KARL EBERHARD HALLERKARL EBERHARD HALLER is a senior research officer at the Federal Research Organization for Forestry and Forest Products, Reinbek, Federal Republic of Germany.
The reason for developing a computer processing program was the inventory of the national forests of Liberia, which has been carried out since 1960 in collaboration between the Liberian Bureau of Forest and Wildlife Conservation and the German Forestry Mission to Liberia in the framework of a bilateral technical assistance project.
Up to 1963 processing of the inventory data was carried out in the Federal Republic of Germany by means of a conventional punch card system. The transmission of documents, the transfer of data from tally sheets to punch cards and the production of the lists of results required a great deal of time and, in particular, the painstaking search for errors presented considerable difficulties and made the work slow and costly.
In 1965, an IBM 1401 data processing system was installed for the iron ore company LAMCO in Liberia and placed at the disposal of the inventory team on an hourly rental basis of about $100. Accordingly, the Forest Inventory Section of the Federal Research Organization for Forestry and Forest Products at Reinbek/Hamburg was commissioned by the Ministry of Food, Agriculture and Forestry of the Federal Republic of Germany to prepare a computer program. This program has permitted the smooth and rapid processing of inventory data in the country of origin, and is, according to the broad range of inventory tasks in Liberia, equally well adapted to extensive reconnaissance surveys as to very intensive inventories of logging areas. The results of these inventories, as well as a detailed description of the data processing program, are being published in a series of reports issued by the German Forestry Mission to Liberia.
On-the-spot evaluation and the possibility of swift production of intermediate results are to be considered as advantages or even as necessities for the effective conduct of forest inventories.
The preparation of a complex computer program always involves a great deal of work for programming and relatively high costs for testing. It was therefore highly desirable to construct a program that could be used not only for the task on hand, but also for other tropical inventories.
In the preparation of the program specific attention was given to the following restrictions and considerations:
1. The program should be suitable for a flexible and comprehensive evaluation of data that are gathered for mixed natural tropical forests either in an initial inventory or in an independent subsequent inventory.The first restriction emphasizes the special features of inventory work in mixed tropical forests, such as:
(a) multitude of tree species;(b) low significance of stand indices (e.g., mean height, average age);
(c) special features of volume definitions, measurement and computation (e.g., stem volume and not tree volume, reductions for defects);
(d) assessment of the present status only, without growth determination.
The second restriction, initial inventory or independent subsequent inventory, states explicitly that the evaluation concerns a single-phase inventory and not a regression sample. However, this does not exclude that, independently of the ground inventory, certain quantitative area information may be obtained from existing sources or from photointerpretation, for use in the selection of sampling units or in the derivation of the total volume. Area information, as acreage of a particular national forest, is considered as a given reference value, and is not derived by the program. However, area proportions, such as soil, or terrain class proportions, are usually determined in the course of the ground phase of an inventory, and their determination must be possible with the aid of the lists produced by the program.
2. The program should be capable of many modifications and not be tied to a particular data recording procedure. It cannot be expected that any specific recording procedure, however carefully designed, will be used completely unchanged now and in the future. A program should therefore be flexible enough:
(a) to be applicable from the beginning to data collected according to various sampling procedures;(b) to continue to be usable as it is, or with only minor program changes, when measuring techniques change.
3. The program should consist of a group of subprograms which can be executed individually, rather than providing integrated data processing, i.e., producing all lists and tables in an unbroken sequence from a single input.
Integrated data processing requires the presence of a computer system with external storage units. Such systems are probably available for use in industrial countries but, for the time being, only in a few developing countries. The IBM 1401 system which was available in Liberia was not equipped with tape or disk storage units.
A further objection to programming an integrated data processing for such inventories is that on initial tropical forest inventories the species or forest types of special interest (and the corresponding output lists) are frequently not apparent prior to the first evaluation. Many such special evaluations can be made through the use of available subprograms by sorting out a particular portion of the punch cards and arranging them in a given sequence. Integrated data processing requires the detailed advance determination of every question to be asked, often at the expense of flexibility.
4. The program should include a detailed and comprehensive check of all input data.
Data processing can be carried out effectively only if the basic data are on hand in their entirety and without errors. For this reason every computation must be preceded by thorough checking of the completeness, correctness and plausibility of the recorded -and punched data.
