-
The Hydrology-Geomorphology Interface: Rainfall, Floods,
Sedimentation, Land Use (Proceedings of the Jerusalem Conference,
May 1999). IAHS Publ. no. 261, 2000. 75
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert
LESLIE SHAN AN Institute of Earth Sciences, Hebrew University
ofJerusalem, Givat Ram Campus, Jerusalem 91904, Israel
Abstract Runoff and erosion processes in desert watersheds were
investigated by studying ancient irrigation systems discovered in
the 100 mm rainfall region of the central Negev. Catchments
delivering flood waters to these areas ranged in size from small
plots to large watersheds. Runoff from small watersheds (less than
50 ha) varied from 4—12 mm year"1 compared to 0.5-2.5 mm year"1 for
large watersheds (greater than 1000 ha). Even in extreme drought
years, small watersheds produced at least 1.4 mm of runoff, while
large watersheds experienced "dry" years (i.e. without any runoff
event) about once in every three years. Ancient irrigation systems
using runoff from small watersheds were much more efficient
"water-harvesting" projects than those diverting flash flood flows
from large watersheds. Rates of erosion from small watersheds
averaged 3.6 mm century"1 (54 t km"2 year"'), originating mainly as
sheet erosion on the hillsides. Rates of erosion from large
watersheds, where main wadis are stable broad depressions with deep
loessial soils and a good winter vegetation cover, are about 4.6 mm
century"1 (701 km"2 year"1). In large watersheds where wadi
incision and headcutting processes are active, rates of erosion can
be expected to range from 7.6-12.6 mm century"1 (115-180 t km"2
year"1). Key words climatic changes; erosion; Negev Desert; runoff;
sustainable irrigation systems
INTRODUCTION
Numerous studies have reported that the technique of using
runoff and flash floods for irrigation has been practised for more
than 3000 years in China, India, Egypt, Iraq and Israel. Parts of
these ancient irrigated areas still produce reasonable yields but
extensive sections are now barren and desolate wasteland. There is
no general agreement on the cause of the decline of these systems.
Some investigators have favoured a theory of increasing aridity and
climatic change. Our research lends support to the conclusion
following the dendro-archaeological studies of Lipshitz &
Waisel (1978), that the climate of the Negev has not changed
significantly during the past 5000 years. The majority of studies
on other ancient systems demonstrate that three principle factors,
acting singly or in combination, led to the deterioration and
abandonment of projects: - The strong central authority that
planned and operated the systems was replaced by
one that was incapable of managing them. - The accumulation of
sediment in canals and fields reduced the amount of water
available for irrigation, and the heavy burden of maintaining
the projects led to their abandonment.
- Inadequate drainage, water logging, and the accumulation of
salts in the soils reduced yields below financially viable
levels.
-
76 Leslie Shanan
This paper focuses on the lessons learned from the ancient
irrigation systems in the Central Negev desert and shows how: (a)
the presence or absence of a strong central authority, and (b) the
processes of runoff and erosion affected their sustainability.
THE CENTRAL NEGEV DESERT
The Central Negev, the southern desert of Israel, is dotted with
extensive remains of ancient habitation and agricultural systems
(Fig. 1). In this 100-150 mm annual rainfall region, irrigation
based on the utilization of surface runoff from the meagre winter
storms was developed to a high technical degree, reaching its peak
during the Nabatean-Roman-Byzantine domination of the region from
about the second century BC to the seventh century AD. Desert
agriculture using hillside runoff was already
Fig . 1 The Negev , showing (I) lowland foothills and (II)
central h ighlands. The dashed and dotted line indicates the
boundary be tween the Negev and Jordan and Sinai. The triple dots
indicate the ruins of ancient cities: Haluza (Halutsa) , Rucheiba
(Rehovoth) , Nizzana (Auja-Hafir), Shivta (Subeita) , Kurnub
(Mamshi t ) and Avda t (Abda) .
-
Runoff, erosion, and the sustainabilily of ancient irrigation
systems in the Central Negev desert 11
practised during the Israelite Period (eighth and ninth
centuries BC) at the time of the Judean Kings (Evenari et al.,
1982). The densest settled areas have been discovered in the
lowlands and foothills and the highlands (Fig. 1).
The lowlands and foothills cover about 150 000 ha. The
morphological structure of this subregion is made up mostly of
Eocene limestone hills separating wide rolling plains, with
elevations ranging from 200-450 m a.m.s.l. A number of large wadis
whose sources are in the highlands, cut through the plains (Evenari
et al, 1982). The hillsides are generally covered with shallow,
gravelly, saline soils with immature profiles (Table 1).
The highlands cover about 200 000 ha and are composed of a
series of parallel anticlines of Cenomanian Turanian limestones and
cherts. Elevations vary between 450-1000 m a.m.s.l. (Evenari et al,
1982). Adjacent to the main wadis (gullies) lie relatively narrow
flood plains, and near the watershed divides, where the wadis have
not cut down to stable base levels, there are a number of expansive
plains (Table 1).
Table 1 S u m m a r y of ecological conditions in the Negev
Highlands (after T a d m o r & Hillel , 1956).
Habitat % o f t h e highlands area
Soils Plant associations Water available for plant g rowth
(mm)
Rocky slopes
Loessial plains
Wad i beds 3
8 0 - 9 0 Shallow, gravelly, saline
10 -15 Deep loessial soils; salts leached to 30 c m or more
Deep loessial soils or gravel and silt fill
Artemisietum herbae-albae and 1 0 - 6 0 Zygophylletum dumosi
Anabasidetum hausknechtii and 2 0 - 5 0 Haloxylonetum
articulati
Retama roetam associat ion with many annuals
Gravel ly wadis : 6 0 - 1 0 0 , loessial wadis : 4 0 0 - 6 0
0
"RUNOFF FARM" SYSTEMS
In order to investigate the techniques that enabled ancient
desert agriculture to exist under extremely marginal climatic
conditions, a research team (the late Professor M. Evenari, the
late Professor N. H. Tadmor and the author) was established in
1954. During 1954-1959 we surveyed and studied more than 100
ancient farming systems and irrigation projects (Evenari et al,
1982). We discovered, amongst other findings, that all the ancient
agricultural projects in this desert were based on utilizing runoff
from small and large watersheds—hence the term "runoff farm"
systems. In order to evaluate the hydrological and agricultural
potential of these methods, in 1958-1959 the research team
reconstructed two ancient runoff farms, one at Shivta and the other
at Avdat, where there are extensive remains of ancient systems
(Fig. 1). Scientific research at the reconstructed farms continued
for 25 years (Evenari et al, 1982).
The selection of the sites was partly influenced by a
controversy concerning man-made heaps of stones that had been
placed in mounds and long strips covering thousands of hectares in
the Central Negev desert (Fig. 2). These structures had been
observed in the 1870s, but were not investigated until the
1950s.
Various theories regarding the purpose of the mounds have been
reviewed by Evenari et al. (1982). Palmer in 1871, had concluded
that they were associated with vine cultivation on the hillsides
because his Bedouin guide had called them tuleilat el enab,
-
78 Leslie Shanan
Fig . 2 A n oblique aerial photo of 2000 year old patterns of
stone strips, near Nizzana. The remnants of ancient terraced
fields, surrounded by stone fences, can be seen in the background
in the broad loessial depression at the foot of the slopes. (Photo
by N. Tadmor , 1954).
i.e. grapevine mounds. His theory failed to explain how the
grapes grew on such extremely shallow saline soils, without any
irrigation in this 100 mm rainfall region. Some investigators
overcame this difficulty by proposing that the additional moisture
was obtained due to the mounds acting as "air-wells", with dew
condensing on the stones. Observations in gravel mounds rebuilt by
us, after the ancient model, neither collected dew (in sealed
containers to prevent evaporation) nor did they improve the soil
moisture conditions below them as compared with the surrounding
soil.
Mayerson, in 1959, suggested that the ancient farmers irrigated
the tens of thousands of hectares of hillside vines by carrying
water from wells or cisterns, completely disregarding the amount of
water that would be required or the human effort involved.
Previously, Kedar (1957) had proposed an entirely different theory
and argued that the main function of the mounds was not
viticulture, but to increase erosion from the hillsides and so
accelerate soil accumulation in the cultivated valleys.
Concurrent with these theories, we proposed that the purpose of
the mounds, and also the strips (Fig. 2, which the previous
investigators never recorded or referred to) was to increase
runoff, not erosion from the hillsides, with the aim of collecting
the maximum possible runoff from the slopes. It was in this
atmosphere of debate regarding the runoff and erosion processes in
the Negev that the two ancient farms were reconstructed and
experiments superimposed on them to study, inter alia, the
hydrological and climatic conditions of the area.
