Editor's note: the illustrations, at the end of the book on the printed version, have been integrated here in the body of the article.

1Water is the most limited resource for agriculture in Yemen, the optimal use of which determines the agricultural and economical development of Yemeni rural areas. The amount of water available for agriculture and domestic use correlates with the social and economic standard of living.

2The renewable water resources of Yemen are estimated at 5,2 km³/a¹ which was equivalent to 445 m³/a/cap. in 1990 or 346 m³/a/cap. in 1995. The per capita volume of renewable water shows a dramatic decline due to the significant increase of population (Table 1). Taking the index for water scarcity defined by Marlin Falkenmark, 1700 m³/a/cap is assumed to be general lower limit for a moderate standard of living, while 1000 m³/a/cap is the benchmark for absolute water scarcity accepted by organisations like the World Bank and used by Falkenmark. Yemen had already exceeded this benchmark in the 1950's. That a water crisis did not occur much earlier is itself a sign of efficient water use.

3The traditional type of water management is both ecologically adapted and sustainable. Therefore it is of scientific interest and practical use to evaluate with today's scientific approaches and techniques Yemen’s traditional water management methods.

Geographical condition

4Throughout history Yemen has been known for its relatively abundant rainfall—especially in comparison with the rest of the Arabian Peninsula—and therefore was called “Arabia Felix” even in ancient times. But with respect to the general geographic conditions, this rainfall is often overestimated. Reasonable annual rainfall occurs only in the western and southern slopes of the Yemen mountain massif.

5In the highland plains, the annual precipitation diminishes to under 300 mm/a (e.g. San^cā’ 153 mm/a, Radā^c 194 mm/a).² The seasonal rainfall distribution shows two recognizable seasons. This pattern prevails in most regions of the Yemen Mountain Massif although the proportions vary across the region. The analysis of the annual and seasonal rainfall pattern of Ta^cizz - the closed weather station of the detailed study area - shows the following results. ³

6The considerable annual average of 587 mm/a is not very representative, and the standard deviation of 160 mm describes the high variability from year to year. With a classification adapted from Hastenrath and Heller,⁴ the rainfall pattern can be described as alternating between extremes. In meteorological terms, every other year is “extreme,” and every third year can be considered “dry.”

7The seasonal distribution can described as equinoxial rainfall, with a clear correspondence with the seasonal movement of the sun reaching maxima in April and September. The error bars in Figure 3 indicate the standard deviation.

8The relief energy in the Yemen Mountain Massif is significant. Apart from the High Plains, good agricultural land is scarce. Never the less this seems to be no handicap for farming. Pure rainfed agricultural is therefore possible only in a very limited area. Most agricultural systems rely on additional irrigation or other methods to augment the limited rainfall.

9The conditions described above cause several constraints on agriculture in the mountains:

• high potential erosion;
• slope conditions allow only small scale fields which require high maintenance efforts;
• limited precipitation for agricultural production in average years;
• high variability of the precipitation.
• All these constraints result in the magnificent design of the terraced landscape for which Yemen is known:
• terraces to provide agricultural area;
• terraces for erosion protection;
• terraces as retention schemes to store precipitation.

Study site

10The detailed study area is called Ma^cāmirah and is located in the Mawasit district of the Ta^cizz Governorate, approximately 30 km south of Ta^cizz city (long. 44°11', lat. 13°18'). The altitude ranges between 1750 to 2000 masl.⁵ The bedrock shifts from highly fragmented Tertiary volcanic rock in the west to Medj-Zir and Tawilah sandstone in the east.⁶ The catchment is approximately 40 ha in area. The normal growing season is May to October. The prevailing crops are sorghum (subsistence production) and qāt (market and subsistence production). Due to the marginal position and high altitude of the catchment, groundwater is not used for irrigation. Irrigation is 1) pure rainfed or, 2) rainfed with additional water-harvesting. No other water sources are available for irrigation. The ridge and upper part of the slopes are used for water-harvesting and goat pasture, and mainly consist of bare rock with a little scattered vegetation. Terraces start at different sections of the slope. In general they are protected by semi-impervious channels that control the amount of water and sediment reaching the terraces. Some of the channels lead the water out of the catchment, which causes the catchment to have additional “micro-outlets.”

