Tittonell & Giller (2013). When yield gaps are poverty traps: The paradigm of ecological intensification in African smallholder agriculture

http://dx.doi.org/10.1016/j.fcr.2012.10.007

to

deﬁne

the

set

of

principles

and

means

necessary

to

increase

primary

productivity

in

the

major

cereal

agroecosystems

of

the

world.

Emphasis

was

placed

on

increasing

the

yield

ability

of

major

crops

and

narrowing

yield

gaps

through

implementing

forms

of

precision

agriculture,

relying

on

scientiﬁc

breakthroughs

in

the

ﬁeld

of

plant

physiology,

crop

ecophysiology

and

soil

science.

Later

reinterpretations

of

this

concept,

particularly

in

the

con-

text

of

European

agriculture

(e.g.,

attached

a

deﬁnition

that

borders

those

of

organic

or

ecological

farming.

Ecological

intensiﬁcation

is

now

understood

as

a

means

of

increasing

agricultural

outputs

(food,

ﬁbre,

agro-fuels

and

environ-

mental

services)

while

reducing

the

use

and

the

need

for

external

inputs

(agrochemicals,

fuel,

and

plastic),

capitalising

on

ecological

processes

that

support

and

regulate

primary

productivity

in

agroe-

cosystems.

Yet,

little

has

been

written

on

how

to

achieve

this.

Single

∗

Corresponding

author.

E-mail

address:

pablo.tittonell@wur.nl

(P.

Tittonell).

efforts

addressing

the

various

challenges

facing

current

agriculture

have

been

often

done

in

isolation

rather

than

holistically.

These

observations

have

prompted

propose

new

sources

of

knowledge

and

methods

in

agronomy

to

strengthen

the

ecological

intensiﬁcation

of

current

agriculture.

The

ecological

intensiﬁcation

of

agriculture

has

seldom

been

addressed

in

the

context

of

the

smallholder

farming

systems

that

characterise

rural

Africa.

There

is

no

doubt

of

the

concept’s

relevance

to

guide

farming

systems

design

in

the

African

con-

text:

producing

more

with

less

external

input,

while

keeping

a

healthy

environment

that

provides

multiple

services.

However,

in

view

of

the

importance

of

agriculture

for

rural

livelihoods

and

national

economies

the

‘intensiﬁcation’

component,

whether

eco-

logical

or

not

has

been

sensed

as

most

urgent

for

Africa.

The

need

to

intensify

African

agriculture

has

recently

led

agricultural

research

for

development

in

a

somehow

opposite

direction,

pro-

moting

the

use

of

mineral

fertilisers,

hybrid

seeds,

new

crops,

irrigation,

herbicide-based

no-till

systems,

genetically

modiﬁed

cultivars

or

mechanisation

as

means

to

increase

productivity

(e.g.,

http://www.agra.org

).

The

sustainable

intensiﬁcation

of

agricul-

ture

through

technologies

that

rely

on

substantial

investment

in

inputs

has

been

seriously

hampered

by

poorly

developed

input

and

0378-4290 ©

2012

Elsevier

B.V.

Open access under

CC BY-NC-ND license.

Open access under

CC BY-NC-ND license.

P.

Tittonell,

K.E.

Giller

/

Field

Crops

Research

143

(2013)

76–90

77

output

markets

but

often

also

by

the

poor

performance

of

technologies

in

the

African

context

or

their

inade-

quacy

to

ﬁt

within

local

smallholder

systems

(e.g.,

Thus,

whether

deliberately

or

not,

much

of

African

agriculture

has

remained

rather

‘ecological’.

Basic

supportive

and

regulatory

ecological

processes

steered

through

local

lay

knowledge

still

con-

stitute

the

backbone

of

smallholder

agriculture

in

many

places.

But

despite

the

genuine

attractiveness

that

surrounds

traditional

systems

and

local

practices,

their

inability

to

sustain

Africa’s

food

sufﬁciency

is

self-evident.

Yet,

there

are

reasons

to

believe

that

supportive

and

regulatory

ecological

mechanisms

that

farmers

are

able

to

master

can

lead

to

synergetic

responses

of

agricultural

sys-

tems

to

external

inputs

(e.g.,

This

forms

the

basis

for

the

design

of

integrated

approaches

to

soil

fertility,

pest

or

crop

management

that

build

on

local

knowledge.

A

large

body

of

evidence

shows

that

labour,

water

or

nutrient

use

efﬁciencies

are

enhanced

through

the

implementation

of

such

knowledge-

intensive

approaches

(e.g.,

between

strategies

necessary

for

ecological

intensiﬁcation

under

unfavourable

(mostly

rain-

fed)

and

favourable

environments,

focusing

mostly

on

the

latter.

He

postulated

that

the

ecological

intensiﬁcation

of

agriculture

in

unfavourable

rain-fed

environments,

where

lack

of

water

would

be

the

primary

constraint,

would

depend

on

reducing

the

reliance

on

subsistence

cereal

production,

integration

with

livestock

enter-

prises,

greater

crop

diversiﬁcation

and

agroforestry

practices

that

may

ensure

higher

economic

value

and

soil

conservation.

Although

some

of

these

principles

may

be

relevant

for

the

rain-fed

pro-

duction

environments

that

predominate

in

sub-Saharan

Africa,

closer

examination

is

necessary.

For

example,

crop

productivity

in

many

parts

of

Africa

is

limited

primarily

by

nutrient

rather

than

water

availability;

smallholder

cereal

production

is

often

oriented

to

both

consumption

and

the

market;

the

integration

of

cropping

and

livestock

activities

is

already

a

common

denominator

to

many

of

these

systems;

in

densely

populated

regions

green

manures

or

agroforestry

do

not

always

ﬁt

the

needs

and

possibilities

of

smallholders;

and

so

on.

Where

natural

resources

have

become

degraded,

farmers

may

be

caught

in

poverty

traps

where

response

to

inputs

is

poor

and

follows

an

‘S-shaped’

curve

(

Africa

needs

a

‘uniquely

African’

strategy

for

the

sustainable

intensiﬁcation

of

its

agricul-

ture

(cf.

capitalising

on

ecological

processes

and

ensuring

efﬁcient

use

of

scarce

external

inputs.

This

paper

examines

current

yield

gaps

in

Africa

and

the

opportunities

and

challenges

that

lay

ahead

for

the

ecological

intensiﬁcation

of

smallholder

agriculture,

placing

emphasis

on

the

efﬁcient

use

of

the

abiotic

resources:

light,

water

and

nutrients.

We

draw

on

a

wide

range

of

experiences

from

Southern,

East,

Central

and

West

Africa.

