AN OVERVIEW OF GIS MANIPULATIONS AND ANALYSES FOR THE STRATEGIC ASSESSMENT OF FISH FARMING POTENTIAL IN AFRICA:WARM-WATER FISH FARMING WITH ONE AND TWO CROPS PER YEAR AT SUBSISTENCE AND COMMERCIAL LEVELS
by
Vittoria Zichella
Viale Mazzini, 50
50132 - Florence, Italy
1. INITIAL DEFINITION OF THE CRITERIA FOR THE ANALYSES
Subsistence fish farming
Agricultural by-products (productivity and variety) as inputs
Commercial fish farming
2. FURTHER DEFINITION OF THE CRITERIA FOR THE ANALYSES BY MERGING INITIAL CRITERIA
Subsistence fish farming
Economic criteria: local market demand and inputs from agricultural by-products
Commercial fish farming
3. ORIGINAL INPUT FILES
In the first step of the analysis, ORIGINAL INPUT FILES created by R. Refk within an earlier phase of the study have been identified and copied in D:\VIC.
AFPSC4CL.GIS>> PRE.GIS (PRECIPITATION)
W10SO.GIS>> WTMR.GIS (WATER TEMPERATURE)
AFSLO10M.GIS>> SLO.GIS (SLOPE)
AFTXT10R.GIS>> TXT.GIS (SOIL TEXTURE)
POPSUBSB.GIS>> POPSUBR.GIS (POPULATION DENSITY - SUBSISTENCE)
POPCOMSB.GIS>> POPCOMR.GIS (POPULATION DENSITY - COMMERCIAL)
LGPCLI.GIS>> LGPCLIR.GIS (COMBINATION of LGP and CLIMATE)
RIVPCMI.GIS>> RIV.GIS (PERENNIAL STREAMS and RIVERS)
RDPMC.GIS>> RDS.GIS (PAVED and MOTORABLE ROADS)
The standard format for all such files is:
Unless otherwise indicated, all GIS operations were performed using ERDAS (Earth Resources Data Analysis System) software, ver. 7.5, by ERDAS Inc., 2801 Buford Highway, Suite 300, Atlanta, GA - 30329 - USA.
Such ORIGINAL INPUT FILES were then thresholded and recoded (RECODE option) in the following way:
| unsuitable areas (land and water) | > >CLASS 0 |
| marginal areas | > >CLASS 1 |
| suitable areas | > > CLASS 2 |
| optimum areas | > >CLASS 3 |
As a general rule, pixels having a 0 (i.e. unsuitable) value in any of the input files were discarded from the analyses.
The only exception to this rule will be highlighted further on.
The sub-directory D:\VIC\INPUTS contains all the RECODED INPUT FILES for ERDAS analysis (.GIS extension).
The next two pages give a detailed description of the individual recoding of each input file, with the meaning of each class.
3.1 Recoded input files for subsistence fish farming
1) PRER1.GIS(ANNUAL RAINFALL)
0 = unsuitable & water
1 = 700 – 999 mm
2 = 1000 – 1199 mm
3 = >1199 mm
2) WTMR1.GIS(WATER TEMPERATURE)
0 = unsuitable & water
1 = =>22^ for 8 consecutive months and > = 14^ for 4 months
2 = 22^–26^ for 12 consecutive months
3 = =>26^ for 12 consecutive months
3) SLOR1.GIS(SLOPE)
0 = water
1 = > 30%
2 = 8–30%
3 = 0–8%
4) TXTR1.GIS(SOIL TEXTURE)
0 = unsuitable & water
1 = 50–75% coarse
2 = mainly medium
3 = mainly fine
5) POPSUBR1.GIS(LOCAL MARKET DEMAND - SUBSISTENCE FARMS)
0 = water
1 = <1 person/sqkm
2 = 1–5 persons/sqkm
3 = >5 persons/sqkm
6) LGPCLIR1.GIS(INPUTS FROM AGRICULTURAL BY-PRODUCTS)
0 = unsuitable & water
1 = low (1–3)
2 = moderate, few crops (4–8)
3 = moderate to good, varied (9–12)
3.2 Recoded input files for commercial fish farming
1) PRER1.GIS(ANNUAL RAINFALL)
0 = unsuitable & water
1 = 700 – 999 mm
2 = 1000 – 1199 mm
3 = > 1199 mm
2) RIVR1.GIS(DENSITY of PERENNIAL STREAMS and RIVERS)
0 = unsuitable & water
1 = 1–4 (2'×2'-pixels containing rivers in a 10'×10'-pixel)
2 = 5–8 (2'×2'-pixels containing rivers in a 10'×10'-pixel)