11There are various types of terraces. Terrace walls usually lack static features. They lean against the soil body of the terrace and simply protect the soil from being eroded. Only a few have a “house wall” structure.

12The traditional irrigation system was modified in several steps during the late 1960s and early 1970s. New channels were cut from the uphill run-off areas through the terraces to the wādī, creating a bypass that cuts some of the terraces off from former water-harvesting possibilities. Adjacent terraces are still irrigated from the channel. The main reason for the change was reduction of sediment transport to the fields. The disadvantage was accepted of reduced water for irrigation to the field. The major advantage was the reduction of labour costs for the maintenance of terraces.

13No long-term data on the climate are available, but due to the higher altitude, rainfall might be slightly higher than in Ta'izz city. Even so Ta^cizz meteorological data are used for long-term considerations. The population density is above 250 p/km², the population growth exceeding 4 %.⁷ The local agricultural production covers about 30 % of the dietary needs. All other food is imported.

Conceptual approach

14As explained above, climatic conditions are the limiting factor for agricultural production. This is clearly shown by the fact that evaporation exceeds precipitation in every month of the year. Figure 4 shows the cumulative graphs of precipitation (P) and evapotranspiration (ETpot) for the May-October growing period, where ETpot constantly rises to more than 800 mm while precipitation levels peak at 360 mm.

15Even though precipitation and ETpot are only rough estimations for agricultural water supply and water demand, the gap between the two curves illustrates the water deficit. In the absence of other water sources the traditional solution to the deficit is the collection of rainwater and its redistribution on terraces. This system is applied and adapted in many ways to the local conditions and can be assumed to be “optimised by tradition”. Many previous authors have described such water-harvesting/surface irrigation schemes.⁸ Quantitative approaches to evaluate the efficiency of this scheme are rare.

16To evaluate these water-harvesting terrace irrigation schemes, the following conceptual design is used (Figure 5). A set of adjoining terraces is connected with a given water-harvesting area. Consequently, the effective rainfall on the water-harvesting (WH) area irrigates the corresponding set of terraces.

17The enhancement of precipitation relative to the rainfed conditions correspond to the proportions of WH-area and terraced area. The relation between those two areas can be described with a “terrace factor”. This factor is 1.0 in the case of directly rainfed terraces, and is above 1.0 if terraces receive additional irrigation from upslope water-harvesting areas. The potential retention storage of the terraces is determined by the volume of the terrace body and the field moisture capacity. The optimal relation depends on the potential water storage volume of the terraces and characteristic rainfall amounts.

18For optimal access to and processing of the data a Geographical Information System (GIS, ArcView 3.1) is used.

Measurement design and field work

19The design of the field work in 1998 was intended to provide all necessary information for the conceptual approach described above. The meteorological parameters⁹ were collected with an automatic weather station at 30-minutes intervals. Precipitation was recorded at 10-minute intervals. In addition, eight manual rain gauges were operated on a daily basis. Run-off was gauged at four locations in the catchment. At three points the surface run-off of the water-harvesting areas were measured with v-notches. The fourth gauge, a cemented flume with a float gauge, was located at the catchment outlet.

20A soil survey was carried out to determine the distribution of soil textures, using RRA-techniques.¹⁰ Gravimetric soil moisture samples were taken from ten selected terraces at three depths twice a week to provide information regarding soil moisture for different types of soils and terraces.

21For terrace features a detailed survey was conducted to produce a high resolution digital elevation model (DEM). On this basis a detailed mapping of the relevant hydrological element such as water-harvesting area, terrace units, and terrace intake and outlet channels was achieved to provide detailed information for the GIS-based analysis of the terraces.