We

postulate

that

current

approaches

to

the

eco-

logical

intensiﬁcation

of

smallholder

agriculture

in

Africa:

(i)

may

be

deterred

by

inherent

characteristics

of

these

agroecosystems,

(ii)

lack

biophysical

references

and

suitable

technical

means

that

embrace

local

preferences

and

knowledge,

(iii)

should

not

overlook

the

integrated

nature

of

smallholder

systems

(e.g.,

crop-livestock

interactions,

communally

owned

resources,

etc.)

in

which

decisions

are

made

at

scales

higher

than

the

ﬁeld

plot,

and

(iv)

will

not

be

achieved

without

prior

efforts

to

restore

productivity

of

already

degraded

land.

The

three

pillars

that

http://www.earthtrends.wri.org

identiﬁed

for

the

ecological

intensiﬁcation

of

cereal

production,

namely

yield

potential,

soil

quality

and

precision

agriculture

are

ﬁrst

analysed

for

their

speciﬁcities

and

suitability

in

the

African

context.

This

is

followed

by

an

estimate

of

the

average

yield

gap

of

major

crops

in

the

continent,

attempting

to

distinguish

between

yield

gaps

caused

by

resource

availability

from

gaps

caused

by

access

to

technology.

Fig.

1.

Average

cereal

productivity

and

fertiliser

use

intensity

(total

fertiliser

use

over

area

cropped)

at

national

level

for

selected

countries

between

1961

and

2008.

Source

:

FAOstat.

We

conclude

by

proposing

a

framework

for

yield

gap

assessment

in

African

agriculture

based

on

biophysical

drivers,

with

the

aim

of

contributing

to

the

delineation

of

a

worldwide

yield

gap

atlas

2.

The

pillars

of

ecological

intensiﬁcation

following

Cassman

(1999)

2.1.

Yield

potential

Food

production

in

sub-Saharan

Africa

is

not

keeping

pace

with

population

growth.

Sub-Saharan

Africa

has

the

lowest

land

and

labour

productivity

rates

in

the

world,

with

annual

growth

in

cereal

yields

averaging

only

10

kg

grain

ha

−

1

yr

−

1

–

about

1%

(

While

cereal

yields

in

most

of

the

developed

and

developing

world

increased

steadily

during

the

last

50

years,

yields

in

African

countries

hovered

around

1

t

ha

−

1

or

less

Similarly,

the

average

yield

of

tuber

crops

(cas-

sava,

sweet

potato,

yam,

etc.)

is

the

lowest

in

the

world

(around

8

t

ha

−

1

),

increasing

at

a

rate

of

50

kg

ha

−

1

yr

−

1

or

0.6%

over

the

same

period.

Counting

growth

in

harvested

area

as

well,

food

pro-

duction

in

sub-Saharan

Africa

increases

at

an

annual

rate

of

ca.

2%,

while

population

growth

rates

average

3%.

If

Africa

seeks

to

rely

on

agriculture

for

economic

development,

an

annual

increase

of

4–7%

in

food

production

is

required

Techno-

logical

progress

in

tropical

agriculture

in

combination

with

more

favourable

socio-economic

contexts

allowed

food

production,

and

particularly

cereal

yields

to

increase

substantially

in

Latin

Amer-

ica

and

Asia

during

the

last

two

to

three

decades.

Although

much

of

such

yield

increase

may

be

explained

by

increased

input

use

(

1

B),

genetic

progress

through

plant

breeding

played

a

central

78

P.

Tittonell,

K.E.

Giller

/

Field

Crops

Research

143

(2013)

76–90

Fig.

2.

A

representation

of

yield-deﬁning,

yield-limiting

and

yield-reducing

factors

determining

respectively

the

potential,

attainable

and

actual

yield

levels.

Factors

were

grouped

into

varietal

characteristics

(G),

uncontrolled

environment

(E)

and

management

(M)

to

illustrate

the

increasing

importance

of

the

latter

as

one

moves

from

potential

to

actual

yields.

From

:

Tittonell

(2009)

.

role

(Hall,

this

volume).

As

the

yield

ability

of

major

crop

varieties

available

for

tropical

environments

keeps

rising,

the

stagnating

average

yield

observed

in

most

African

countries

means

that

yield

gaps

are

widening

across

the

continent.

(Theoretical)

potential

yields

in

sub-Saharan

Africa,

those

that

may

be

achieved

under

no

water

or

nutrient

constraints,

vary

for

the

various

cereal

crops

and

their

cultivars

and

are

deﬁned

chieﬂy

by

latitude,

altitude

and

cloud

cover.

Potential

yields

can

be

calcu-

lated

using

simulation

models

that

input

the

length

of

the

growing

season,

diurnal

and

nocturnal

air

temperatures,

day

length

and

the

total

amount

of

solar

radiation

received

during

a

season

by

a

given

cultivar

(e.g.,

In

this

paper

we

refer

to

the

water-limited

yield

potential

(

Y

w

)

as

deﬁned

by

is

the

maximum

yield

achievable

under

rain-

fed

conditions

without

irrigation

if

soil

water

capture

and

storage

is

optimal

and

nutrient

constraints

are

eliminated.

The

attainable

yield

level

following

the

classical

deﬁnition

of

production

situa-

tions

by

to

water

and

nutrient-limited

yields.

Here,

we

propose

an

adaptation

of

this

to

the

African

situ-

ation

by

deﬁning

a

locally

attainable

yield

level

(

Y

L

),

which

is

the

maximum

yield

achievable

by

resource

endowed

farmers

in

their

most

productive

ﬁelds.

Two

yield

gaps

can

be

then

calculated:

a

yield

gap

1,

between

Y

W

and

Y

L

,

and

a

yield

gap

2,

between

Y

L

and

actual

farmers

yield

levels

(

Y

A

).

2.1.1.

The

(G

×

E

×

M)

FS

interaction

Taking

de

Wit’s

deﬁnitions,

the

relative

importance

of

manage-

ment

decisions

with

respect

to

the

genotype

or

the

uncontrolled

environment

increases

as

we

move

from

yield-deﬁning

to

yield-

reducing

factors,

from

potential

to

attainable

and

actual

yields

There

is

consensus

that

nutrient

supply,

rather

than

water,

is

the

main

yield-limiting

factor

in

sub-Saharan

Africa

However,

whilst

fertiliser

use

has

expanded

in

some

countries

such

as

Kenya,

it

remains

anecdotal

in

many

parts

of

the

continent.

Where

land

is

not

limiting,

the

area

cultivated

is

a

more

important

determinant

of

household

food

security

than

the

yield

per

unit

area.

In

such

cases,

farmers

often

prioritise

investments

to

hire

labour

or

ox-

ploughing

rather

than

purchasing

fertilisers

or

improved

seeds

to

intensify

production

(

Cultivating

large

areas

with

limited

labour

available

often

leads

to

late

planting,

exposing

bare

soil

to

the

ﬁrst

torrential

rains

of

the

season,

or

to

late

or

inefﬁ-

cient

weeding

during

the

season.