3 = 9–23 (2'×2'-pixels containing rivers in a 10'×10'-pixel)
3) WTMR1.GIS(WATER TEMPERATURE)
0 = unsuitable & water
1 = =>22^ for 8 consecutive months and > = 14^ for 4 months
2 = 22^–26^ for 12 consecutive months
3 = =>26^ for 12 consecutive months
4) SLOR1.GIS(SLOPE)
0 = water
1 = > 30%
2 = 8–30%
3 = 0–8%
5) TXTR1.GIS(SOIL TEXTURE)
0 = unsuitable & water
1 = 50–75% coarse
2 = mainly medium
3 = mainly fine
6) POPCOMR1.GIS(LOCAL MARKET DEMAND - COMMERCIAL FARMS)
0 = water
1 = <5 persons/sqkm
2 = 5–25 persons/sqkm
3 = >25 persons/sqkm
7) RDSR1.GIS(DENSITY of PAVED and MOTORABLE ROADS)
0 = unsuitable & water
1 = 1–4 (2'×2'-pixels containing roads in a 10'×10'-pixel)
2 = 5–8 (2'×2'-pixels containing roads in a 10'×10'-pixel)
3 = 9–22 (2'×2'-pixels containing roads in a 10'×10'-pixel)
8) LGPCLIR1.GIS(INPUTS FROM AGRICULTURAL BY-PRODUCTS)
0 = unsuitable & water
1 = low (1–3)
2 = moderate, few crops (4–8)
3 = moderate to good, varied (9–12)
4. COLOUR PLOTTING
The RECODED INPUT FILES for both SUBSISTENCE and COMMERCIAL ANALYSIS were overlayed (OVERLAY option) - with output of the maximum value - with a mask called SEA_LAND.GIS that scores 0 at sea and 1 on land. This operation produced a separation of water (score 0) and unsuitable land areas (score 1) - which was needed for plotting purposes - and a new range of scores, from 0 to 4 (filenames *R04.GIS).
Three files - SLOR1.GIS, POPSUBR1.GIS and POPCOMR1.GIS - do not have an unsuitable land areas class.
The overlayed files were then converted (*12.IMG) to IDRISI format (ver. 3.2, by Graduate School of Geography, Clark University, Worcester, MA - 01610 -USA) and individual colour plots were made through PIZZAZ PLUS software (by Application Techniques Inc., 10 Lomar Park Drive, Pepperell, MA - 01463 - USA).
Another mask called ALCOM150.IMG was created - scoring 1 for NONALCOM countries, 5 for ALCOM countries and 0 for water - as well as a COUNTRY BOUNDARIES file (CNTBD6.IMG).
Each RECODED INPUT FILE was overlayed (OVERLAY option in IDRISI) with ALCOM150.IMG and CNTBD6.IMG - with output of maximum value - and a window was defined for new plots showing only Southern Africa (WINDOW option in IDRISI).
5. AREA CALCULATION
Surface extension was calculated on a country-by-country basis for each of the 5 classes (0–4) of each criterion and the results were converted through QUATTRO PRO (ver. 5.0, by Borland International Inc., PO Box 660001, Scotts valley, CA - 95067-0001 - USA) to EXCEL format (PRE.XLS, WTMR.XLS, SLO.XLS, TXT.XLS, POPS.XLS, LGPCL.XLS, RIV.XLS, POPC.XLS and RDS.XLS).
6. PRELIMINARY WORK FOR FINAL ANALYSIS
6.1 Subsistence fish farming
According to the re-definition of the criteria for analysis (see first page), 4 out of 6 initial layers were merged into 2 new categories.
This was achieved through the MATRIX option, which combines two input files and outputs to a new file scoring 0 for any combination containing class 0, and gives other values for each different combination of input classes. Such an output allows each combination to be identified and recoded as needed.
The final categories used for SUBSISTENCE ANALYSIS are listed below, with the corresponding files:
category ANNUAL RAINFALL;
file PRER1.GIS: recoded input file (range 0–3), as it is.
category WATER TEMPERATURE;
file WTMR1.GIS: recoded input file (range 0–3), as it is.