This paper focuses on case studies related to the techniques
used by the ancient farmers in their endeavours to establish
sustainable irrigation systems in this harsh
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 79
desert region, and refers to the results of our research work
over three decades (1954-1985) that deal with runoff and erosion
processes of the region.
The case studies include: (a) diversion systems from large
watersheds, (b) runoff farm systems associated with small
watersheds, (c) research data from two reconstructed ancient runoff
farms, and (d) sedimentation measurements in two large watersheds
not directly related to ancient agricultural systems.
The erosion data is presented as "depth per unit of time",
namely, mm per century (mm century"1) together with the
conventional unit, tons per square kilometre per year (t Ion"2
year"1). The unit 1.0 mm century"1, is approximately equal to 14 t
Ian"2 year"1, assuming an average sediment bulk density value of
1.45.
Diversion systems
Case study 1: The Nahal Lavan system Nahal Lavan (Wadi Abiad)
(Figs 3and 4) is the largest wadi in the vicinity of the ancient
town of Shivta (Fig. 1) and drains from the high plateau of the
Matred Plain through an area of barren rocky Cenomanian-Turonian
hills. Torrential floods have cut a deep wadi into the alluvial
plain that is narrow in the upper reaches but in the lower reaches,
widens out into extensive flood plains. Today, the wadi flows in a
gravel-bed watercourse typical of the area. Numerous remnants of
ancient walls and terraces are found in the alluvial flood plain
(Fig. 5). At a point where the drainage area of Wadi Abiad is about
53 km 2, an area covering 200 ha of terraces was studied in detail
(Fig. 4).
Because of the superimposition of many terrace systems one on
the other, it is often difficult to differentiate between projects
of different periods. However, the size and capacity of the
spillways, canals and drop structures, provided a key to their
understanding. Three types of spillway, all serving to lower water
from one terrace to the next, were found: (a) spillways with crest
lengths of 30-60 m for handling flows of 10-30 m 3 s"1 (Fig. 6).
(b) spillways with a crest length of 3-8 m for flows in the range
of 1-5 m 3 s"1. (c) small spillways up to 1 m wide, for flows less
than 1 m 3 s"1. Using these criteria, three different types of
developments were distinguished. The earliest flood irrigation
devices were discovered on the west bank of the wadi where massive
stone spillways with 30-60 m crest lengths are the common
structure. These spillways were connected to low earth embankments,
stretching across the plain, of which only faint traces remain
today in the form of low banks (Fig. 3). The spillways are of such
large capacity that they were capable of handling the entire flood.
The topographic location of this system indicated that it was used
when Wadi Abiad was a shallow depression and the earth embankments
were built to spread the runoff waters across the broad flood
plain. The wide stone spillways served to control and direct the
water as it passed from higher to lower elevations. This flood
plain water-spreading system was in use before Wadi Abiad had
become a deep gravel-bed watercourse.
Water-spreading systems with spillways of 3-8 m crest length and
diversion canals able to handle 1-5 m 3 s"1, were found mainly on
the northeast bank of the wadi, in the middle and upper reaches of
the survey area (Fig. 3). Some of these canals are more than 1 Ion
long, 5-10 m wide and aligned with a gradient of 0.4-0.5%. Each
diversion
-
00 o
ANCIENT AGRICULTURE IN WADI ABIAD
I:
n"bn sunn
D7) TO
yxn nmUl - namb
250 m
Cotiarime feetra) Biycrmn dail Orep structure.splmy Or Control
structure Dislriiufm Hitch Cistern
Bmcl Mounts Brircl Strip
• Itma mill Ba*a»Ba» lieiuit mllmttd irp i Stwc firci Fcae
m Ngvsf. âr WitcMewer F i g . 3 A map of Wadi Abiad (Nahal
Lavan) surveyed b y L. Shanan and N. H. Tadmor in 1955. No te the
series of diversion canals on bo th banks of the w a d i leading
part o f the flood flows to the terraced fields. Stage I sys tems
are located on the southwes t side of the presen t day wadi , and
stage II and III systems on the northeast side. Note h o w a wadi
has cut through Stage II sys tem towards the lower end of the
surveyed system. This tr ibutary wadi was "captured" b y W a d i
Abiad in relatively recent t imes (probably about the R o m a n Per
iod, 50 B C - 1 5 0 A D ) .
-
Runoff, erosion,
and the sustainability
of ancient irrigation
systems
in the Central
Negev desert
81
-
82 Leslie Shanan
Fig . 5 A wadi (gully) bank stabilization wall in Wad i Ab iad
(Fig. 4) . Note that the wall was buil t during at least three
different per iods . The terraced fields are 4.5 m above the wadi
bed at this point.
F ig . 6 A mass ive spil lway about 50 m long on the southwest
flood plain of Wad i Abiad (Fig. 4) , belonging to Stage I
development of a large wadi diversion sys tem (900 B C - 7 5 0 B C
, Middle Bronze II Per iod) .
canal irrigates an area of about 2-4 ha. The original stone
diversion dams have been washed away. Most of these terraces are
still in excellent condition.
The walls of the diversion canals and the associated terrace
walls, were built in stages (Fig. 3) because the silt and sediment
that accumulated in the fields and canals
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 83
compelled the irrigators periodically to raise the terrace walls
and diversion structures. Some of the walls reach 5 m in
height.
An interesting example of river-capture was found cutting
through the system (Fig. 3). The water-spreading system on the
northeastern side of Wadi Abiad was once a continuous
project—despite the fact that today, a 4 m deep tributary wadi
joins Wadi Abiad from the north. The alignment and elevation of the
walls on both sides of this tributary wadi, as well as the sizes
and capacities of the structures, clearly indicate that they once
belonged to the same system. In earlier times this tributary was
not connected to Wadi Abiad at all, but continued to flow west to a
separate water-spreading system. This tributary wadi was "captured"
by Wadi Abiad in relatively recent times, probably after the late
Byzantine period (about 650 AD).
The third use of the area was as "runoff farms" (see below)
connected to the adjoining small watersheds and not to the main
wadi. These farms adapted the existing structures and stone walls
of the diversion system to their needs. They no longer exploited
the floods in Wadi Abiad but used runoff from small catchments in
the adjoining hills. It is in connection with these runoff farms
that the small spillways are found.
Most of the flood plain, particularly the area lying west of the
wadi is badly damaged because it has been used for military
manoeuvres since the 1960s.
Case study 2: Wadi Kurnub Wadi Kumub cuts a narrow, steep gorge
through a limestone ridge 2 km south of the ancient town of Kurnub
(Fig. 7). At the point where the gorge opens onto the Tureibeh
plain, the ancient settlers constructed a large channel to divert
part of the Wadi Kumub flood waters (generated over a 27 km 2
drainage basin). The diversion chamiel is a solidly built stone
structure 5-9 m wide, with a gradient of 1:2000 over its 400 m
length. The channel led the water to an extensive (10-12 ha) system
of terraced fields that are all still in good condition. Excess
water from each terrace flowed to the next lower one, through drop
structures.
The diversion structure in the wadi diverted large quantities of
silt into the terraced areas and, in a manner similar to that
described above for the Wadi Abiad systems, the level of the
terraced fields rose continually, forcing the farmers from time to
time to raise the level of the terrace walls and diversion
structures. The three types of development—flood plain, diversion
system, and runoff farm—as described for the Wadi Abiad system,
were also discovered in the Wadi Kumub system.
Runoff farm systems
The term "runoff farm" (Evenari et al, 1982), denotes a group of
adjoining terraced fields surrounded by a stone-wall fence, forming
an integral unit of about 0.5-2.0 ha of cultivated land. A house,
cistern and/or watch-tower are often found within the boundaries of
this fence, and are indicative of a sedentary agricultural
population (Fig. 8). The hillside surrounding the farm served as a
catchment from which conduits channelled runoff water to the
terraced fields. Catchment and cultivated areas are thus seen as a
clearly defined unit—an integral part of an overall plan of
watershed subdivision.
In the desert, rainfall only wets the hillside soil to a shallow
depth, and is soon lost by evaporation. Ancient runoff fanners had
to depend on winter runoff water from the surrounding slopes to
supplement the meagre rainfall. Cultivation practices of the
-
Fig . 7 M a p of the Kurnub (Mamshit) system, surveyed by L.
Shanan and N. H. Tadmor in 1954. The remains of ancient walls
related to the evolut ion of the wadi and the terraces during three
diversion stages dating from 900 B C to 600 A D were clearly
visible in the field and are shown on the m a p . Note the divers
ion ditch leading runoff from two small catchments to a farm unit
belonging to the Byzantine Period (300-650 AD).