In

some

cases,

however,

extended

periods

of

sowing

may

be

a

strategy

to

deal

with

erratic

rainfall,

minimising

the

risk

of

complete

crop

failure

in

space

and

time

Thus

probably

more

than

anywhere

else,

potential

yields

under

African

smallholder

conditions

are

the

result

of

a

tight

interaction

between

the

genotype,

the

environment

and

the

local

farming

practice

(i.e.,

FS

in

the

G

×

E

×

M

model).

2.1.2.

Local

reference

yields

Although

it

may

be

hard

to

estimate

the

achievable

yields

of

local

varieties

under

local

circumstances

and

management

prac-

tices,

approximations

are

necessary

in

order

to

quantify

current

yield

gaps.

An

example

is

the

use

of

boundary-line

analysis

(e.g.,

of

large

yield

datasets

across

sites,

seasons

and

management

practices.

Boundary

lines

may

reveal

ceil-

ing

yields

for

a

given

crop

in

a

certain

environment.

Ceiling

yields

may

sometimes

represent

the

maximum

achievable

yields

under

farmer

management,

or

locally

attainable

yield

Y

L

,

or

be

close

to

the

water-limited

yield

potential

Y

W

when

yields

are

measured

under

controlled

conditions.

An

alternative

is

to

use

simulation

models

to

establish

the

reference

yield

Y

W

,

or

an

approxima-

tion

to

the

locally

attainable

(water

and

nutrient-limited)

yield

Y

L

when

proper

model

calibration

and

validation

are

possible

(e.g.,

Such

an

exercise

is

illustrated

in

A

for

P.

Tittonell,

K.E.

Giller

/

Field

Crops

Research

143

(2013)

76–90

79

Fig.

3.

(A)

On-farm

maize

biomass

yields

in

farmers’

ﬁelds

(current

practice)

and

in

researcher-managed

micro-plots

established

on

the

same

ﬁelds

with

or

without

mineral

fertilisers

(NPK,

control),

and

yields

simulated

for

each

ﬁeld

with

the

model

DYNBAL

(B)

means

and

standard

deviations

in

each

case

(ﬁrst

column

from

the

left

is

average

simulated

yields);

(C)

comparing

average

grain

yields

on-farm

(current

practice

and

researcher-managed)

against

yields

obtained

in

nearby

experimental

stations

(control,

NPK),

with

the

observations

grouped

accord-

ing

to

proximity

to

two

stations

located

in

lower

and

higher

potential

agro-ecological

zones

(AEZ)

in

western

Kenya.

maize

(

Zea

mays

L.)

grown

in

the

highlands

of

Kenya,

using

above-

ground

biomass

as

a

measure

of

productivity

(harvest

indices

were

highly

variable

under

farmer

management).

Simulated

water

and

nutrient-limited

biomass

yields

(corresponding

to

Y

L

)

are

plotted

against

soil

organic

carbon

together

with

biomass

yields

measured

on

farmers’

ﬁelds

under

their

own

management

(current

practice)

and

under

researcher

management

with

(NPK)

or

without

(control)

fertilisers.

Soil

C

(0–20

cm)

is

used

here

as

a

surrogate

of

inherent

and

current

soil

fertility,

with

soil

clay

contents

varying

from

18

to

44%,

altitude

from

1100

to

2100

masl

and

rainfall

from

1300

to

1900

mm

in

the

study

area.

Across

this

environmental

range,

the

simulated

attainable

biomass

yields

varied

from

ca.

5

to

20

t

ha

−

1

(15.5

on

average),

and

the

respective

gaps

between

2

and

15

t

ha

−

1

under

farmer

prac-

tices,

between

0

and

14

in

control

plots,

and

between

0

and

8

t

ha

−

1

with

a

full

NPK

fertiliser

application.

The

average

response

to

a

full

NPK

fertilisation

was

in

the

order

of

5.3

t

ha

−

1

of

aboveground

biomass.

It

is

striking

that

researcher-managed

plots

receiving

no

fertilisers

yielded

on

average

better

than

the

same

ﬁelds

under

farmer

management,

which

may

include

using

organic

and/or

mineral

fertilisers.

The

effect

of

germplasm

was

more

difﬁcult

to

unravel,

given

that

the

hybrid

used

under

research

manage-

ment

was

also

often

seen

in

farmers’

ﬁelds.

When

examining

grain

yields

(

C),

such

differences

were

wider

for

the

ﬁelds

grouped

in

the

lower

potential

zone

(dominated

by

ferric-Acrisols,

1300–1500

mm)

than

in

the

higher

potential

zone

(Nitosols

and

humic-Ferralsols,

1500–1900

mm).

Yields

under

controlled

man-

agement

with

fertiliser

were

greater

on

nearby

experimental

stations

than

in

farmer’s

ﬁelds

(although

not

during

the

same

sea-

son),

and

such

difference

was

also

wider

in

the

high

potential

zone.

The

potential

yields

simulated

with

the

soil-crop

model

DYNBAL

(

in

response

to

water

and

nutrient

availability

varied

between

10.8

and

11.4

t

ha

−

1

grain

(22.6

and

24.5

t

ha

−

1

aboveground

biomass)

for

this

environmental

range

(

),

cutting

through

one

of

the

areas

of

highest

agricultural

potential

in

sub-Saharan

Africa.

2.1.3.

Non-cereal

crops

Although

the

analysis

of

focuses

on

the

yield

potential

of

major

cereal

crops,

the

economy

of

large

areas

of

sub-

Saharan

Africa

depends

also

on

other,

equally

important

staple

or

cash

crops.

Yield

gaps

for

cassava

(

Manihot

esculenta

Crantz)

grown

in

vast

areas

of

medium

to

marginal

agro-ecological

potential

may

also

be

analysed

with

respect

to

management

practices

or

single

limiting

factors,

and

reference

ceiling

(rather

than

potential)

yields

be

derived

through

boundary

line

analysis

(

4

A).

Maximum

fresh

root

yields

obtained

on

eastern

Uganda

farms

were

in

the

order

of

25

t

ha

−

1

,

while

yields

of

50–60

t

ha

−

1

have

been

obtained

under

experimental

conditions

in

East

Africa

and

as

high

as

75–90

t

ha

−

1

in

Colombia

or

India

(

the

yield

gap

of

cassava

in

the

East

African

highlands

and

determined

the

individual

yield

gains

and

synergies

that

may

be

expected

from

improved

agronomy,

cultivar

choice

or

fertiliser

use.

Although

it

is

generally

believed

that

cassava

responds

poorly

to

fertilisers,

this

research

showed

once

again

that

responses

were

substantial

when

proper

agronomic

(establishment

and

weeding)

practices

were

in

place.