Furthermore, it was assumed in preparing this program that only a limited amount of computer time is available. It was attempted, therefore, to obtain a maximum amount of data from each run, even at some sacrifice of clarity in the lists. However, these lists are supposed to be so arranged that all of the information needed for an inventory report can be extracted from them without difficulty.
In Liberia the following configuration was available:
(a) Central processing unit IBM 1401/12K with optional feature punch in same card;(b) Card read punch IBM 1402;
(c) Printer IBM 1403 with 132 print positions.
This is more or less the minimum equipment for such a system. Since external storage units are lacking, the data cards must partly fulfill the storage function during the processing. For each subprogram the necessary cards must be picked out by mechanical sorting and fed to the card read punch in a definite sequence. Therefore, a sorter must also be available.
The principle use of the IBM 1401 system is for commercial purposes. It is especially well adapted for the production of extensive lists and tables, but less well for mathematical analyses, such as variance or standard error calculations. For example, the mere extraction of square roots presented almost insuperable technical difficulties.
The program for IBM 1401 can be either written in machine or in the assembler languages SPS or AUTOCODER. Problem oriented languages, such as FORTRAN, ALGOL or COBOL are usable to a very limited extent only. The programs under consideration were written in AUTOCODER.
The IBM 1401 system is still the most widely distributed electronic computer in the world. Almost every country boasts of firms or agencies that have this system available, and the possibility of renting such a system on an hourly basis should exist almost everywhere, although its production has already ceased. It will mainly be replaced by third generation systems such as the IBM 360/series. However, these new systems are mostly 1401 compatible, which means that programs written for the IBM 1401 can automatically be transformed into machine programs for the new systems.
The respective terms are identified by italics.
Each punch card contains 80 vertical columns which are identified by numbers. A group of columns, used to record a single item of information is called a card field.
Cards with the same layout and function are called the same type of card and all belong to one deck. All of the cards in a given deck are marked with the same deck number.
FIGURE 1. - Arrangement of the recording from used in Liberia.
In the course of evaluation several decks are needed or produced:
1. The cards of the basic deck are punched from the recording sheets (see sample plot form, Figure 1). For particular subprograms only certain portions of the basic cards - subdecks - are needed.2. The cards of a master deck supply information not contained in the basic deck but needed during processing.
3. The cards of a summary deck are produced automatically during the evaluation. They contain condensed data and are used in subsequent subprograms.
4. The cards of a program deck tell the computer in machine language which operation to perform and in what order.
The programming of data processing normally begins with the definition of the contents of the necessary output lists. Following this, it is decided what information is needed for the assembling of these lists and how it can be derived. This phase of program preparation is called the problem study, which consists of the following:
1. Design of the output list (unit of report, subject matter, fields and groups)
2. Design of suitable number codes for identification and classification of data
3. Design of the basic punch card
4. Design of the auxiliary lists, master and summary cards
5. Definition of the complete work Bow
6. Detailed specification of the subprograms.
Figure 1 shows the arrangement of the recording form used in Liberia. Since the form contains preprinted column numbers, it represents simultaneously the content of the basic punch card. One card has to be punched for each line with recorded data. Each card contains always the plot data together with the tree and/or reproduction data.
The form is designed in such a way that normally only one form must be filled out in the field for each unit of record (sample plot), and that card punching is simple and reliable.
The form has been prepared for a specific inventory design, but the content of the single card fields is changeable within certain limits in order to allow the processing of differently designed inventories with the same program.
In Liberia the main objective of the inventories is to give information for single national forests or parts of them (units of report) with a minimum size of about 25,000 hectares. The forests were demarcated by means of aerial photographs. A further subdivision into photo strata is not possible at present. Therefore, a pure ground survey has been carried out in the following way:
Each unit of report is subdivided into squared blocks of equal size (primary units).
In each block or in a randomly selected part of the blocks two transects of fixed length (secondary units) are selected by random (restricted random or two-stage sampling).
Along these transects circular plots are distributed at fixed intervals (units of record).