••III
• • l l
. V . *<
• • 1 1 1
F i g . 8 A n oblique aerial photo of a 2000 year old runoff
farm near Nizzana . The homes tead and the stone fence surrounding
the terraced fields are clearly seen. A wadi (gully) has b roken
through the farm, destroying sections of the terraced walls and
fields. (Photo by N . Tadmor , 1954.)
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 85
farmer in more humid regions aimed to prevent hillside runoff
and enhance the infiltration of rain into the soil. The desert
runoff farmer, operates the contrary principle. He aims to minimize
infiltration on the slopes, maximize runoff, and lead the runoff
from a relatively large area on the slope to small cultivated field
in the bottomlands.
Agricultural development under the physiographic conditions of
the Negev (shallow, gravelly soils and steep gradients), required
not only a mastery of the techniques of using surface runoff for
irrigation, but also an understanding of the skills of land
reclamation. One of the important aspects of ancient desert
settlement was terracing the small narrow wadis (Table 1), and it
is in these wadis that runoff farm units or groups of farm units
are found.
Case study 3: Runoff farms using runoff from small watersheds
Figure 9 is a map of a number of farm units in the Avdat region
where the fields were irrigated by a
Fig. 9 A detailed map (surveyed by L. Shanan and N . T a d m o r
in 1955) of a number of runoff farm systems in the Avda t area. The
contour interval is 10 m. The farm units, collecting conduits and
the mound and strip systems are easily discernible in the fields.
See text for further explanation.
-
86 Leslie Shanan
system of hillside collection channels. The small eastern valley
A (Fig. 9), is terraced in its lower reaches and gets part of its
water from the upstream wadi. The terraces, receive additional
runoff from stone built channel systems on the adjoining hills. The
western valley B (Fig. 9), is terraced from its lower end to the
head gully. Water is collected from adjoining slopes that have been
partly cleared of stones, to form patterns of stone mounds, strips,
and conduits.
The total area of all these systems is about 100 ha, of which 3
ha were cultivated. The ratio between the water-collecting and the
water-receiving area is about 33:1.
The catchments were subdivided into subcatchments, with conduits
transporting runoff from specific parts of the slopes to specific
fields. A single conduit generally collected water from a
relatively small area, 0.1 ha to 1.5 ha in size. The runoff was
thus divided into small controllable streams of water, suited to
the dry stone structures built by the ancient farmers. These small
flows could be easily controlled during a flood period.
Figure 10 is a detailed map of a complete runoff farm unit,
which is also shown in the centre of the flood plain on the
southwest bank of Wadi Abiad in Fig. 3. The stone mounds and strip
systems (A, B and C, Fig. 10) increased the runoff from the
catchment. Five runoff collecting channels (a, b, c, d and e, Fig.
10) directed the runoff to the terraced fields. The ratio of the
catchment to the cultivated area is a low 3:1 compared to the
normal average of 20:1, because the cultivated terraces were
originally part of a larger project planned to receive runoff water
from the early 900 BC-750 BC diversion system shown in Fig 3.
The earliest runoff farm units were found in the Mishor Haruach
and Matred Plains, dating to the Israelite III period (850 BC-600
BC). The farms were generally constructed near forts and water
cisterns located along the caravan routes (Evenari et al, 1982).
Negev (1979) dated several runoff farm units in the Avdat area to
the late Nabatean period (c. 100 AD). He suggests that the
Nabateans had applied the technology of collecting hillside runoff
water for filling their cisterns, to irrigating the terraced
fields.
Runoff farms were used continuously until about the seventh
century AD, when the rising tide of Islam swept the region. They
were abandoned because the Arab civilization had neither religious,
economic, or military motives for maintaining them (Negev,
1979).
RECONSTRUCTED ANCIENT FARMS
Case study 4: Reconstructed farms at Avdat and Shivta
Detailed descriptions of studies carried out at the two
reconstructed ancient runoff farms have been published elsewhere
(Shanan & Schick, 1980; Evenari et al, 1982). The hydrological
studies include 20 runoff plots and 13 watersheds.
Avdat The watersheds include eight catchments ranging in size
from 1 ha to 345 ha. The large watershed is a third order basin in
which many of the ancient terraced walls have collapsed. The other
seven watersheds are ancient subdivisions of a 30 ha watershed that
resulted from reconstructing the ancient "water-harvesting"
hillside collecting ditches. The subcatchments vary in size from 1
to 7 ha.
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 87
i l o u r l l n r
-
88 Leslie Shanan
subcatchments of the watershed. In Nahal Haroeh, a
detention-storage dam was constructed in 1954-1955 by the author to
use flood flows from a 43 km 2 catchment area (Fig. 11). Nativ
(1976) provides details of the watershed, the dam and its
construction, and daily and seasonal rainfall and runoff
measurements.
Most of the rainfall in the Negev is localized, often coming in
convective cells; the typical cell diameter is less than 10 km. The
proportion of the area receiving rainfall on a given day may be as
low as 20% (Sharon, 1972). Rainfall data were recorded at Sde Boqer
at a point 9 Ion south of the catchment centre and 3 Ion outside
the watershed boundary. Because of the spotty, erratic, and unequal
distribution of the storms on the
Fig. 11 A vertical aerial photograph of the Naha l Haroeh D a m
, constructed near Sde Boker in 1 9 5 4 - 1 9 5 5 . Downs t r eam
of the d a m are the terraced fields that receive this supplemental
irrigation. Note also the ancient terraced fields in some wadis ,
particularly the broad terraced sys tem southeast of the dam. A
tributary wadi cut back from Naha l Haroeh and completely bypassed
the terraced fields, p robably during the late Byzant ine Period,
about 600 A D .
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 89
watershed, the recorded rainfall values are only approximate
estimates of the probable average values for the catchment. Annual
rainfall for the period 1951-1981 ranged from 30 mm to 170 mm and
averaged 93 mm.
The maximum storage capacity of the dam is 250 000 m 3 and the
water is used for irrigation. Seepage and evaporation losses are
high (15 cm day"1) and an estimated 20% of the flood flows are
wasted through the spillway that functions about once every 7-10
years.
Case Study 6: Nahal Boqer
The Nahal Boqer watershed (35 km 2) adjoins Nahal Haroeh and is
physiographically similar to Nahal Haroeh. Observations relevant to
this paper included measurements of a wadi "back-cutting" process
that has been followed by the author since 1951. We have observed
similar wadi incision processes in Ramat Matred, Mishor Haruach and
Nahal Lavan areas.
RESEARCH RESULTS AND OBSERVATIONS
Runoff processes
Small watersheds and runoff-plots Based on two decades of
comprehensive rainfall and runoff measurements in small watersheds
and runoff plots in the Avdat and Shivta reconstructed farms, the
runoff processes have been studied in detail (Shanan & Schick,
1980; Evenari et al, 1982). The principle factors affecting runoff
in these two regions include: soil cover, seasonal infiltration
rates, basin shape, basin size, hillside slope, differential slope
contribution, overland flow lengths, and channel losses.
Our studies showed that both stoim and seasonal runoff are
mainly functions of catchment size and slope gradients (Table 2).
Average seasonal yields from the three catchment sizes studied
ranged from 2.4 mm (for third order catchments) to 26 mm (for small
plots, 80 m 2 in size). One of the most important factors
contributing to these differences in yield is the initial loss
(i.e. the threshold rainfall needed to initiate runoff, Table 2).
The initial loss in the third order basin (7.0-8.0 mm) can be
accounted for in the following manner:
crust wetting 2.5 mm overland flow losses 2.5-3.0 mm wadi
channel loss 2.0-2.5 mm
Table 2 Average annual runoff and initial losses from catchments
in the Central Negev .
Catchment Area (ha) Average annual runoff (mm) Initial loss
(mm)
Plots: 1% slope 0.08 26 2.5 10% slope 0.08 22 2.5 2 0 % slope
0.08 11 2.5
Sub-catchments 1-7 4 - 1 2 5.5 Third-order catchments 345 2.4 7
.0 -8 .0
-
90 Leslie Shanan
These conclusions explain why ancient runoff farming systems
based on small watersheds were relatively efficient water
harvesting projects. In the ancient systems, subdividing the first
order basins by artificial channels reduced initial losses from 5.5
mm to about 3.0 mm and so increased the frequency of runoff events
and the seasonal water yield.
The results of the runoff plot studies showed that slope
gradient is an extremely important factor affecting runoff. The 10%
gradient, for example, produced about 60% more total annual runoff
than the 20% gradient. Gradient is not only an indicator of the
topographic condition of an area, but is an index summarizing the
physiographic site conditions including soil depth, stone cover,
rock outcrops and vegetation cover. Moderate slopes with low
infiltration rates were found to be the principal contributors of
runoff in the Avdat area. Unequal areal distribution of rainfall
further reduced the runoff contribution from peaks of hills,
because valley side slopes were found to receive almost twice the
rainfall received by the peaks and knolls.