The

incre-

mental

contribution

of

these

different

factors

to

narrowing

the

yield

gap

on

farmers’

ﬁelds

was

1.5

t

ha

−

1

for

improved

crop

implanta-

tion

(from

an

average

yield

of

fresh

cassava

roots

8.6

to

10.1

t

ha

−

1

),

3.5

t

ha

−

1

for

improved

cultivar

choice

(from

10.1

to

13.6

t

ha

−

1

)

and

7.2

t

ha

−

1

with

fertiliser

use

(from

13.6

to

20.8

t

ha

−

1

).

The

choice

of

cultivars

with

resistance

to

cassava

mosaic

virus

was

crucial.

For

an

indeterminate

crop

such

as

cotton

(

Gossypium

hirsutum

L.),

the

engine

of

rural

economies

across

regions

of

West,

Central

and

Southern

Africa,

the

relationships

between

single

limiting

fac-

tors

and

yield

are

more

elusive.

Examining

data

from

a

number

of

research

trials

established

across

the

cotton

growing

area

of

south-

ern

Mali,

Cretenet

(1994)

arrived

at

establishing

yield

thresholds

with

respect

to

soil

indicators

such

as

exchangeable

K

(

4

B)

or

organic

C

contents,

to

rainfall

or

to

sowing

dates

using

a

similar

boundary-line

approach.

Cotton

has

a

relatively

strong

K

demand.

A

negative

K

balance

of

200

kg

ha

−

1

over

a

certain

period

of

time,

which

corresponds

to

a

change

in

0.025

cmol

(+)

kg

−

1

of

K

in

the

ﬁrst

40

cm

of

the

soil,

results

in

a

reduction

of

up

to

570

kg

ha

−

1

in

the

attainable

seed-cotton

yield.

Each

day

of

delay

in

the

date

of

plant-

ing,

or

each

day

of

reduction

in

the

rainy

period

led

to

an

average

reduction

of

16

kg

ha

−

1

in

the

attainable

seed

cotton

yield,

irrespec-

tive

of

fertiliser

use.

More

difﬁcult

to

predict

in

the

case

of

cotton

is

the

potential

quality

of

the

ﬁbre

obtained,

which

has

a

strong

impact

on

the

price

received

by

the

farmer.

The

examples

above

illustrate

the

use

of

boundary

line

analy-

sis

to

study

the

effect

of

single

abiotic,

yield-limiting

factors

such

as

water,

nutrients

or

dates

of

planting.

Similarly,

the

method

can

be

used

to

study

the

effect

of

biotic

yield-reducing

factors

such

as

weeds,

pest

and

diseases.

The

East

African

highland

banana

(

Musa

spp.,

AAA-EA

genome)

is

a

staple

crop

in

large

areas

of

Burundi,

the

Democratic

Republic

of

Congo,

Rwanda

and

Uganda.

In

a

detailed

investigation

on

plots

on

159

smallholder

farms

in

Uganda,

yields

of

highland

banana

were

signiﬁcantly

greater

80

P.

Tittonell,

K.E.

Giller

/

Field

Crops

Research

143

(2013)

76–90

0

5

10

15

20

25

30

0.0

0.3

0.6

0.9

1.2

1.5

1.8

2.1

C

a

ssava r

oot

yi

el

d (

t ha

-1

)

On-farm observations

Boundary-line model

0.0

0.6

1.2

1.8

2.4

3.0

3.6

0.0

0.1

0.2

0.3

0.4

0.5

S

e

ed-

cot

ton yi

el

d (

t ha

-1

)

On-station observations

Boundary-line model

Source: Crétenet (1994)

A

B

Banana bunch yield (t ha

-1

)

Nematode root necrosis (%)

Weevils corm damage (%)

Weed pressure index (%)

C D E

Source: Fermont et al. (2009)

Total soil N (g kg

-1

)

Exchangeable K (cmol

(+)

kg

-1

)

40

20

0

40

20

0

40

20

0

Source: Wairegi et al. (2010)

0

2.5

5

7.5

10

0

15

30

45

0

5

10

15

20

Fig.

4.

Examples

of

using

boundary

line

analysis

to

establish

ceiling,

or

locally

attainable

yields

as

a

function

of

different

explanatory

variables.

(A)

Cassava

yield

in

Uganda

and

Kenya

against

total

soil

N;

(B)

Cotton

yield

in

Mali

against

soil

exchangeable

K;

(C–E)

Highland

banana

yield

against

different

biotic

constraints.

Boundary

line

models

ﬁtted

to

the

maxima

at

each

level

of

the

independent

variable

are

represented

by

(continuous

or

dashed)

lines.

in

the

Southwest

region

(20

t

ha

−

1

year

−

1

)

than

in

the

Central

(12

t

ha

−

1

year

−

1

)

and

South

(10

t

ha

−

1

year

−

1

)

(

whereas

the

estimated

national

average

yield

in

2007

was

only

5.5

t

ha

−

1

year

−

1

(FAOSTAT).

On

the

same

farms

demonstration

plots

yielded

3

to

10

t

ha

−

1

year

−

1

more

when

(on

average)

71

N,

8

P

and

32

K

kg

ha

−

1

year

−

1

was

applied

(

Using

the

boundary

line

approach

(

4

C–E)

they

found

that

the

yield

gap

expressed

as

percentage

of

attainable

yield

in

the

Central

region

was

caused

by

pests

(nematodes

10%

loss,

weevils

6%)

and

subop-

timal

crop

management

(mulch

25%)

indicating

that

past

research

efforts

were

mistaken

in

neglecting

abiotic

constraints.

2.2.

Soil

quality

The

second

pillar

of

ecological

intensiﬁcation

proposed

by

was

the

maintenance

or

improvement

of

soil

quality,

deﬁned

as

the

capacity

of

soils

to

sustain

biological

pro-

ductivity

while

ensuring

environmental,

plant

and

animal

health.

Both

severe

and

subtle

forms

of

soil

degradation

are

associated

worldwide

with

the

practice

of

agriculture,

even

under

favourable

production

environments.

Four

soil

degradation

phenomena

may

be

distinguished:

water

erosion,

wind

erosion,

deterioration

of

physical

properties,

and

chemical

degradation.

The

latter

includes

nutrient

depletion

and

loss

of

organic

matter,

salinisation,

acidiﬁca-

tion,

and

chemical

pollution.

The

relationship

between

poverty

and

land

degradation

has

been

highlighted

for

sub-Saharan

Africa

(e.g.,

and

many

of

the

forms

of

degrada-

tion

listed

take

place

simultaneously

in

the

continent.

Rather

than

cataloguing

the

extent

of

land

degradation,

as

recently

done

(cf.

we

focus

on

the

particularities

of

the

degradation

processes

taking

place

in

sub-Saharan

Africa,

and

on

their

implica-

tions

for

the

design

of

strategies

to

rehabilitate

degraded

land.

