Thursday, September 29, 2011

Export land model analysis of food production and consumption—deriving food energy conversion factors

As I indicated in the last post, food energy conversion factors are vital for converting quantities of different foods into a common energy unit.  It is this conversion that will allow me to compare the food production, consumption, and import/export rates for different years within the same country as well as between countries or regions. 

As also indicated in my last post the problem, for me, is that FAOSTAT doesn’t present the conversion factors for all of the food types.  Well, actually FAOSTAT, in providing its Food Balance sheets, doesn’t present any conversion factors at all.  It’s just that I can derive conversion factors for certain food types, based on the information presents about human food energy consumption per capita.  However, there are significant types of food used for purposes other that human consumption (i.e., as feed for livestock), and not accounting for these other foods presents a real problem for within- and cross-country comparisons, as well as getting an accurate estimate of total food energy production.

This post discusses how I addressed this problem and derived my own set of conversion factors for all of the food types typically listed in the Food Balance Sheets that FAOSTAT reports.

Part of the answer: the Food Composition Table in Annex 1 of the Handbook

I already mentioned the Food Balance Sheet Handbook as an informative reference in the context of providing definitions for various terms presented in the Food Balance Sheets.  However the Handbook in an appendix also provides a table of food composition energy contents in units of kcal/100 gm for over 500 different food items.  Again for illustrate purposes I will just show a portion of the Food Composition Table (henceforth, "Table") corresponding to various cereal food products.

At first this might seem to solve the problem—we have all the major food types an their energy contents, so we should be able to just plug this value into to appropriate categories in the Food Balance Sheet of interest and thereby derive the totals of food energy production, consumption etc...

The problem with doing this is that many, but not all of the food categories presented in the Food Balance Sheets are actually composites of several different food items.

Again, wheat provides a good example.  In the working example for Australia 2003 shown in the last post there is a single row presented for “wheat.”  As explained in the last post, I derived a food energy conversion factor of 2.97 kcal/gm for the “wheat” category.

However, the Table shows ten different wheat-based food types, and, there energy content ranges from 2.13 to 3.82 kcal/gm (see above table red circle).

From this, I surmise that the energy conversion factor I derived for Australia 2003 is actually a composite of these ten different wheat-based food types.  Moreover it's likely a weighted average, because a simple un-weighted average from the ten wheat-based food items equals 3.52 kcal/gm, and not 2.97 kcal/gm (evidently a heavier weighting to bread or bran of wheat than the others).

My surmise is supported by a document at FAOSTAT that lists various food types that are “aggregated or standardized” a second document that defined the various Food Balance Sheets, and a third document at FAOSTAT that presents various  technical conversion factors, which is done as the latter document explains:

to compile commodity balances and supply/utilization accounts for nearly all the countries in the world. The main scope of these balances and accounts is ultimately to arrive at approximate estimates of the total availability of food in each country, expressed in terms of quantity as well as in terms of calories, protein and fat

To generate the Food Balance Sheets, FAOSTAT has made general estimates of the energy content of a number of aggregated food types. 

However, not every food item listed in the Food Balance Sheets is an aggregate of several different foods.   For instance, bananas, dry beans, honey, plantains, sugar cane, sugar beets, sweet potatoes, yams are among a few dozen single-item foods that are presented in the Food Balance Sheets.  In these cases the conversion factor in the Handbook and the conversion factor derived from Australia 2003 Food Balance Sheet are generally in good agreement.  The only exemption to this is when the food item quantity is so small that there is a rounding error in the conversion because the Food Balance Sheets only presents food items in quantities of 1 thousand tonnes or higher.  It appears that quantities less than this are either not reported or rounded to 1. 

The remaining part of the answer: deriving conversion factors from world food balance sheets

To minimize these rounding errors, and to derive conversion factors for a broader number of weighted and aggregated food items, I decided to look at the Food Balance Sheets for the largest entity reported in FAOSTAT: the world.  Specifically, I looked at world-wide Food Balance Sheets for 1963, 1973, 1983, 1993 and 2003.  Using the human food energy consumption reported for each of these years, I derived conversion factors for most of the food items in the Food Balance Sheets, and then calculated an average.  The procedure I used is exactly the same as described in the last post from Australia 2003.

By considering world-wide values, a number of food items with negligible human food consumption quantities for an individual country, (e.g., like barley and sorghum in Australia), are now in significant quantities, and therefore I could expand the number of derived conversion factors for more food items.  Also taking the average for five different years spread-out by a decade gave me some assurance of the consistence of the derived conversion factor over the 46 year time span that FAOSTAT reports.  

Generally, for food items with larger quantities of human food consumption there was a small variation in the conversion factors within the five years considered (e.g., typically the percent standard deviation < 10%) and no discernable trend for change with time.  For single item food types, the agreement with the value from the Table was generally within a few percent.

