | ||
| swidden agriculture (shifting cultivation) | secondary forest (flooded or not) | ||
| field crops horticulture | flooded forest | ||
| mangrove forest | |||
| Permanent Crops | woodlands | ||
| orchards | natural shrub land | ||
| olive trees | Mediterranean maquis | ||
| deciduous fruit trees | broadleaf forest | ||
| citrus or bananas | |||
| vineyards | Other land | ||
| highly dissected and eroded land | |||
| Permanent Pastures | low shrub and bare rocks – garigue | ||
| pastures and related grasslands | urban areas | ||
| swampy areas | marginal lands | ||
| marshes | barren land or rocks | ||
| grassland (flooded or not), | eroded land and beaches | ||
| abandoned grassland | |||
| savannah | |||
| abandoned shrub land | |||
| degenerated forest or maquis | |||
The above list is a mix of categories coming from three different land cover legends worked out during the study.
OTHER RECENT EFFORTS UNDERTAKEN TO IMPROVE THE QUALITY OF DATA AND INFORMATION
In addition to the above-mentioned efforts in improving the reliability and coverage of the data, ESS has undertaken reviews to adjust the area under forest cover in relation to the results of the FAO Forest Resources Assessment Project whenever possible, thus recognizing that such assessments are complementary to national statistics. Furthermore, the Division has also taken on board some of the data from the Water Resources, Development and Management Service of FAO on irrigated areas as presented in their AQUASTAT files to make adjustments similar to those made with remote sensing information.
BIBLIOGRAPHY
FAO. 1985. The ICS: The Interlinked Computerized Storage and Processing System of Food and Agricultural Commodity Data Users’ Manual. FAO/ESS, Rome.
FAO. 1995. Programme for the World Census of Agriculture 2000. FAO Statistical Development Series No. 5. FAO/ESS, Rome.
FAO. 1996. Multiple Frame Agricultural Surveys – Current Surveys based on Area and List Sampling Methods. Volume I. FAO/ESS, Rome.
FAO. WAICENT. (World Agricultural Information Centre). Netscape location: http://www.fao.org
Narain, P. 1995. Crop Cutting Surveys: Planners’ View. Paper submitted to the 50th Session of International Statistical Institute, Beijing.
World Bank. 1995. Land quality indicators. C. Pieri, J. Dumanski, A. Hamblin and A. Young. World Bank Discussion Papers,315. World Bank, Washington, D.C.