Soil

nutrient

depletion

and

loss

of

organic

matter

are

treated

as

inte-

grative

measures

of

land

degradation,

which

reﬂect

the

combined

effect

of

management

practices,

inherent

soil

fertility

and

other

forms

of

degradation.

Two

decades

ago

alarming

ﬁgures

on

negative

nutrient

balances

at

country

level

for

sub-Saharan

Africa.

Since

then,

nutrient

balances

calculated

in

dif-

ferent

ways

have

been

used

extensively

as

indicators

of

soil

nutrient

depletion

and

of

the

long-term

sustainability

of

agricultural

sys-

tems

at

scales

ranging

from

the

individual

ﬁeld

plot

to

entire

regions

or

countries

(e.g.,

Nutrient

balances

calculated

at

regional

and/or

national

scales

provide

coarse

but

relevant

information

for

policy

mak-

ers.

For

example,

it

was

estimated

that,

every

day,

up

to

100

trucks

with

a

payload

of

6

tonnes

of

cooking

bananas

enter

the

capital

city

of

Uganda,

Kampala,

representing

an

annual

export

from

rural

areas

of

over

1.5

million

kg

K

and

0.5

million

kg

N

(

However,

less

than

5%

of

banana

farmers

in

Uganda

use

any

type

of

mineral

fertiliser

(

Estimations

of

nutrient

balances

at

ﬁeld

scale,

by

differ-

ent

authors

and

through

slightly

different

methods

show

almost

always

negative

values

in

different

African

farming

systems.

For

continuous

cereal

cropping

in

the

central

highlands

of

Kenya,

nitrogen

balances

as

negative

as

−

44

to

−

75

kg

N

ha

−

1

year

−

1

,

clearly

contrasting

with

the

values

pre-

sented

for

other

African

systems

that

were

calculated

using

the

same

method

(cf.

Table

1

One

of

the

weakest

points

in

the

calcula-

tion

of

nutrient

balances

is

the

estimation

of

ﬂows

that

are

difﬁcult

P.

Tittonell,

K.E.

Giller

/

Field

Crops

Research

143

(2013)

76–90

81

Table

1

N

balances

(kg

ha

−

1

season

−

1

)

at

ﬁeld

scale

across

African

farming

systems

calculated

using

comparable

methods

and

assumptions.

a

Inputs

and

outputs

consider

only

those

mediated

by

farmers

(e.g.,

fertilisers,

harvest

of

crop

residue,

etc).

Case

study

Farming

system

Variability

component

Total

N

inputs

N

removal

in

harvest

N

balance

Calculation

procedure

Source

Central

Zimbabwe

Integrated

cereal-livestock

systems,

free

grazing

Best

plots

26

24

−

38

Complete

balance

using

NUTMON

Average

plots

22

18

−

34

Worst

plots

5

11

−

31

Western

Kenya

Integrated

cereal-livestock

systems,

zero

grazing

Home

gardens

28

+21

Complete

balance

using

dynamic

simulation

Close

ﬁelds

43

36

−

22

Mid-distance

ﬁelds

11

25

−

17

Remote

ﬁelds

4

12

−

24

Eastern

Uganda

Cereal-based

farming

systems

(Averaged

for

a

representative

farm)

86

138

−

48

Complete

balance

using

static

model

Northwest

Tanzania

Banana-based

farming

systems

Kibanja

(banana)

18

26

−

8

Partial

balance

using

NUTMON

Kikamba-maize

4

13

−

9

Kikamba-S.

potato

1

6

−

5

Kikamba-Cassava

0

2

−

2

Northern

Ghana

Cereal-based

farming

systems

Mucuna/maize

72.4

+26

Complete

balance

using

a

modiﬁed

NUTMON

Sole

maize

(burning)

0

118

−

120

Southern

Mali

Cereal/cotton/pastoral

systems

Village

settlings

58

45

−

15

Complete

balance

using

NUTMON

Ramisch,

2005

Hamlet

settlings

81

47

−

3

Fulawere

settlings

128

41

+21

a

When

results

were

presented

for

farms

of

different

wealth

classes,

only

the

middle

class

farms

were

considered.

82

P.

Tittonell,

K.E.

Giller

/

Field

Crops

Research

143

(2013)

76–90

to

measure,

such

as

losses

by

leaching

or

erosion,

or

the

ﬂows

gener-

ated

by

denitriﬁcation,

wet/dry

deposition

and

N

2

-ﬁxation

(

For

instance,

N

losses

by

leaching

assumed

by

different

authors

in

African

conditions

were

widely

variable:

8–15

kg

N

ha

−

1

year

−

1

10

kg

N

ha

−

1

year

−

1

11–26

kg

ha

−

1

year

−

1

(

Ramisch,

2005

)

or

36–153

kg

N

ha

−

1

year

−

1

Estimations

remain

uncertain,

and

may

often

lack

quantitative

rigour

(

Nonetheless,

the

work

of

highly

inﬂuential

in

bringing

attention

to

the

problem

and

in

prioritising

research

agendas

on

soil

fertility

management.

Fortunately

the

nutrient

balances

they

predicted

were

not

always

realised,

for

if

they

were,

agriculture

would

by

now

have

disap-

peared

from

one

third

of

the

continent.

Soil

fertility

zoning

is

a

well-known

process

in

the

savannah-

derived

agroecosystems

of

West

Africa,

in

which

nutrients

tend

to

be

concentrated

in

the

village

ﬁelds

to

the

detriment

of

the

fertility

of

the

so-called

bush

ﬁelds

(e.g.,

a

similar

relationship

between

the

magnitude

and

sign

of

nutrient

balances,

as

determined

by

management

decisions,

and

the

creation

of

patterns

of

spatial

soil

heterogeneity

within

indi-

vidual

farms

in

the

highly

fragmented

landscapes

of

western

Kenya.

Large

differences

in

input

use

(e.g.,

0.7–104

kg

N

ha

−

1

year

−

1

),

food

production

(e.g.,

0.6–2.9

t

DM

ha

−

1

year

−

1

),

partial

C

(e.g.,

−

570

to

1480

kg

ha

−

1

year

−

1

)

and

N

(e.g.,

−

92

to

57

ha

−

1

year

−

1

)

bal-

ances

were

observed

between

home-

and

outﬁelds

located

less

than

50–100

m

apart,

which

showed

also

wide

differences

in

extractable

P

(e.g.,

2.1–19.8

mg

kg

−

1

)

and

exchangeable

K

(e.g.,

0.14–0.54

cmol

(+)

kg

−

1

)

contents

in

their

soils.

Differential

manage-

ment

of

the

various

ﬁelds

of

the

farm

led

to

the

establishment

of

gradients

of

soil

fertility,

notably

decreasing

with

distance

from

the

homestead.