Stone clearing was found to have a significant effect on runoff.
Stone clearing increased average annual yields by 24% and 49% from
10% and 20% slope plots, respectively. The studies indicate that
this increase can be explained by regarding infiltration as a
two-phase process that takes into account the movement of air
escaping upwards through the soil profile and water infiltrating
downwards. This explanation supports the theory that the purpose of
the gravel mounds, tuleilat el enab, found in the region, were a
result of stone clearing by ancient farmers to increase the runoff
yield (Evenari et al, 1982).
Avdat large watershed (345 ha) Runoff from this watershed was
erratic with about 50%o of the years producing less than 0.5 mm.
However, in about 10% of the years, runoff exceeded 10 mm. The
average annual yield was 2.4 mm (Table 2).
This explains why the catchment-to-cultivated area ratio in the
large diversion projects is at least 35:1 (compared to 20:1 for the
small watershed runoff farms), indicating that the large systems
are relatively inefficient "water-harvesters".
These studies contributed towards understanding of the water
balance of the Upper Nahal Zin Basin. By evaluating the
differential contribution from first, second, and third order
catchments, the research showed that the amount of water available
for plant growth both on hillsides and on wadi bottoms was minimal.
On the hillsides, only about 30-40% of the annual rainfall
penetrates below the soil crust and becomes available for plant
growth because a 10-15 mm rainfall wets only 5-10 cm depth,
depending on local soil cracks and fissures. The threshold
conditions required for third order basins to contribute to the
regional water table indicate that: (a) a 2-3 h flow in the wadi
wets an average depth of about 40-60 cm; (b) that after a 3-4 h
flow, the maximum saturation depth in the centre of the wadi does
not exceed 1.2 m; and (c) flood water penetrates to the deeper
gravel layers in the wadi beds of third order streams only during
exceptional floods lasting more than 8 h.
Nahal Haroeh dam (43 km 2 watershed) Based on 35 years of
records at Kibbutz Sde Boqer (1955-1991), average rainfall was
calculated as 95 mm. In 13 out of the 35 years (37%) there was no
runoff, i.e. about one out of every three years is likely to be a
drought year. However, the spillway functioned five times during
the same 35 year period, averaging once in seven years (15%). The
average annual runoff was 2.4 mm.
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 91
The records also indicated that: (a) a threshold daily rainfall
of 8 mm was required to initiate a runoff event in Nahal
Haroeh; (b) a daily rainfall of about 32 mm produced a large
enough flood for the spillway to
operate. Daily runoff, for rainfall amounts of up to 50 mm
day"1, can be represented by the equation:
RD = K{PD-%) where, RD = daily runoff [mm], PD = daily rainfall
(
-
92 Leslie Shanan
Erosion from the Shivta cistern watershed Sediment accumulated
in the cistern over the 20 year period (1960-1980) at an average
rate of 0.042 mm year"1 from this 1.2 ha watershed. Average annual
rainfall was 93 mm year"1 and runoff 12 mm year"1.
1 - 2 1
Total sediment yield was therefore equivalent to 4.2 mm century"
(63 t Ion" year" ). Taking into account that only 10% of the total
sediment yield was coarse particle
bed load (see above), these low sediment values explain why the
silt-traps built by the ancient settlers at the entrance to these
runoff-collecting cisterns (Evenari et al., 1982) had capacities of
generally less than 100 1. The Shivta cistern, for example,
produced an average of about 80 kg (50 1) bed load sediment per
year, a quantity that could be caught and stored in the silt-trap.
Cleaning it once per year, or after each storm, was a simple
maintenance task.
Erosion processes in large watersheds with diversion systems All
the diversion projects studied showed a remarkable similarity in
their evolution that is characterized by three development stages:
- Stage I, Flood plain development Many of the major present-day
gravel-bed wadis
were wide shallow depressions in the alluvial plains, prior to
1500 BC. Intensification of the agricultural use of these fertile
areas necessitated the construction of the stone walls to prevent
the flood flows concentrating in the lowest depressions and
creating gullies. These walls were, over time, extended to spread
the water to extensive sections of the flood plain. The main
spillways of these systems were characterized by 30-60 m wide
openings capable of passing the entire flood (Fig. 3).
- Stage II, Diversion system At some period these flood plain
spreading projects were abandoned and the system deteriorated
through lack of maintenance. After abandonment, the floods cut a
1-3 m deep gully through the flood plain. The next settlers in the
area developed the technique of diverting the flood water by
constructing low stone barrages in the wadi and building canals to
lead the water to the flood plain. Each diversion canal served a
relatively small area (2-6 ha) and in most cases the new settlers
built on the remnants of earlier walls and structures they found in
the floodplains. The wadi continued to erode, and silt from eroding
banks and from upstream head-cutting gullies was deposited in the
terraced fields. This sediment raised the level of the fields until
a stage was reached when the height of the terraced walls had to be
increased and a new diversion structure and canal constructed
further upstream to irrigate the fields at their higher elevations.
The wadi bed was eventually several metres below the terraced
fields in the flood plain, and channelled between high stone
retaining walls (Fig. 3). Any break in these walls would cause
serious damage to the whole system. The construction of these
structures required an understanding of hydrology and hydraulics.
Thus, the period of these diversion systems must have been one in
which the science of engineering was well developed and the
projects were under the control of a central authority that was
able to manage the entire watershed and to enforce rules for
distributing the water during the short flood periods that occurred
two or three times every year.
- Stage III, Runoff farms The area was again abandoned at some
point in time. The systems may have become unmanageable because of
the silting problem or exceptional floods may have destroyed the
main structures. The next settlers in the
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 93
area no longer relied on using flood water from the main wadis,
but used runoff from small watersheds adjoining their terraced
fields to obtain supplementary irrigation water. These runoff farms
usually adapted older walls and structures to their new
requirements and used only small sections of the original diversion
system area that were situated close to hillsides.
Sediment in the Nahal Haroeh dam (43 km 2 basin) Sediment yields
in the dam (Fig. 11) were calculated on the basis of a revised
survey made in 1976 by the Department of Agriculture (E. Ador) and
on flood flows and discharges recorded by Kibbutz Sde Boker (Nativ,
1976, and personal communication, R. Yahel of Sde Boker) for the
period 1955-1991.
Average annual runoff was 3.0 mm year"1 and the total volume of
sediment that accumulated in the dam during a 21 year period
(1955-1976) was 23 000 m 3. Assuming a trap efficiency for the dam
of 55% (25-35% delivered to irrigation and 15-20% wasted over the
spillway), average annual sediment yield was 0.046 mm year"1
(4.6 mm century"1, 70 t km"2 year"1). This relatively small
sediment yield is due to the physiographic and ecological
conditions prevailing in Wadi Haroeh. It is a stable, wide
depression with a lush winter vegetation cover, and in sections of
the wadi the ancient terraces are still in good condition. The
relatively low gradients of the main wadi—about 1%—have also
contributed to the stability of the wadi and few gullies have cut
back into the flood plains.
Gully headcutting in Nahal Boqer (40 km 2 basin) At the lower
end of Wadi Boqer, the flood waters flow in a wide depression that
in some sections was stabilized by ancient stone walls built to
keep the flood spread across a 30-50 m wide zone that was covered
with perennial and annual vegetation (Table 1). Base levels of the
area have not changed appreciably since ancient times, perhaps for
5000-10 000 years. However, in 1950, at the lower end of the
valley, probably because a stone stabilizing wall broke during an
exceptional flood, an ever deepening and widening gully began
cutting back into the flood plain destroying the vegetation and
ancient walls as it moved upstream.
In 1951 the gully was about 1-2 m wide and 1 m deep. Thirty
years later it had become a wadi 30 m wide with vertical banks 2.5
m high, and it had cut back some 500 m into the original stable
floodplain. These observations indicate that an average of about
1250 m 3 soil had eroded annually during the 30 year upstream
advance of the headcut. This is equivalent to 0.031 mm year"1 on
the entire watershed or 3.1 mm century"1 (46 t km"2 year"1). This
is of the same magnitude as the rate of erosion from small
watersheds.
This degradation in the main wadi results in initiation of a
correlated incision process in tributary wadis, and a cycle of
upstream headwater erosion can be expected to further increase the
total rate of erosion from the watershed.
Sedimentation in ancient terraced fields The silt-laden flood
water used for irrigation brought about cumulative changes in the
levels of the terraced fields. The silt deposited in diversion
canals and in fields raised their levels about 2 m during a 600
year period, averaging about 3.3 mm year"1 (33 cm century"1).