Selected results of the field campaign 1998

22Precipitation in 1998 was exceptionally abundant, and total rainfall during the May-October growing period exceeded 800 mm. Local farmers reported that 1998 was the wettest year for “more than a decade”.

23The comparison between precipitation, ETpot and crop reference evapotranspiration for sorghum (ETcr, according to FAO standards) indicated in Figure 6 reflect those conditions. Although the rainfall did not reach the level of the ETpot, it is significantly higher than the ETcr. Under these conditions pure rainfeed was sufficient and no additional irrigation was necessary. Never the less this a unique snapshot and not a representative situation. Considering the wet-year conditions, the low discharge was noteworthy – it never exceeded 5% of the rainfall in any measured event. Figure 7 depicts rainfall versus soil moisture [%] and discharge [% of P]. The discharge record starts on 11 August 1998. Before that date only two floods were reported by the local inhabitants. The graph shows the rise of rainfall events beginning mid-July. The changes in soil moisture are very interesting. Except for the event on 15 May, soil moisture correlates very poorly with rainfall until 5 August. After the latter date the responses to rainfall are more noteworthy. This effect is caused by the regulated irrigation on the sample terrace. The in-flow was opened for the event of 15 May then probably closed until 5 August. From that point the inflow was left open, which caused a much more sensitive response of the soil moisture to the precipitation.

GIS based analysis of the terraces

24A detailed geomorphological map was developed on the basis of the field survey. Figure 8 shows the water-harvesting areas and the classified terrace zones. The relationship between the water-harvesting areas and the corresponding terrace units are set by the field survey. Figure 8 clearly shows that terraces adjacent to water-harvesting areas benefit more than any other terraced area. This result is still preliminary because it only describes the run-off from the water-harvesting area to the terraces but not the interaction of different terrace units. But the result does allow an estimation of the terrace factor F, which achieves values up to 4. Most zones range between terrace factor 2 and 3.

25The potential storage capacity is derived from the terrace volume, stone content of the terrace and soil texture. The terraced volume is simplified to an average soil depth on a 1m grid to facilitate data processing. Most terraces are 1-2m deep, which is considered as a reasonable value for calculation. The stone content (SC) clearly deceases from the uphill terraces (> 70 %) to the wādī bottom (< 5 %). The soil texture (according to USDA standards) varies only in the limited range from clay loam to sandy loam.

26As plant-available water (PAW), soil porosity is assumed to be proxy for field moisture capacity (FMC). The total plant available water is calculated as:

PAW = FMC * (T_Vol - SC)

27The result, shown in Figure 9, clearly classifies the terraces on the wādī bottom as those with greatest potential water storage. Water volumes available to plants reach up to 40 mm. In contrast the upslope terraces can retain less than 5 mm of water available to plants. This clear distribution seems to be determined mainly by the stone content of the terraces: stone content as much as 70 % significantly reduces the storage capacity.

Conclusion and open questions

28With the GIS-based analysis of the terraced catchment, it is possible in principle to quantify the water-harvesting–terrace irrigation systems. However, a detailed data collection of the catchment is an essential prerequisite. Potential storage volume can be computed and used as an indicator to overcome a certain dry period. Although the results presented here need further elaboration to achieve greater precision, it can be stated that even adjacent terrace areas may possess significantly different water storage capacities. These results can be used to determine the ability of different terrace zones to retain enough moisture to support agriculture during dry periods.

Figure 1. Annual average rainfall

Figure 2. Variability of annual rainfall

Figure 3. Monthly average rainfall

Figure 4. Relationship between terrace and water harvesting area

Figure 5. Cumulative precipitation and ETpot

Figure 6. Measured relationship between precipitation and potential evapotranspiration, 1998

Figure 7. Measured relationship between precipitation and soil moisture, 1998

Figure 8. Water harvesting area and terrace factor

Figure 9. Plant available water