Farmers

tended

to

allocate

their

scarce

nutrient

and

labour

resources

in

the

ﬁelds

they

perceived

as

most

fertile

or

less

risky,

or

in

ﬁelds

around

the

homestead

where

high

value

crops

were

better

protected

from

marauding

livestock

or

theft.

A

close

interaction

was

also

found

between

soil

fertility

gradients

and

topography

in

these

highly

dissected

landscapes,

with

home-

steads

located

on

the

upper

positions

of

the

slope.

Such

interactions

between

inherent

soil-landscape

variability,

historical

and

current

management,

nutrient

balances

and

current

soil

fertility

were

later

documented

for

smallholder

systems

in

different

parts

of

Africa;

e.g.,

in

Zimbabwe

(

Ghana

and

Uganda

(

Ebanyat,

2009

Losses

of

organic

matter

from

agricultural

soils

are

the

result

of

the

imbalance

between

inputs

as

plant

litter

or

animal

manure

and

outputs

through

decomposition

and

soil

erosion.

This

balance

is

regulated

by

environmental

conditions,

soil

type,

litter

quality

and

management

practices.

Vegetation

clearance

for

cultivation,

as

practiced

in

the

African

savannahs

triggers

positive

feedback

loops

–

or

vicious

cycles

–

characterised

by

the

disturbance

of

soil

physi-

cal

properties,

increased

erosion,

accelerated

decomposition

rates

and

gradually

decreasing

C

inputs

to

the

soil

in

the

form

of

crop

residues

due

to

declining

crop

yields

(e.g.,

Such

process

may

be

counterbalanced

by

application

of

organic

fer-

tilisers

such

as

animal

manures,

or

further

aggravated

when

crop

residues

are

removed

from

the

ﬁelds

or

grazed

by

livestock.

The

magnitude

of

soil

carbon

losses

would

vary

for

different

soil

types

and

environments,

and

be

affected

by

the

characteristics

of

the

local

farming

(or

natural

resource

management)

system.

Abrupt

productivity

losses

(a

fast

responding

variable)

take

place

during

the

ﬁrst

5–10

years

after

woodland

clearance,

especially

on

sandy

soils

or

under

intensive

double-cropping

in

areas

with

a

bimodal

rainfall

regime

(

5

During

the

ﬁrst

decade

of

maize

cultivation

in

Gagnoa,

Ivory

Coast

maize

yields

decreased

to

about

one

ﬁfth

of

their

initial

level

when

no

nutrient

input

was

applied,

but

yields

could

be

sustained

with

application

of

large

amounts

of

mineral

and

0

1

2

3

4

5

6

7

1970

197

2

197

4

197

6

197

8

198

0

Maize yield

(t

ha

-1

)

4

6

8

10

12

1970

197

2

197

4

197

6

197

8

198

0

Soil organic

C (g kg

-1

)

Control

Comp

ost

Miner

al fertilisers

Comp

ost

+

Fertilis

ers

Source: Guibert

(1

999)

A

B

Fig.

5.

Results

of

the

ﬁrst

10

years

of

a

long-term

trial

conducted

in

Gagnoa,

Ivory

Coast

illustrating

how

yield

gaps

increase

in

time

(years

of

cultivation)

due

to

soil

fertility

depletion

when

no

nutrient

inputs

are

used

(control).

(A)

Maize

grain

yields;

(B)

soil

organic

carbon.

Source:

Guibert

.

organic

fertilisers,

or

both

combined

Soil

organic

carbon

(a

slow

variable)

decreased

in

about

20%

over

the

same

period

in

soils

that

did

not

receive

organic

amendments.

These

results

and

similar

ones

from

long-term

experiments

in

West

Africa

indicate

that

the

magnitude

of

the

yield

gap

increases

substantially

after

the

ﬁrst

ﬁve

years

of

cultivation

of

soils

cleared

from

savannah

vegetation.

Rehabilitating

soils

that

have

become

degraded

often

requires

substantial

investment.

The

amounts

of

manure

necessary

to

restore

productivity

of

degraded

outﬁelds

on

the

granitic

sandy

soils

of

Zimbabwe

were

as

much

as

17

t

ha

−

1

year

−

1

,

complemented

with

100

kg

ha

−

1

year

−

1

of

N

as

mineral

fertiliser

(

Earlier

calculations

indicated

that

an

equivalent

of

30

ha

of

grazing

land

would

be

necessary

to

sustain

productivity

in

one

hectare

of

cropping

land

on

these

soils

through

annual

application

of

8

t

ha

−

1

of

manure

(

Analysis

at

village

scale

indicated

that

only

one

third

of

the

village

cropland

could

be

covered

with

the

manure

produced

with

its

livestock

popula-

tion,

in

equilibrium

with

the

carrying

capacity

of

local

grasslands

A

quick

glance

at

livestock

population

densi-

ties

in

Africa

shows

that

sustaining

soil

fertility

exclusively

through

manure

applications

is

not

a

viable

option

in

most

places

(

6

Cattle

(as

much

as

human

population)

densities

are

greatest

in

the

highland

regions,

where

soils

are

inherently

more

fertile

and

agriculture

most

intensive.

Although

crop

productivity

may

be

sus-

tained

through

use

of

mineral

fertilisers

coupled

with

restitution

of

crop

residues,

this

does

not

appear

to

be

sufﬁcient

to

maintain

the

soil

organic

matter

contents

needed

(i.e.,

soil

C

output

is

larger

than

C

input,

resulting

in

soil

C

decline

–

cf.

5

http://www.earthtrends.org

As

a

consequence,

yields

often

plummet

when

fertiliser

applications

are

interrupted.

Discontinuous,

insufﬁcient

or

no

fertiliser

application

over

a

cer-

tain

period

of

time

may

lead

to

severe

soil

degradation

through

nutrient

depletion

and

loss

of

organic

matter.

When

fertiliser

or

organic

matter

applications

restart

after

a

certain

period

of

cultiva-

tion

without

them

soils

may

not

respond

immediately.

Often

crop

productivity

may

not

be

raised

back

to

the

yields

attained

before

P.

Tittonell,

K.E.

Giller

/

Field

Crops

Research

143

(2013)

76–90

83

Fig.

6.

Cattle

densities

in

sub-Saharan

Africa

(World

Resources

Institute;

).

Denser

cattle

populations

(between

20

and

more

than

50

cattle

per

km

2

)

are

distributed

across

an

east-west

band

of

northern

grassland,

and

along

a

northeast-southeast

band

of

eastern

grassland.

Countries

with

the

high-

est

densities

include

Ethiopia,

Kenya,

Uganda,

Tanzania,

Zambia,

Zimbabwe,

South

Africa,

and

Madagascar.

fertilisation

was

interrupted,

creating

a

new

system

state

at

lower

equilibrium

and

consequently

a

very

resilient

yield

gap

The

two

states,

responsive

and

non-responsive

soils

do

not

necessarily

represent

a

continuum.