The rate of rise in the levels of the fields irrigated from
diversion canals can also be estimated from the annual water use by
agricultural crops in the ancient systems
-
94 Leslie Shanan
which was about 300-400 mm year"1 (Evenari et ai, 1982). Given
that the average silt content of the flood waters reaching the
fields is about 1% by weight, the amount of silt deposited in the
fields was about 3-4 mm year"1, i.e. 30-40 cm century"1.
This magnitude of rise would require a concomitant raising of
the terraced walls at the same rate (Fig. 5). These estimated
sedimentation rates are considerably higher than those reported for
ancient irrigation systems in Mesopotamia (Iraq) which averaged
about 20 cm century"1 over a 5000 year period (Jacobson &
Adams, 1966). The situation was less acute in the runoff farm
systems, because the rate of sedimentation from the small
watersheds was only 6 cm century"1 (see above).
DISCUSSION
Rates of erosion
The rates of erosion reported above, are summarized in Table 3.
These observations indicate that erosion from small watersheds
averaged an extremely low 3.0-4.2 mm century"1
(45-63 t km"2 year"1). The large Nahal Haroeh watershed, where
the main wadi is a shallow wide loessial depression with a good
winter annual and perennial vegetation cover, produced about 4.6 mm
century"1 (70 t km"2 year"1) of sediment. Incremental sediment load
resulting from gullies headcutting back into a stable deep loessial
wadi (Nahal Boqer), was equivalent to additional 3.1 mm century"1
(46 11cm"2 year"1).
Large watersheds, where gully erosion and headcutting processes
are active, can be expected to produce at least 7.6-9.5 mm
century"1 (115-1521 km"2 year"1) of sediment as estimated in Table
4. In catchments where the headstream erosion process is also
taking place in the tributary wadis, rates of erosion may increase
by an additional, say, 3.1 mm century"1 and reach a total of 12.6
mm century"1 (180 t km"2 year"1).
Rates of erosion reported for other regions in the Negev are
given in Table 5. Several factors account for the relatively lower
rates of erosion in the Central Negev, principally:
T a b l e 3 Eros ion rates in the Central Negev: summary .
Watershed Area Rates of erosion: ( m m century" 1) (t km" 2
year" 1)
Avda t small watersheds 1-7 ha 3.0 45 Shivta cistern 1.2 ha 4.2
63 Nahal Haroeh d a m 43 k m 2 4.6 70 Nahal Boqer headcut 35 k m 2
3.1 46
T a b l e 4 Est imated rates of erosion in subcatchments of a
large watershed with active headcutt ing.
Source Est imated rate of erosion: ( m m century"') (t km" 2
year"')
Small watersheds 3 .6-4 .0 5 4 - 6 0 Stable wadis 1.0-2.0 15 -30
Act ive headcuts 3 .1-3 .5 4 6 - 6 2 Total 7 .6-9 .5 115 -152
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 95
Table 5 Rates of erosion in three selected regions in the Negev
.
Region and reference Annua l rainfall (mm)
Watershed area ( k m 2 )
Descript ion Erosion ra tes : ( m m century"')
(t km" 2 year" 1)
Eilat, Nahal Yael 31 0.5 Natural watershed 11.5 170 (Shick &
Lekach, 1993) Machtesh HaKatan 80 7.3 Natural watershed 7.5 120
(Greenbaum & Lekach, 1997) Beersheba (Lemanim) 220 2.9 4 7 %
contour ploughed; 16.0 259 (Laronne, 1989) 51 % loess-mantled
slopes
(a) The number of days per year with a rainfall greater than 5
mm averages about 6, of which only about 3 exceed 10 mm. Rainfall
of 25 mm day"1 has a probability of less than once every few years.
The maximum intensities of these rainfalls are much lower than in
the rest of Israel (Evenari et al, 1982).
(b) Overland flow distances are short, seldom exceeding 40 m.
(c) The loess soil of the area forms a crust after 2-3 mm of
rainfall have wetted the
surface, and it remains relatively stable under the low-velocity
laminar flow conditions prevailing before runoff concentrates in
rills and gullies.
(d) The stone cover, particularly in the hamada areas, protects
the soil surface and acts like a mulch.
(e) In the deep loess wadis, annual winter and perennial
vegetation stabilizes the depressions. Erosion rates increase
significantly however, after a head-cutting incision process has
been initiated.
The ancient farmer, by subdividing the watersheds into
relatively small subcatchments achieved two important advantages:
(a) He increased the amount of runoff that could be harvested from
the hillsides by
reducing the overland flow distances and seepage losses. (b) He
decreased the rates of erosion from the watersheds by minimizing
or
eliminating the development of rill, gully, and wadi erosion.
Herzog (1998) discovered an ingenious Israelite III period (850
BC-600 BC)
water supply system at Tel Beersheva that diverted flash floods
from a large 25 km 2
watershed into a tunnel leading to four underground cisterns
with a total storage capacity of 500 m 3. The system was abandoned
after 200 years because of serious sedimentation problems in the
tunnel and the cisterns.
During the Hellenistic period (350 BC-167 BC), one of the
cisterns (with a capacity of about 100 m 3) was again put into use
to store runoff water, collected this time not from a large
watershed, but from a small watershed.
The Tel Beersheva water cistern complex is the only system
discovered in the Negev that was supplied with runoff water from a
large watershed. All other cisterns, dating from the Israelite
period through to the Byzantine period (850 BC-650 AD), used runoff
from small watersheds (Evenari et al, 1982). Apparently after about
600 BC, the engineers realized that the high silt and sediment
loads carried by flash floods in large watersheds, caused rapid
rates of sedimentation in the diversion canals, tunnels, and
cisterns. Unable to meet the heavy burden of maintaining these
diversion systems, they decided to use runoff flows only from small
watersheds to fill their cisterns.
-
96 Leslie Shanan
Role of steep rocky slopes in producing runoff
Yair (1983), based on studies in a small watershed near Sde
Boqer, concluded inter alia, that from the point of view of runoff
production, a greater emphasis should be placed on the role played
by steep rocky hillsides and less on the gentle slopes and the
stoneless bare soils. Evenari et al. (1982) had recognized the
relative contribution of the rocky hillslopes, but their runoff
plot results showed that significant amounts of runoff from these
areas could only reach the fields if two conditions were met,
separately or in combination: (a) the overland flow distances of
the natural catchments were reduced to 40 m or less
by dividing the hillsides into subcatchments with cross-slope
collecting conduits; (b) the annual runoff yield could be increased
20-60% when the stones were cleared
from the surface and placed in mounds and/or strips. First, it
is important to point out that Yair's Sde Boqer experimental site
is not typical of watersheds where ancient agricultural systems are
found and the results may not be applicable to the ancient
runoff-collecting systems for the following reasons: (a) At Yair's
experimental site there are neither runoff farm systems or gravel
mound
and/or strip systems. Several terraced wadis and wadi
stabilization walls are found in a few subcatchments of the Sde
Boker watershed. However, ancient agricultural systems had been
constructed on no more than 1.3% of the Boqer-Ashalim catchment
area, compared to 2.7%, 5.4%, and 3.1% for the Shivta, Avdat and
Nizzana watersheds, respectively (Table 6). The ratio of the
catchment-to-cultivated area (Table 6) reflects different densities
of development and a diversity in landforms. In the Avdat
watershed, the systems are predominately runoff farms with mounds,
strips, and collecting conduits enhancing runoff production from
the hillsides, with an average catchment-to-cultivated area ratio
of 18:1; in the Shivta and Nizzana watersheds, flood plain
diversion systems adjacent to the main wadis are the principal form
of development with catchment-to-cultivated area ratios of 38:1 and
32:1, respectively. However, the catchment-to-cultivated area ratio
for Sde Boqer is 84:1 and reflects the paucity and infrequent
occurrence of stabilization walls in the watershed.
(b) This low level of runoff farm development in the Sde Boqer
watershed is one of the reasons why Yair (1983) did not find any
"agricultural installations" in the 13 km 2 loessial plain of
Mishor Zin. A second reason for the absence of development is that,
although Mishor Zin is only 8 km due north of Avdat city, it is
separated from the Avdat area by the deep canyon of Nahal Zin,
about 150 m deep and 2 km wide in parts. Access to Mishor Zin from
Avdat necessitates a 15 km trek circumventing this canyon (Fig.
12).
T a b l e 6 Ancien t agricultural cultivated areas in selected
watersheds in the Negev (after Kedar , 1967).