category OTHER PHYSICAL CRITERIA: SOIL TEXTURE and SLOPE;
file SLTXR1.GIS: output of MATRIX analysis between SLOR1.GIS and
TXTR1.GIS, recoded to 0–3 with equal weights on both files, as shown
below:
| TM | TS | TO | |
| SM | SM-TM > 1 | SM-TS > 1 | SM-TO > 2 |
| SS | SS-TM > 1 | SS-TS > 2 | SS-TO > 3 |
| SO | SO-TM > 2 | SO-TS > 3 | SO-TO > 3 |
category ECONOMIC CRITERIA: LOCAL MARKET DEMAND and INPUTS
from AGRIC. BY-PROD.;
file LMDIR1.GIS: output of MATRIX analysis between POPSUBR1.GIS and
LGPCLIR1.GIS, recoded to 0–3 with equal weights on both files, as shown
below:
| PM | PS | PO | |
| LM | LM-PM > 1 | LM-PS > 1 | LM-PO > 2 |
| LS | LS-PM > 1 | LS-PS > 2 | LS-PO > 3 |
| LO | LO-PM > 2 | LO-PS > 3 | LO-PO > 3 |
6.2 Commercial fish farming
According to the re-definition of the criteria for analysis (see first page), 5 out of 8 initial layers were merged into 2 new categories.
Again, this was achieved through the MATRIX option.
The final categories used for COMMERCIAL ANALYSIS are listed below, with the corresponding files:
category WATER AVAILABILITY: ANNUAL RAINFALL and DENSITY of
PERENNIAL STREAMS and RIVERS;
file WATAVR1.GIS: output of MATRIX analysis between PRER04.GIS and
RIVR04.GIS, recoded to 0–3 with double weight on PRER04.GIS, as shown
below:
| PU | PM | PS | PO | |
| RU | RU-PU > 0 | RU-PM > 1 | RU-PS > 2 | RU-PO > 3 |
| RM | RM-PU > 1 | RM-PM > 1 | RM-PS > 2 | RM-PO > 3 |
| RS | RS-PU > 1 | RS-PM > 1 | RS-PS > 2 | RS-PO > 3 |
| RO | RO-PU > 2 | RO-PM > 2 | RO-PS > 3 | RO-PO > 3 |
N.B.: In this case, a special condition existed for the analysis.
Unlike other criteria, ANNUAL RAINFALL and DENSITY of PERENNIAL STREAMS and RIVERS can alternatively be suitable, since they are both sources of water. Therefore, either of the two can be absent (but not both).
Since the MATRIX operation would eliminate any combination containing class unsuitable (score 0), files recoded to 0–4 were used (having class unsuitable = 1), instead of those recoded to 0–3.
category WATER TEMPERATURE;
file WTMR1.GIS: recoded input file (range 0–3), as it is.
category OTHER PHYSICAL CRITERIA;
file SLTXR1.GIS: same as SUBSISTENCE FISH FARMING.
category ECONOMIC CRITERIA: LOCAL MARKET DEMAND, INPUTS and
ROAD DENSITY;
file LMDRIR1.GIS: output of a two-step MATRIX analysis between
POPCOMR1 and LGPCLIR1 (same procedure as for POPSUBR1.GIS and
LGPCLIR1.GIS, in SUBSISTENCE FISH FARMING), and RDSR1.GIS, recoded
to 0–3 with equal weights on the three files (labelled A, B and C), as shown
below:
| FILE A | M | M | M | M | S | M | S | M | S | O |
| FILE B | M | M | M | S | S | S | S | O | O | O |
| FILE C | M | S | O | S | S | O | O | O | O | O |
| RECODING | 1 | 1 | 1 | 2 | 2 | 2 | 2 | 3 | 3 | 3 |
6.3 Colour plotting
The new recoded input files (SLTXR1.GIS, LMDIR1.GIS, WATAVR1.GIS and LMDRIR1.GIS) were overlayed with SEA_LAND.GIS - producing files: SLTXR04.GIS, LMDIR04.GIS, WATAVR04.GIS and LMDRIR04.GIS, with scores 0–4 - and then converted to IDRISI format (TXSL.IMG, LMDI.IMG, WATAV.IMG and LMDRI.GIS).
Then, they were further overlayed (OVERLAY option in IDRISI) - with output of maximum value -with the COUNTRY BOUNDARIES file CNTBD5.IMG (TXSLBD.IMG, LMDIBD.IMG, WTAVBD.IMG and LMDRIBD.IMG) and plotted.
Such new files were also overlayed with the mask ALCOM150.IMG and the COUNTRY BOUNDARIES file CNTBD6.IMG, then windowed and plotted again with the usual colour scheme.
6.4 Area calculation
Surface area was calculated on a country-by-country basis for each of the 5 classes (0–4) of each new criterion and the results were converted to EXCEL format (SLTX.XLS, LMDI.XLS, WATAV.XLS and LMDRI.XLS).
7. FINAL ANALYSIS
In order to combine the 4 final categories for both SUBSISTENCE and COMMERCIAL FISH FARMING, a model called “NOV17” was written with GISMO option and saved into D:\VIC\MODELS.MLB.