Reversibility

may

be

lost

when

a

certain

threshold

of

soil

degradation

is

surpassed.

The

magnitude

of

the

distance

between

these

two

alternate

states

is

known

as

hysteresis.

This

is

a

concept

common

in

ecology,

but

it

has

also

been

used

to

characterise

phases

of

land

rehabilitation

(e.g.,

In

the

francophone

literature,

it

has

been

termed

the

‘

memoire

du

sol

’

and

the

param-

eters

that

deﬁne

response

curves

to

fertilisers

by

crops

(

In

a

long-term

experiment

in

southern

Benin

ceiling

maize

yields

ﬂuctuated

between

3

and

4

t

ha

−

1

when

large

amounts

of

mineral

and

organic

fertilisers

were

applied

together,

and

crop

residues

incorporated

into

the

soil

every

year.

Control

yields

without

fertilisers

coupled

with

residue

removal

dropped

from

about

1

t

ha

−

1

to

practically

nothing

after

8

years.

When

the

same

fertiliser

and

residue

management

regimes

were

applied

to

these

control

ﬁelds

from

the

10th

year

maize

yields

recovered

grad-

ually,

but

hardly

achieved

the

ceiling

yields

after

a

decade.

Current

research

aims

at

characterising

the

determinants

of

such

soil

mem-

ory,

and

ways

to

overcome

it

to

ensure

hysteretic

soil

rehabilitation,

for

different

types

of

soils

and

cropping

systems.

2.3.

Precision

agriculture

Precision

agriculture

was

proposed

by

as

a

means

of

ensuring

a

more

efﬁcient

use

of

applied

agricultural

inputs,

reducing

losses

and

thus

environmental

pollution.

Although

this

may

seem

rather

foreign

to

the

reality

of

subsistence

agricul-

ture,

smallholder

farmers

in

sub-Saharan

Africa

practice

several

forms

of

precision

agriculture.

To

start

with,

farmers

recognise

niches

of

soil

fertility

to

which

they

ascribe

different

local

names.

Thus

the

pervasive

localised

soil

heterogeneity

is

much

more

than

a

curiosity.

The

efﬁciency

with

which

nutrients

added

as

fertilizer

or

manure

are

captured

and

used

by

crops

is

strongly

reduced

by

soil

degradation

(

Farmers

recognise

the

existence

of

soil

fertility

gradients.

They

tend

to

plant

crops

earlier

and

more

densely,

weed

earlier

and

more

frequently,

and

apply

nutrients

as

fertilisers

and

manure

to

the

plots

that

are

already

more

fertile

(

Thus

the

resulting

differences

in

yields

are

due

to

gradients

of

management

inten-

sity

rather

than

soil

fertility

alone

(

Such

decision-making

patterns,

local

soil

classiﬁcation

and

soil

quality

indicators

may

form

the

basis

of

a

new

form

of

precision

agriculture

adapted

to

the

African

smallholder

context.

To

ensure

efﬁciency

this

new

form

of

precision

agriculture

should

recognise

and

target

resources:

(i)

To

diverse

regions

and

agricultural

contexts.

(ii)

To

diverse

rural

livelihood

systems.

(iii)

To

agro-ecological,

cropping

systems

and

soil

fertility

niches.

(iv)

Through

concentration

of

limited

resources

in

space

and

time.

(v)

In

synchrony

with

crop

demands.

points

out

that

in

most

cases

technical

solutions

are

available,

but

that

socio-economic

factors

deter

their

imple-

mentation.

We

listed

some

of

such

factors

relevant

for

African

agriculture

in

the

introductory

paragraphs,

and

many

publications

deal

with

issues

such

as

diversity

of

livelihood

strategies,

land

tenure,

integration

of

crop-livestock

activities,

climate

change

and

other

risks.

Likewise,

there

is

an

important

body

of

literature

con-

cerning

means

to

improve

the

synchrony

between

crop

nutrient

demands

and

nutrient

release

from

different

organic

resources

used

in

Africa

and/or

with

micro-dosing

and

point-placed

application

of

nutrient

sources

Point

(iii)

deserves

most

attention

in

the

context

of

yield

gaps

and

their

biophysical

causes.

7

illustrates

the

challenges

in

moving

from

recommendations

based

in

on-station

trials

to

deci-

sion

rules

for

niches

of

soil

fertility

within

heterogeneous

farms.

Fertiliser

experiments

provide

information

on

crop

responses

to

nutrient

inputs.

Based

on

this,

a

range

of

sensible

input

rates

can

be

identiﬁed

that

ensure

biophysically

efﬁcient

input

use,

avoiding

negative

externalities

to

the

environment.

Theory

indicates

that

the

amount

of

inputs

to

be

added

depends

on

the

balance

between

nec-

essary

investments

and

economic

returns

(both

affected

by

market

conditions).

However,

adding

nutrient

inputs

may

result

in

highly

variable

crop

responses

across

spatially

heterogeneous

farms.

In

smallholder

farms

as

small

as

0.5

ha

efﬁciencies

will

vary

enor-

mously

from

poorly

responsive

fertile

ﬁelds

(normally

the

home

ﬁelds),

to

responsive

or

poorly

responsive

infertile

ﬁelds

(normally

the

outﬁelds).

Applying

nutrient

inputs

in

the

most

responsive

ﬁelds

of

the

farm

will

ensure

most

efﬁcient

use

of

them.

Fertile

home

gardens

may

be

managed

with

‘maintenance

fertilisation’,

whereas

poor

ﬁelds

should

be

rehabilitated

with

long-term

addi-

tions

of

organic

matter

before

they

can

respond

to

nutrient

inputs.

This

means

also

that

the

impact

of

input

use

should

be

analysed

considering

time

horizons

longer

than

a

single

season.

A

major

challenge

in

designing

such

forms

of

precision

agricul-

ture

resides

in

identifying

these

three

categories

of

ﬁelds

(or

zones

of

fertility)

within

an

area

(cf.

Table

2

This

information,

together

with

knowledge

about

major

biotic

constraints

(weeds,

pest

and

diseases

–

cf.

exam-

ple

of

highland

banana

in

Section

2.1

)

in

a

certain

location

can

be

used

to

estimate

actual

yields

and

their

variability,

not

for

each

individual

farm,

but

for

the

entire

area

(cf.

8

While

yield

esti-

mates

are

still

likely

to

show

strong

stochasticity,

the

relevant

area

unit

to

estimate

yield

gaps

is

likely

to

vary

across

regions

depend-

ing

on

inherent

spatial

variability

and

demographic

patterns.

Well

calibrated

simulation

models

of

the

cropping

system

(cf.

A),

allowing

for

correction

factors

due

to

weed,

pest

or

disease

pres-

sure

could

be

used

to

estimate

actual

yields

–

although

obtaining

the

necessary

data

to

parameterise

such

models

would

be

often

more

demanding

than

measuring

yields

in

a

rigorously

identiﬁed

sample

of

farmers’

ﬁelds.