Watershed Catchment area ( k m 2 )
Cult ivated area (ha)
% watershed cult ivated
Rat io of ca tchment- to-cultivated area
Sde Boquer , Ashal im 255 303 1.2 84:1 Shivta 188 495 2.7 38:1
Avda t 125 678 5.4 18:1 Nizzana-Ruth-Lotz 560 1750 3.2 32:1
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 97
Fig . 12 A vertical aerial photo of the deep wide canyon of Naha
l Zin, cutt ing through Mishor Zin. Nor th of the canyon, the
buildings of the Ben Gurion Universi ty, Sde Boker Campus , can be
seen on the brink of Mishor Zin. No te the ancient runoff farm
systems located on that section of Mishor Zin situated south of the
canyon are only 4 k m distant from the ancient town of Avdat.
The predominant intensive runoff farm developments in the Negev
(particularly the gravel mound and strip systems and the collecting
conduits), are found mainly within 5-6 km distance of the main
ancient towns (Avdat, Shivta and Nizzana). Farmers were not
prepared to journey more than 6 km to reach and tend their
irrigated fields or walk tens of km to maintain their water
collecting systems. Furthermore, they were not prepared to live too
far from the main centres for security reasons. (In many developing
countries in which the author has worked, it was observed that
villages with 3000-5000 inhabitants cultivate irrigated lands
extending over 300-600 ha but located no more than 3 km from the
village, for the same two reasons). On the southern side of the
Nahal Zin canyon, and only a 4 km trek from Avdat, is a
continuation of the landforms of Mishmar Zin (Fig. 12). On this
loessial plain,
-
98 Leslie Shanan
ancient runoff farms with collecting conduits on the almost
stoneless gentle sloping area are clearly seen in the photo and in
the field. This loess plain was found by the ancient farmers to be
satisfactory from the point of view of its water harvesting
potential and its distance from the town of Avdat.
(c) The physiographic and ecological conditions of the Sde Boqer
watershed differ significantly from those at Shivta and Avdat.
First, the Sde Boqer geological formations are Cretaceous
limestones, dolomites and chalks; those at Avdat and Shivta are
younger Eocene limestones, "hamadas" and conglomerate "hamadas".
Second, the plant associations present are also significantly
different. In the Sde Boqer experimental site they are primarily
Vartenia phionedes-Originum dayi, indicative of relatively high
soil moisture conditions (Yair, 1983); in Shivta the Zygophyllum
dumosum association dominates while Artemesia herba alba represents
the common association in the Avdat area (Evenari et al, 1982).
(d) The studies by Evenari et al. (1982) were carried out at
Avdat and Shivta on an experimental layout superimposed on
reconstructed ancient agricultural runoff collecting systems, while
the experimental site at Sde Boqer was in no way comiected to any
ancient runoff inducement and collecting systems.
(e) The Avdat runoff plots, from which Evenari et al. (1982)
drew their conclusions regarding the effect of slope, cover, and
rainfall, were carefully designed and constructed in four separate
blocks with the slope of each plot uniform as well as equal within
blocks. The site was chosen so that the geological, pedological,
and ecological conditions were uniform. The experiment was planned
in a random block design, with four treatment replicates and
control plots. The runoff plots of the Sde Boqer site in contrast,
differ widely in their shape, size, dominant hillslopes, stone
cover, and geology. The site comprised three limestone formations:
Dorim, Shivta, and Netser, each with its particular ecological
environment. The majority of plots include two different geological
formations (Yair, 1983). In addition, overland flow lengths vary
from 55-76 m and hillside gradients from 12-29%, Furthermore there
are wide variations in the slope gradients within the plots
themselves. Consequently, in the Sde Boqer experimental site, it is
impossible to separate out from the data, the effects of
interrelated variables in an analytical, statistical, or simulation
analysis of the complex nonlinear relationships. This problem is
discussed in further detail below.
(f) The runoff plots at Avdat were uniform in size and shape (20
m long and 4 m wide), and the slope was uniform in each plot and
equal within blocks. The geological formation is also unifonn on
the site. The overland flow length on all plots was standardized
(20 m) and the rainfall micro-distribution was recorded with a
representative number of recorders.
The lack of uniformity and wide variation in attributes between
and within the Sde Boqer plots results in basic differences in the
statistical populations under study. For example, Yair et al.
(1980, Table 1) found that 82% of the total erosion and 46% of the
runoff originated on three plots representing only 35% of the area,
due to the differential bioturbation activity of porcupines and
isopods that was concentrated on these plots.
In contrast, the experimental design of the runoff plots of
Avdat limited the number of variables under study, and evaluated
them in a manner that overcame the
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 99
problem of equifmality (i.e. similar changes resulting from
disparate combinations of input, throughput, and output acting over
different periods of time).
Ancient terraces—sediment traps or stabilization walls?
An important aspect of understanding the runoff and erosion
processes in the Negev, is the question of whether the ancient
stone terraces on the hillsides and in the valley bottoms were
constructed to trap the silt washed down from the hillsides, or
whether they were built to stabilize existing soils in the wadis
and depressions by preventing gullies "cutting-back" into these
potentially productive areas. This issue has been referred to
widely in the literature during the past four decades. There are
several major objections to the hypotheses that the walls were
sediment collecting structures: (a) Our studies have shown that
erosion from small and large watersheds in the Negev
does not exceed 5 mm century"1 and 10 mm century"1,
respectively. Assuming a catchment-to-cultivated area ratio of
30:1, soil accumulates behind the terraces at a rate of 15-30 mm
century"1 This means that the ancient farmers would have to wait at
least 200 years until they had trapped 30-60 cm of soil behind
their terraces before they could expect to produce any viable
agricultural crop.
(b) Our archaeological discoveries in the field support the
hypothesis that the soils in the wadis and depressions pre-date the
construction of the runoff collecting systems: (i) The three-stage
evolution of the large wadi floodplains described previously, shows
that the first walls were built specifically to stabilize the wide
depressions and spread the flood flows across the floodplains (Figs
2, 8 and 11) and so prevent the development of gullies in the
depressions. (ii) Many of the flood plain development projects were
operated during the Israelite Period (1200 BC-1000 BC), at least a
millennium before the Roman-Byzantine period of development. (iii)
We discovered stone mound systems superimposed on the Roman road
just north of Avdat, showing that the Roman road predated these
runoff collecting systems (Evenari et al, 1982). (iv) Numerous
examples were recorded of wadi incision and "wadi capturing" (Figs
2, 8 and 11) clearly indicating that many of these head-cutting
processes occurred after the areas had been abandoned (probably
after 650 AD) when the systems became dilapidated through lack of
maintenance and gullies broke through the terraced walls.
(c) Terraces to enable the intensive cultivation of hillsides
and bottomlands have been built continuously throughout the ages,
for example, in Middle America and Peru during the first millennium
AD, in the USA during the Navaho period about 800 AD, as well as
more ancient examples in China, Nepal and North Africa. Modern
terrace development continues in many regions, particularly those
bordering the Mediterranean. All these projects are constructed
only on sites with soil profiles of at least 50 cm depth, and soils
that can be cultivated intensively and adapted to profitable upland
crops, vines, olives and orchards. They are always built as
stabilizing structures and are never constructed on barren
areas.
-
100 Leslie Shanan
It must be pointed out that the hillside terraces in the ancient
runoff farm systems in the Negev desert are always constructed in
hillside depressions or wadis, and the runoff from the hillsides
collected and led in conduits to the terraced fields. However, in
humid regions (1000-2500 mm annual rainfall)—for example in Nepal,
Korea, the Philippines and other southeast Asian countries—hillside
terraces are constructed as level-bench terraces to collect and
hold all the rainfall falling directly on the hillside slopes.
Hence, these are not runoff farm systems. The level-bench terraces
are usually constructed of earth embankments and the fields used
for rice, upland crops or orchards. Similarly, in the 500-100 mm
annual rainfall region, for example in the Jerusalem Hills in
Israel and in other countries in the Mediterranean region, the
walls of the bench terraces on the rocky slopes are constructed of
stone, the fields are not always level and are used for growing
vines, olives, orchards and grain production.
Stabilization terrace walls in both arid and humid regions
enhance the infiltration of water into the soil profile and so
contribute to improving yields and production. Based on the records
of the Nizzana Papyri (Evenari et al, 1982), the yields of barley
in the ancient farm systems in Nizzana (Fig. 1) in the seventh
century AD were recorded as being 8.0-8.7 times the amount
originally sown, compared to yields obtained at Avdat in the 1960s
of about 10-11 fold increases. The higher modem yields were due
probably to the use of fertilizers. Based on these production
estimates, terracing in the Negev desert enabled hillside and
valley fields to be cultivated intensively, either to increase the
range carrying capacity 10-20 times on steep slopes, or for grain
production, of about 1-3 t ha"1 barley (grain) on the deeper soils
of the gentle slopes (Evenari et al, 1982).