Such model is designed to combine the 4 input files and create an output file with the following classes:
The syntax of the model is the following:
Data
input A file ask “temperature”;
input B file ask “water supply”;
input C file ask “physical criteria”;
input D file ask “economic criteria”;
output OUT file ask “desired output”;
output SCREEN display;
Start
OUT=CONDITIONAL {
(A eq 3 and B eq 3 and C eq 3 and D eq 3) 8
(A eq 3 and B ge 2 and C ge 2 and D ge 2) 7
(A eq 2 and B eq 3 and C eq 3 and D eq 3) 6
(A eq 2 and B ge 2 and C ge 2 and D ge 2) 5
(A eq 1 and B eq 3 and C eq 3 and D eq 3) 4
(A eq 1 and B ge 2 and C ge 2 and D ge 2) 3
(A eq 1 and B eq 1 and C eq 1 and D eq 1) 2
(default) 0};
SCREEN=OUT;
End
Such model has the advantage of being suitable for both SUBSISTENCE and COMMERCIAL ANALYSIS, since it prompts the operator for input and output filenames (“ask” option).
In order to do a random check of the results on the input files, a cursor display list (D:\VIC\INPUTS\4LAYSC.CDL) was created with INQUIRE option. Such list allows the reading of attributes - as file and map coordinates, class values…-of multiple files (not on the display) just by typing a certain key.
The randomly-checked pixels confirmed the proper functioning of the model.
7.1 Colour plotting
The two output files - SUBS.GIS and COMM.GIS - were overlayed with SEA_LAND.GIS (this is why the model leaves class 1 empty) and converted to IDRISI format for plotting (SUB1.IMG and COMM1.IMG).
Since class 2 (all marginal) contained only 3–5 pixels, the two files were recoded within the range 0–7 (SUB2.IMG and COMM2.IMG).
A new COUNTRY BOUNDARIES file (CNTBD8.IMG) was created for the new outputs and overlayed with maximum output (SUB2BD.IMG and COMM2BD.IMG).
Also, a new colour scheme was prepared - as consistent as possible with the previous one - and plots were made.
In this case, too, plots were repeated for ALCOM countries, by overlaying the two output files with a new mask called ALCOM180.IMG and a new COUNTRY BOUNDARIES file called CNTBD9.IMG.
7.2 Area calculation
Surface extension was calculated on a country-by-country basis for each of the 8 classes (0–7) of the two output files and the results were converted to EXCEL format (SUB.XLS and COMM.XLS).
8. ADDENDUM: ADDITIONAL WORK ON AFRICA STUDY TO UPDATE ROADS LAYER WITH MADAGASCAR
Due to the lack of roads for Madagascar in WDBII coverage - which had been used in the first phase of the study - the EROS DATA CENTER coverage (AFRDS) was restored from tape and rasterized.
Rasterization was performed only on a subset of the coverage centered on Madagascar (LL corner = 43.137, -25.953, UL corner = 43.137, -11.277), to speed up the process.
Cell size for rasterization was 2'×2' (0.0333×0.0333 dd). Once in raster format (MADROADS.GIS, and 8-bit equivalent MADRD8B.GIS), hard surface and gravel roads were merged into one class and remaining road types were reclassified as background (MADRD01.GIS).
High speed roads are missing in Madagascar.
Conversion to IDRISI was necessary before performing aggregation. Every 5×5 pixels were aggregated into one, giving a cell size of 10'×10' (0.1666×0.1666 dd). After the aggregation procedure, class values ranged from 0 to 15 (MADRD10.GIS).
Such values were reclassified according to the following thresholds:
no roads = unsuitable (recoded to 1 for overlay)
1 – 4 = marginal (recoded to 2)
5 – 8 = suitable (recoded to 3)
> = 9 = optimum (recoded to 4)
Different versions of ROADS layer for the whole Africa were updated by stitching of the Madagascar subset with proper recoding.
Matrix analysis was performed between LOCAL MARKET DEMAND & INPUTS layer and ROADS layer and the resulting ECONOMIC LAYER checked by comparison with its previous version.
Final overlay with WATER TEMPERATURE, WATER AVAILABILITY and SOIL TEXTURE & SLOPE layers was achieved through GISMO model “NOV17” (see Africa report).
All amended layers and final “commercial” overlay were converted to IDRISI format and plotted in colour.
Area calculation programmes were run on updated files and outputs converted to Excel format.
At the same time, Africa country boundaries were also updated with Eritrea, through rasterization of coverage AFCNTLLC at 10'×10' (both as lines and as polygons).
Updated boundaries were used in new colour plots (NEWBND8 - lines) and in area calculations (AFCNTNEW - polygons).