Still, there were still about two dozen food items (e.g., sugar beets, sunflower and cotton seed, palm kernels, crustaceans, cephalopods, mollusks, aquatic mammal meat, aquatic plants) whose world-wide human consumption was relatively small, and therefore, either the conversion factors could not be derived from the world-wide human energy consumption, or in my judgment, there were likely to be rounding errors.  For those items, I used the single item value, or simple average of multiple items, as presented in the Table.

Summary of Conversion Factors

Alright, here is my summary of my best estimates of the food energy conversion factor either drived from the five world-wide Food Balance Sheets ("D") or taken from the Annex 1 Table from the Handbook ("T").

Table 1:  Summary of Food Energy Conversion Factors for Food Items Listed in FAOSTAT Food Balance Sheets
(http://crash-watcher.blogspot.com/)
Food Balance Sheet Item
Food Item Code
Best estimate CF
(kcal/gm)
Source
Population



Grand Total + (Total)



Vegetal Products + (Total)



Animal Products + (Total)



Cereals - Excluding Beer + (Total)



Wheat
2511
2.864
D
Rice (Milled Equivalent)
2805
3.676
D
Barley
2513
2.507
D
Maize
2514
3.021
D
Rye
2515
2.655
D
Oats
2516
2.133
D
Millet
2517
2.997
D
Sorghum
2518
3.052
D
Cereals, Other
2520
2.824
D
Starchy Roots + (Total)



Cassava
2532
0.956
D
Potatoes
2531
0.672
D
Sweet Potatoes
2533
0.984
D
Yams
2535
1.000
D
Roots, Other
2534
1.081
D
Sugar crops + (Total)



Sugar Cane
2536
0.306
D
Sugar Beet
2537
0.7
T
Sugar & Sweeteners + (Total)



Sugar, Non-Centrifugal
2541
3.535
D
Sugar (Raw Equivalent)
2542
3.574
D
Sweeteners, Other
2543
3.161
D
Honey
2745
3.622
D
Pulses + (Total)



Beans
2546
3.388
D
Peas
2547
3.352
D
Pulses, Other
2549
3.463
D
Treenuts + (Total)
2912
2.581
D
Oilcrops + (Total)



Soyabeans
2555
3.371
D
Groundnuts (Shelled Eq)
2556
5.469
D
Sunflowerseed
2557
3.08
T
Rape and Mustardseed
2558
4.81
T
Cottonseed
2559
2.53
T
Coconuts - Incl Copra
2560
1.341
D
Sesameseed
2561
7.148
D
Palmkernels
2562
5.14
T
Olives
2563
1.42
T
Oilcrops, Other
2570
3.487
D
Vegetable Oils + (Total)



Soyabean Oil
2571
8.740
D
Groundnut Oil
2572
9.085
D
Sunflowerseed Oil
2573
8.796
D
Rape and Mustard Oil
2574
8.933
D
Cottonseed Oil
2575
8.455
D
Palmkernel Oil
2576
9.649
D
Palm Oil
2577
8.652
D
Coconut Oil
2578
8.599
D
Sesameseed Oil
2579
11.531
D
Olive Oil
2580
8.870
D
Ricebran Oil
2581
8.840
T
Maize Germ Oil
2582
9.319
D
Oilcrops Oil, Other
2586
9.050
D
Vegetables + (Total)



Tomatoes
2601
0.185
D
Onions
2602
0.364
D
Vegetables, Other
2605
0.242
D
Fruits - Excluding Wine + (Total)



Oranges, Mandarines
2611
0.282
D
Lemons, Limes
2612
0.266
D
Grapefruit
2613
0.67
T
Citrus, Other
2614
0.404
D
Bananas
2615
0.621
D
Plantains
2616
0.868
D
Apples
2617
0.445
D
Pineapples
2618
0.355
D
Dates
2619
1.542
D
Grapes
2620
0.649
D
Fruits, Other
2625
0.438
D
Stimulants + (Total)



Coffee
2630
0.324
D
Cocoa Beans
2633
2.078
D
Tea
2635
0.345
T
Spices + (Total)



Pepper
2640
2.76
T
Pimento
2641
3.453
D
Cloves
2642
3.23
T
Spices, Other
2645
3.303
D
Alcoholic Beverages + (Total)



Wine
2655
0.694
D
Beer
2656
0.448
D
Beverages, Fermented
2657
0.516
D
Beverages, Alcoholic
2658
2.910
D
Alcohol, Non-Food
2659
0.6
T
Meat + (Total)