Gravel mounds and strips-a unique phenomenon
The tuleilat et enab man-made gravel-stone mounds and strips
(Figs 2, 9 and 10), are confined to specific areas in the Negev.
They have not been observed in any other part of the region, nor
reported elsewhere in the world. They are therefore a unique and
remarkable aspect of the Negev desert. Their occurrence in the
central Negev Highlands is always associated with hillside
runoff-collecting conduits and together they form an integral part
of the runoff farm systems. Our research has shown that they result
from the ancient settlers clearing stones from the hillside slopes
and placing them in mounds and/or strips. The bared soils on the
slopes increased the seasonal amount of runoff harvested from the
hillsides, particularly from the rocky slopes and the rock
outcrops.
Why is their occurrence a unique phenomenon? We concluded that a
combination of many physiographic, environmental and sociological
conditions must exist in a particular area to justify their
establishment, operation and maintenance. These preconditions
include: (a) Availability of land with agricultural potential which
only requires the addition of
200-300 mm year"1 of supplemental water to make it productive
and economically viable.
(b) Location of these potential agricultural areas close to
existing towns (within a 6 km radius) so that the farmer does not
spend more than about three hours a
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 101
work-day, walking to and from his fields to cultivate, sow,
irrigate, harvest, and keep guard over his crops and his
water-collecting systems.
(c) The soil in the fields is at least 1.5 m deep so that the
water holding capacity of the profile is not less than 250 mm, the
average depth of water that would be supplied by a flood
irrigation.
(d) The only reliable source of water for this supplementary
irrigation is runoff from the adjoining hillsides.
(e) Runoff from hillsides can be increased significantly if the
stones are cleared from the hillside surfaces; the cleared soils
comprise impermeable rock outcrops and/or have the characteristic
of forming an impermeable crust after a few mm of rainfall have
wetted the surface, so that most of the subsequent rain becomes
runoff.
(f) The natural overland flow distances are less than 40 m, or
alternatively, the catchments can be subdivided into subcatchments
with collecting conduits that also serve to reduce the overland
flow length to less than 40 m.
(g) Rainfall occurs mostly in the winter months when the soil
and rainfall temperatures are 0-5°C. Infiltration rates at these
low temperatures are significantly less than the rainfall
intensities of the average storm, thus ensuring runoff occurring
even with light rainfalls of 5 mm depth.
(h) The rates of erosion from the catchments are low, less than
4 mm century"1
(60 t km"2 year"1) so that sedimentation in the ditches, fields
and cisterns is not a critical problem.
(i) The area is under the control of a strong central authority
that has the political will and competence to plan and operate the
systems and enforce regulations for water rights and water
distribution procedures. The agricultural sector includes farmers
who are willing to introduce and use new techniques,
(j) An economic and social structure that does not rely only on
agriculture for its livelihood but insures a satisfactory and
stable income from several productive economic sectors. This
includes: desert caravan convoys continually moving through the
region along international trade routes; military camps and defence
installations, for local and regional security objectives (like the
large Nabatean army camp for about 1000 permanent soldiers at
Avdat); churches and monasteries serving as local and national
religious centres of learning and study found in all the ancient
towns (Fig. 1); and a class of entrepreneurs who are prepared to
initiate new economic ventures in the region (like the
horse-breeding enterprise at Kurnub during the Nabatean period
(about 100 BC), or the large Nabatean factory at Avdat
manufacturing exquisite hand-painted delicate pottery of high
quality and supplying to the entire region, or the luxurious public
bath-house projects at Avdat and Kurnub serving the caravan convoys
along the regional trade routes (Negev, 1979).
The concurrent presence of all these circumstances in the
central Negev Highlands enabled the ancient settlers to introduce
runoff inducing water-harvesting systems into limited areas near
the ancient towns of Avdat, Shivta and Nizzana.
Analytical solutions to the runoff process
The scientist often finds himself having to chose between simple
or elaborate analytical mechanisms for understanding complex
relationships and has to select from
-
102 Leslie Shanan
a number of widely differing analytical
methodologies—mathematical, statistical and simulation modelling.
Systems that can be solved mathematically are generally very simple
sub-processes and essentially only of academic interest, with
limited practical value. Analytical analysis is nevertheless
important because it gives an insight into fundamental aspects of a
problem.
Methods of studying the behaviour of involved, even chaotic
interacting systems, have been developed using advanced statistical
methods and/or with the simulation of continuous and parallel
systems. In simulation modelling, the state of a system at any
particular point in time, is expressed quantitatively; changes in
the system are described in mathematical statements or as input
data. Ecological, mathematical and programming aspects are
interwoven into a simulation model, and the use of
continuous-system-modelling (CSMP) languages has been developed
specifically for this purpose (de Wit & Goudrian, 1978; Shanan
& Schick, 1980).
Two independent methodologies have been used to analyse the
complex process of runoff from the Negev desert watersheds: a
multivariate analysis for predicting annual runoff yield, and a
digital simulation model for predicting individual storm and total
seasonal runoff (Shanan & Schick, 1980; Evenari et al, 1982)
and are briefly reviewed below.
Multivariate analysis Annual runoff for watersheds varying in
size from micro-catchments (less than 0.1 ha) to third order
watersheds (up to 300 ha) were correlated with watershed size,
annual rainfall, hillside slope, and stone cover. The results are
presented as a nomogram in Fig. 13 (Evenari et al, 1982, Fig. 91).
Yair (1983) used Fig. 13 to extrapolate an extreme value for
predicting the threshold amount of annual rainfall needed to
initiate runoff production from a 1.5 ha catchment with a hillside
slope exceeding 20%. He concluded that the nomogram predicts that a
minimum annual rainfall of 75 mm is required for the catchment to
provide "a minimum amount
0 100 200 300 Fig . 13 N o m o g r a m of rainfal l-runoff relat
ionships at Avdat , showing effect of size, slope, and surface
cover of the catchments on annual runoff. Scale 0 - 3 0 0 , runoff
( m ha" 1 ); scale 0 - 1 5 0 , rainfall (mm) (after Evenar i et al,
1982).
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 103
of water". He then commented that annual rainfall amounts of
less than 70 mm are common in the Negev, and if the nomogram result
is correct, then the ancient runoff farm systems were inefficient
users of runoff. Unfortunately, Yair used the nomogram erroneously.
Instead of applying the curve titled "Watersheds 5-70%", he applied
the curve titled "Microcatchments >20%." If he had used the
nomogram correctly, he would have concluded that the threshold
annual rainfall for this specific watershed to produce runoff is 25
mm (not 75 mm). During the 38 years of records at Avdat, the lowest
recorded rainfall (1962-1963) was 28 mm, a one-time occurrence.
Even in 1962-1963, all the watersheds produced significant amounts
of runoff, averaging 1.4 mm (giving an average coefficient of
runoff of 5.0% for the extreme drought year).
Simulation modelling The simulation model (Shanan & Schick,
1980) developed in CSMP language to simulate the runoff process,
used storm rainfall as input, and runoff measurements from plots
and catchments as output. The structure of the model was verified
in three stages, with an optimization algorithm minimizing the sum
of the squares of the residuals between observed and simulated
events. In the first stage, the model was confined to runoff plot
data so that time and areal variations in rainfall and the effects
of overland flow would be minimized. Functions to account for the
effects of slope, stone cover, infiltration, and evaporation were
evaluated. In the second stage, the model was expanded to simulate
the processes on seven sub-basins that were considered as a
combination of plots, modified by areal rainfall distribution
patterns and overland flow conditions. Finally, in the third stage
the model was adapted to a third order basin, which was considered
an assemblage of sub-basins, modified for main channel losses.
The basic premise of the model states that runoff is initiated
after a crust is formed. Infiltration rates of the soils are
considered to be greater than rainfall intensities until the crust
is puddled and saturated. This threshold requirement, called the
"maximum saturation deficit", is regarded as including depression
and interception storage. When the crust is partly saturated, the
amount of water required to bring it to saturation is called the
"saturation deficit". Runoff is initiated when two conditions are
fulfilled: (a) the crust is saturated (saturation deficit is zero),
and (b) the rate of rainfall exceeds the infiltration rate of the
crust.
The sub-basin model was modified to include the effect of areal
rainfall distribution, differential contribution of areas, seasonal
infiltration rates as functions of soil and rainfall temperatures,
raindrop impact, stone cover, hillside slope, overland flow and
channel losses. Maximum saturation deficit values are about 2.5 mm
and infiltration rates of the saturated crust about 1.0 mm h"1. A
normal distribution in the areal variation of infiltration rates
was found to give the best fit for the simulated results of the
plots. The simulation of storm and annual runoff was performed in
two stages: (a) parameter development stage: trial and error runs
of 63 storms during the 1964-1967 period; and (b) validation stage:
simulation runs of 42 storms during the 1967-1969 period using the
"best fit" values developed from the first period.