Trees above girth limit (40 centimeters DBH) are enumerated within circles of 500 square meters. For these trees the species, the reference diameter (at breast height or above buttress) and the quality of logs 5 meters in length (the reason for rejecting a particular log) were determined and recorded.
For trees from 10 centimeters at DBH to girth limit (reproduction trees) only the species and the diameter at breast height (in 5 centimeter classes) are recorded. These trees are enumerated within a concentric circle of 100 square meters.
The sampling unit field (columns 6 to 9) can have considerable variation in content, depending on the sampling design.
1. Simple random selection of sampling units of equal size in this case the four-column block-transect field serves to identify the sampling unit; the sample plot number field remains empty or may serve for the subdivision of sampling units.2. Restricted or stratified selection of sampling units (plots or lines) of equal size within blocks or strata this corresponds with the system normally used in Liberia.
3. Two-stage sampling with blocks as primary and lines or plots of equal size as secondary sampling units.
The evaluation of the data for one unit of report is carried out in the frame of 8 subprograms, which are partly dependent on each other. Figure 2 shows a flow chart in which the various media (records, cards and lists) and the different machines - key punch (KP) sorter (S) and computer (1401) - are represented by different symbols. The data flow is marked by arrows. The type of card is identified by a number and by a letter - (P = program deck, M = master deck, S = summary deck, and 01 = basic deck).
The following items are required before processing can start:
1. Codes used during enumeration
2. Type and number of the enumerated sampling units
3. Log volume tables to be used.
KEY PUNCHING OF DATA CARDS
Processing starts with the key punching of the data from the recording forms. The following card checking program catches all errors regarding the entirety of the data, or errors made in the field or during the punching, where a code number has been used which is not within the frame set by the coding instructions. For the classification data (columns 12 to 31 and 54 to 62) punching errors which are within the frame set by the coding instructions can only be caught by verification of each punched card, eventually by control-punching. In Liberia, applying the card checking program and punching with great care obviated the need to verify each card.
The basic cards (01 ic) must be sorted before starting the first subprogram because there are no entries in the columns for log and tree volume.
CARD CHECKING AND VOLUME COMPUTATION PROGRAM
This program is the basis for all the following programs. It checks the correctness, plausibility and completeness of the contents of all cards punched from the recording forms. Any errors or discrepancies in the basic card deck are printed in the errors field of a check list. In addition, this subprogram computes and punches into the still incomplete cards of deck 01 (deck 01 ic) the volumes of the merchantable logs together with the merchantable tree volume. The contents of all cards are printed out in the check list.
Besides the program deck (P20), the program needs master deck M21 containing codes for unit of report, strata and gradient classes; master deck M22 containing codes of all species belonging to each species group; and master deck M23 containing the volume table with one card for each 5 centimeter of diameter class. If several volume tables are to be used, the volume computation and the card checking program must be carried out in two phases.
The result of this subprogram is the basic deck 01 (now complete) and the exact figure on the number of cards of deck 01 and of the subdecks Plot, Repro, and Trees.
STRATUM AREA PROGRAM
Apart from program deck P30 and master deck M31, containing the relevant strata codes, the subdeck Plot, which consists of all cards with card number 01 in columns 16 and 17, is used for running this program.
The result of this run is the area list and a summary deck S32. The area list is used for calculation of blow-up factors needed for the reproduction and the diameter program (M41, M46, M52, M56). The area list is produced for each unit of report, whereas the summary cards S32 are used for the production of list NVS and the standard error of ratio estimates. However, both subprograms have not been necessary in Liberia and are not yet programmed.
REPRODUCTION PROGRAM
The reproduction program consists of two parts, one which produces the list RF (Reproduction/Forest) by means of program deck P40 and blow-up factor card M41 for the forest as unit of report and a second one with program deck P45 and one blow-up factor card for each stratum (M46) for the production of the list RS (Reproduction/Stratum) for each stratum. Both lists contain only reproduction trees; therefore only the subdeck Repro (all cards with 1 in column 54) is needed. List RF is produced first and list RS second, since this arrangement calls for less sorting.
DIAMETER PROGRAM
Only subdeck Tree (cards with no "blank" in column 23) is used.