For

yield

gaps

to

be

informative

at

this

scale

of

analysis

it

is

not

only

important

to

consider

average

yields

in

farmers’

ﬁelds

but

also

their

strong

variability

in

time.

The

method

proposed

by

Lobell

(2013)

of

comparing

the

average

yield

of

the

last

one,

two,

three,

four

and

ﬁve

years

to

assess

yield

differences

would

be

most

pertinent

here,

as

this

provides

a

measure

of

variability.

Lobell’s

interpretation

does

of

the

resulting

patterns

of

variabil-

ity

to

infer

their

possible

causes

is

however

questionable.

For

instance,

if

these

averages

were

calculated

for

the

yields

in

the

control

treatment

in

5

,

their

value

would

vary

enormously

depending

on

the

5-year

period

considered,

whether

1971–1976,

1973–1978,

or

1975–1980,

while

in

all

these

periods

the

underlying

cause

of

yield

decline

was

the

same.

We

hypothesise

that

the

gap

between

actual

and

locally

attainable

yields

is

largely

attributable

to

farmer

resource

endowment

and

access

to

nutrient

inputs,

and

a

categorisation

of

yield

ranges

across

farm

types

and

production

environments

will

provide

very

good

approximation

of

this

yield

gap

and

its

space-time

variability

within

smallholder

African

farms.

5.

Discussion

and

conclusions

Yield

gaps

in

African

smallholder

farming

are

among

the

largest

in

the

world.

Given

demographic

projections

that

the

human

pop-

ulation

in

Africa

will

grow

most

rapidly

of

all

continents

in

the

coming

years,

there

is

an

urgent

need

for

productivity

to

increase.

The

underlying

causes

of

poor

productivity

of

African

agriculture

are

diverse,

but

the

challenges

faced

by

farmers

often

include

the

lack

of

access

to

agricultural

inputs,

the

intense

labour

demands

caused

by

lack

of

mechanisation,

the

small

size

and

increasing

frag-

mentation

of

farms

and

the

lack

of

capital

to

invest

in

building

productive

soils

in

harsh

environments.

Thinking

about

yield

gaps

or

yield

ceilings

makes

much

sense

when

examining

current

yields

in

the

most

productive

agroecosystems

of

the

world,

where

actual

yields

are

constantly

narrowing

the

gap

with

respect

to

the

yield

potential.

In

the

case

of

African

smallholder

agriculture,

we

believe

that

the

concept

of

yield

gaps

can

be

meaningful

when

at

least

two

main

components

of

the

total

absolute

gap

can

be

distinguished

and

studied

separately:

(1)

the

gap

between

the

water-limited

yield

potential

(

Y

W

)

and

the

locally

attainable

yield

(

Y

L

:

or

best

yields

attained

in

farmers’

ﬁelds),

which

provides

a

measure

of

yield

gap

attributable

largely

to

access

to

adequate

technologies,

and

(2)

the

gap

between

Y

L

and

the

average

yields

(

Y

A

)

farmers

obtain

across

their

heterogeneous

farms,

which

differences

across

farms

and

largely

attributable

to

access

to

resources.

Given

the

context

of

poor

agricultural

productivity,

of

about

1

Mg

ha

−

1

year

−

1

of

cereal

grain

across

much

of

sub-Saharan

Africa,

increasing

resource

use

efﬁciency

means

increasing

crop

primary

productivity

per

unit

resource

invested.

In

other

terms,

radiation

interception

in

space

and

time

and

its

conversion

through

pho-

tosynthesis

are

the

obvious

ecological

processes

that

need

to

be

addressed

ﬁrst.

Of

the

three

pillars

of

ecological

intensiﬁcation

pro-

posed

by

potential,

soil

quality

and

precision

agriculture,

the

second

is

the

most

urgent

in

Africa.

Little

or

no

productivity

gain

can

be

expected

by

raising

the

yield

potential

when

current

yields

ﬂuctuate

around

only

20%

of

that

achievable

(cf.

Degraded

and

poorly

responsive

soils

cover

large

areas

of

Africa,

and

represent

the

majority

of

poor

farmers’

ﬁelds

in

certain

regions

(cf.

Table

2

Yet

poorly

responsive

soils

are

hidden

due

to

the

patchwork

of

productive

and

unproductive

soils

across

landscapes,

coupled

with

the

location

of

more

fertile

ﬁelds

closer

to

roads

and

homesteads

which

can

mislead

those

con-

ducting

rapid

and

superﬁcial

assessments.

The

fertilisers

that

are

generally

available

simply

do

not

work

on

degraded

soils.

Substan-

tial

investment

to

build

soil

organic

matter

is

needed

to

restore

such

soils

to

a

responsive

state.

A

form

of

precision

agriculture,

as

a

means

of

making

an

efﬁcient

use

of

natural

resources

and

agricul-

tural

inputs,

can

be

redeﬁned

for

Africa

as

a

practical

approach

for

targeting

technologies

across

scales.

There

is

a

need

for

targeting

in

a

“best

ﬁt”

approach

from

a

basket

of

options,

rather

than

pushing

best-bet

approaches

or

“silver

bullet”

solutions

(

The

water-limited

yield

potential

seems

to

be

a

more

sen-

sible

reference

to

calculate

yield

gaps

than

the

yield

potential

determined

by

radiation

and

temperature

in

such

rainfed

African

farming

systems.

However,

the

use

of

such

theoretical

yield

lev-

els

may

yet

mask

important

yield

differences

within

and

across

farms.

For

instance,

doubling

on-farm

cereal

yields

from

0.5

to

1

t

ha

−

1

year

−

1

may

have

a

substantial

impact

on

local

livelihoods;

but

such

changes

may

be

barely

detectable

when

relative

yields

are

calculated

against

a

yield

potential

of

e.g.,

12

t

ha

−

1

year

−

1

(from

4.1

to

8.3%).

Likewise,

yield

gaps

calculated

with

respect

to

yields

obtained

in

experimental

stations

may

also

be

misleading

as

they

are

often

located

in

the

most

productive

environments,

overestimating

attainable

yields

under

farmer

conditions.

Hence,

we

recommend

effort

should

be

invested

in

deriving

sound

esti-

mates

of

locally

attainable

yields,

Y

L

,

which

as

the

maximum

yields

attainable

in

farmers’

ﬁelds

over

years

are

a

more

meaningful

refer-

ence.

We

suggest

a

framework

for

yield

gap

analysis

that

recognises

the

heterogeneous

farming

systems

and

landscapes

of

smallholder

agriculture

(cf.

9

A

sound

understanding

and

accurate

mea-

surements

of

yield

variability

is

essential

for

estimating

yield

gaps

(cf.