Although the problem of equifmality was successfully resolved in
the plot stages of the model, it was not solved for the sub-basins
and the third order basin because of the differential effects of
hillside slope, slope distribution, overland flow distances, and
channel losses. Nevertheless, the model gives satisfactory results
and the inter-
-
104 Leslie Shanan
100 9 0 80 70 60 50 iO 30 20 !0 0
/,Area with infiltration capacity
equal or less than indicated value
Fig . 14 N o m o g r a m (after Evenari et al., 1982), showing
the distribution of infiltration rates on a runoff plot as
influenced by season, stone cover, s lope, and moisture content of
the crust. M a x i m u m saturation deficit ( M D E F ) is the
amount of water required to saturate the crust to initiate runoff.
The n o m o g r a m shows that M D E F is 3.4 m m , and 4.3 m m for
the spring, winter, and autumn seasons respectively. T w o examples
are given in the figure: for a 6.0 m m rainfall in the spring
season ( ) on a plot with a 10% slope and with stone cover, the m e
a n infiltration rate of the plot would be about 4.0 m m h"1
varying from a m i n i m u m of 2 m m h " ' to a m a x i m u m of 6
m m If1 for different points of the plot; and for a 6.0 m m
rainfall in the winter season ( ) for the same 10% slope and stone
cover, the m e a n infiltration rate would be 3.2 m m h " 1
varying from a m i n i m u m of 1 m m h"1 to a m a x i m u m of
5.0 m m h"'.
relationship between the factors affecting infiltration rates on
plots (cumulative rainfall, maximum saturation deficit, season,
stone cover, hillside slope angle, and a normal distribution of
infiltration rates) is presented in nomogram form (Fig. 14).
The model includes: (a) functions for evaluating the sensitivity
of the results so as to enable the research scientist to decide
which parameters mainly control the complex process and so guide
him to further experiments and studies, and (b) functions for
formulating the model stochastically. Consequently, the model has
applicability both as a research tool and an engineering planning
technique.
THE SUSTAIN ABILITY OF THE ANCIENT IRRIGATION SYSTEMS
Our studies highlighted the factors that enabled several
civilizations (Israeli, Nabatean, Roman and Byzantine) to establish
irrigation projects in the Central Negev desert during a 1500 year
period from about 850 BC to 650 AD. Their operation depended on the
ability of the settlers to understand the complex environmental
conditions of the
-
Runoff, erosion, and the sustainability of ancient irrigation
systems in the Central Negev desert 105
desert (particularly the climate, pedology and hydrology), to
master innovative technologies that enabled them to exploit
hillside runoff and flash flood flows in the wadis, and to
establish irrigation layout and design criteria for constructing
dry stone structures to serve as stabilizing terrace walls,
diversion canals, distribution systems, and erosion-control
measures.
The research also highlighted the environmental and
socio-economic constraints following in the wake of this
development and limiting the sustainability of many of the
projects. These included: (a) Environmental constraints: although
erosion, transportation and deposition of
sediment are processes that have occurred throughout geological
time, man's interference with the balance of nature results in
changes in the relative levels of the flood plains and the wadi
bottoms at accelerated rates of 10-30 cm century"1. This
necessitated the raising of the height of the terraced walls and
diversion structures. The increasing burden of maintaining the
projects (cleaning the canals and raising the heights of the
terraced walls) eventually exceeded the capabilities of the
farmers, and the larger systems were abandoned.
(b) Socio-economic constraints: the planning, operation and
maintenance of the projects required a central authority to manage
entire watersheds and possess the power to enforce the laws for
distributing the water during the short flash flood periods. The
projects were abandoned when the central authority was no longer
interested or capable of carrying out these duties.
Furthermore, the economic and financial viability of irrigated
agriculture based on small family-sized plots of less than 1 ha in
size, was dependent on the overall economic viability of the
community in the region. Agriculture was economically sustainable
provided that it was integrated into a regional economy that
comprised several economic sectors including: large-scale military
installations and/or army camps protecting the security of the
region; regional and international trade routes passing through the
area that served as a ready cash market for local goods and
services (such as fresh foods and bath-house facilities); the
existence of local entrepreneurs of industries for exporting goods
to other regions (horse-breeding, exquisite pottery, etc.); and the
presence of large-scale regional religious institutions of
learning.
Understanding the reasons for the failure of irrigation projects
in the past is a prerequisite to proposing feasible ways of
improving the sustainability of irrigation projects today. During
the last four decades, numerous attempts have been made by
governments and international agencies to improve the present
short-comings in irrigation projects, particularly in developing
countries (Shanan, 1998). Unfortunately, the results of improvement
schemes have fallen far below the planners' expectations. The
lessons learned from the ancient irrigation projects in the Central
Negev desert can be of great value to planners who are searching
for ways to make present-day irrigation projects sustainable over
long periods of time.
REFERENCES
De Wit, C. T. & Goudrian, T. (1978) Simulation of Ecological
Processes. Centre for Agricultural Publication and Documentation,
Wageningen, The Netherlands.
Evenari, M., Shanan, L. & Tadmor, N. H. (1982) The Negev:
The Challenge of a Desert (second edn). Harvard University Press,
Cambridge, Massachusetts, USA.
-
106 Leslie Shanan
Greenbaum, N. & Lekach, Y. (1997) The potential water
resource and efficiency of detention storage reservoirs in the
Machtesh Hakatan: final report (in Hebrew). Department of
Geography, Hebrew University of Jerusalem, Israel.
Herzog, Z. (1998) Water supply in Iron Age Tel Beersheva. Paper
presented at the First International Congress in Rome on the
Archaeology of the Ancient Near East. Department of Archaeology,
Tel Aviv University, Israel.
Ionides, M. G. (1939) Report on the Water Resources of Jordan
and their Development. Publication of the Government of
Transjordan, London.
Jacobsen, T. & Adams, R. M. (1966) Salt and silt in ancient
Mesopotamian agriculture In: New Roads to Yesterday (ed. by J. R.
Caldwell), 466-479. American Association for the Advancement of
Science, Basic Books Inc., New York, USA.
Kedar, Y. (1967) The Ancient Agriculture in the Negev Mountains
(in Hebrew). Bialik Institute, Jerusalem, Israel. Laronne, J. B.
(1989) Probability distribution of event sediment yields in the
northern Negev, Israel. In: Soil Erosion in
Agricultural Land (ed. by J. Boardman, I. Foster & J.
Dearing), 481-492. John Wiley & Sons, London, UK. Lipschitz, N.
& Waisel, Y. (1987) The climatic and vegetation history of the
last 5000 years in the Sinai Penninsula: a
dendroarchaeological study. In: Sinai (ed. by Y. Grados). Israel
Ministry of Defense, Geographic Research and Publications, Tel
Aviv, Israel.
Nativ, R. (1976) Seepage losses in the Nahal Haroeh Dam Near
Kibbutz Sde Boqer (in Hebrew). MSc Thesis, Institute of Desert
Research, Ben Gurion University, Sde Boker Campus, Israel.
Negev, A., (1979) The Nabateans in the Negev. In: The Land of
the Negev (ed. by A. Shmueli, Y. Gradus & F. Mor), 226-269. Tel
Aviv University and Ben Gurion University, Israel.
Schnitzis, E., Merrowitz, L., Yanowitz, Y., Chavilz, V. &
Ben-Zvi, A. ( 1997) Peak flood flows in the Negev. Hydrological
Service, Water Commissioner, Hydro Rep. 97/2, Jerusalem,
Israel.
Shanan, L. (1998) Irrigation development: Proactive planning and
interactive management. In: The Arid Frontier (ed. by H. J. Bruins
& H. Lithwick), 251-276. Kluwer Academic Press, Dordrecht, The
Netherlands.
Shanan, L. & Schick, A. P. (1980) A hydrological model for
the Negev Highlands: effects of infiltration, runoff and ancient
agriculture. Hydrol. Sci. Bull. 25(3), 269-282.
Sharon, D. (1972) The spottiness of rainfall in a desert area.
J. Hydrol. 17, 161-175. Tadmor, N. & Hill, D. (1956) Survey of
the possibilities for agricultural development in the southern
Negev. Bull. Agric.
Res. Station, Rehovolh, Israel, no. 61. Yair, A. (1983)
Hillslope hydrology water harvesting and areal distribution of some
ancient agricultural systems in the
Northern Negev desert. J. Arid Environ. 6, 283-301.