Even here the lists DNVF (Diameter, Number of trees, Volume/Forest) and DNVS (Diameter, Number of trees, Volume/Stratum) differ only with regard to unit of report. The same program deck (P50) and masterdeck (M51), which contain the species within each species group, can be used for the production of both lists. The master card M52 gives the blow-up factor for the forest, and the masterdeck M56 the respective factors for each stratum.
BLOCK TRANSECT PROGRAM
The cards of the subdeck Tree are also used for this subprogram, which is only partly programmed because the list NVS (Number of trees, Volume/Stratum) has not been of urgent need within the Liberian inventories.
List NVF (Number of trees, Volume/Forest) gives a general view of the spatial distribution of the trees and their merchantable volume.
Summary cards S61/62 are punched for the merchantable volume of trees tallied in the sampling units and serve as data cards in the standard error program 1. Summary cards S61/62 of incomplete blocks or sampling units are selected on the basis of the area list.
STANDARD ERROR PROGRAM 1
This program gives in list SE1 for the forest, the species and species group in:
Line 1. relative variance (square of standard error per cent) derived from the variance within blocks;Line 2. relative variance (square of standard error percent) derived from total variance.
Apart from program deck P71 and master deck M51 (species within species group) a further master card M72 which contains the number of complete blocks, the number of sampling units and the number of degrees of freedom, is needed.
The subprograms previously described have been used for the evaluation of nearly all inventories carried out by the Liberian-German inventory team. Thus the data from a total of about 50,000 sample plots and 200,000 basic cards have been processed. Accordingly, the programs have been adequately checked in practice. The listed programs or the relevant card sets can be made available by the author upon request.
The card checking program has proved to be extremely valuable. This program makes it possible to produce practically error-free lists in a way impossible to achieve even with the aid of exhaustive, thorough and precisely prescribed hand checking of the recording forms and the checking of each punch card by double key punching.
Difficulties may be caused by the need for repeated sorting which causes heavy physical stress on the cards. This is aggravated by the temperature and humidity changes that take place while the cards are in storage. These changes are especially noticeable in the tropics and cause distortion which leads to the tearing of cards during sorting.
It is therefore necessary to carry out the evaluation as soon as possible after the cards have been punched.
Improvement seems to be possible by further increasing the flexibility. Only some examples may be mentioned:
1. Further extension of the sampling unit field in order to allow even three or more stage samplings.2. Double species classification. At present each species belongs to one species group only. An additional classification field (e.g., utilization classes, such as veneer, construction timber) should be introduced so that each species can belong to several classes.
3. The present kind of volume computation cannot be considered as a general solution. It is based on a sectional appraisal and considers only whether a particular log is merchantable or not. Instead of the volume of merchantable logs and the merchantable tree volume, the basal area, the total stem volume and stem volumes according to different specifications (A, B and C) could be punched in the respective columns and several lists could be produced with the diameter and block-transect program by calling the various arguments.
4. It has been shown that the programming of variances and standard errors is extremely difficult for such a computer, which is best suited for commercial problems. It seems wise, therefore, to use this computer for the processing of lists and summary cards and to carry out more complicated standard error calculations by means of summary cards and a computer which is better suited for mathematical analyses.
INTERNATIONAL UNION OF SOCIETIES OF FORESTERS
The International Union of Societies of Foresters will hold its initial Congress at Washington, D.C. in 1969. Host for the Congress will be the Society of American Foresters. The agenda for the Congress will include two half-day plenary sessions open to all delegates, and two closed meetings of the International Council and the Governing Committee. The International Council will consider the organization's constitution and by-laws with a view to their adoption by the Congress, and will elect officers and members of the Governing Committee.
The International Union, organized in 1966 during the Sixth World Forestry Congress in Madrid, is composed of national societies of professional foresters. Its purpose is to advance the progress, status, and effectiveness of its members by means of societies throughout the world. Union activities include seeking higher educational standards for professional foresters; emphasizing the responsibility of the profession to the public; and functioning on an international basis in the interests of national societies and of the forestry profession.
Forestry societies may obtain information regarding the Congress and membership requirements of the Union by writing to the Executive Director, Apt. 841, 1500 Massachusetts Avenue N. W., Washington, D.C. 20005.