Bovine Meat
2731
1.525
D
Mutton & Goat Meat
2732
2.019
D
Pigmeat
2733
2.469
D
Poultry Meat
2734
1.416
D
Meat, Other
2735
1.258
D
Offals + (Total)
2945
1.103
D
Animal Fats + (Total)



Butter, Ghee
2740
7.432
D
Cream
2743
1.939
D
Fats, Animals, Raw
2737
7.170
D
Fish, Body Oil
2781
9.02
T
Fish, Liver Oil
2782
9.02
T
Eggs + (Total)
2949
1.425
D
Milk - Excluding Butter + (Total)
2948
0.558
D
Fish, Seafood + (Total)



Freshwater Fish
2761
0.727
D
Demersal Fish
2762
0.598
D
Pelagic Fish
2763
0.975
D
Marine Fish, Other
2764
0.706
D
Crustaceans
2765
0.996
T
Cephalopods
2766
1.496
T
Molluscs, Other
2767
1.322
T
Aquatic Products, Other + (Total)



Meat, Aquatic Mammals
2768
1.46
T
Aquatic Animals, Others
2769
0.77
T
Aquatic Plants
2775
1.937
T
Miscellaneous + (Total)
2928



Not very scintillating reading, I know, but a fair amount of work to come up with this.

Some Preliminary Testing

Now that I have a decent set of conversion factors, let's go back to my example of the 2003 Food Balance Sheet for Australia and see what kind of food energy totals are obtained for the main headings listed in the sheet.

To review, for each food item listed in the Food Balance Sheet, I multiplied the quantity by the conversion factor in Table 1 above, made the appropriate conversion to Peta Joules, and then added up all the items to derive a grand total food energy estimate for each of the headings: production, imports, exports, feed, seed etc....  I also decided to break the grand total down into subtotals of plant and animal food energy.

Here's a summary of my calculations as applied to the 2003 Food Balance Sheet for Australia:


 My estimate of human food energy being equal to 101 PJ is fairly close to the estimate of 95 PJ that I made in the previous post, which in turn, was based on the food energy per capita values reported by FAOSTAT in the 2003 Food Balance Sheet for Australia. 

In terms of food energy, the total production of 790 Peta Joules per year comprises about 90% plant product and 10% animal product food energy, although the food energy embodied in animal Feed (125 PJ) is about 16% of the total food energy produced.  Clearly, animal product production is energy expensive—the production of 77 PJ worth of animal food energy takes 125 PJ worth of plant energy just in the form of feed.  It is interesting that the total food energy for humans (101 PJ) is actually less that the total food energy for feed of 125 PJ.  Also interesting is that of the total food energy for humans, roughly 1/3 comes from animal products (34 PJ).   Australia, in 2003, was clearly a net exporter (i.e., 270 PJ of food energy, exporting 270 PJ of food energy versus importing only 32 PJ of food energy—the food exports amount to about 1/3 of the total food energy produced.

Finally, how do these numbers compare to that Table 1 of in Arizpe?  

Unfortunately, Arizpe doesn't define exactly what the terms "Gross Food Production" and "Gross Food Production" as used in Table 1 actually mean, or how they were calculated, so a comparison isn't that easy to make.  I asked the contact author, Mario Giampietro, if he could define these terms but his response didn't clarify this. In a reply email, Professor Giampietro told me that his student made the calculation so he wasn't certain, but he thought "gross production" was referring to the energy of the plant production used both as feed and used as food. 

Well, in my opinion, total human food energy plus total feed energy would seem to be a better of a measure of food consumption than food production. 

Indeed, the sum of human plant food (67 PJ) plus plant feed (121 PJ), equal to 188 PJ, is much closer to "Gross Food Consumption" than "Gross Food Production" in Table 1 of Arizpe (615 PJ).   Similarly, the sum of total human food energy (101 PJ) and total feed energy (124 PJ) equals 225 PJ is a bit higher than Arizpe's 210 PJ, and, Arizpe's "Gross Food Production" of 615 PJ is much less that the either the Grand total of food energy (790 PJ) or plant food energy (713 PJ) in my Table 2. 

After comparing a few other countries in the Arizpe's Table 1 (Argentina and Spain) and getting about the same discrepancies, and, getting no response to a second inquiry email, I gave up trying compare my calculations Arizpe's and decided to move on.

Of course, I think that my numbers are correct—and I could be just comparing apples-to-oranges since Arizpe didn't define his terms or describe how he made his food energy energy estimates in Table 1.  In any event, I hope that I have provided enough details that, if it is my error, then it should be possible for someone to detect it.   At least my numbers should be internally consistent, since I will be applying the same conversion factors throughout my analysis in subsequent posts.
-----------------------------

Okay, that is it for the preliminary work, it is time to move on and look at some country or region data—but, where to start? 

Let's start with the world.