Nitrogen cycling—water use efficiency interactions in semi-arid ecosystems in relation to management of tree legumes (prosopis)**

Peter Felker, Peter R. Clark, Joseph Osborn,

and G. H. Cannell

University of California, Soil Laboratory, Riverside, California

Introduction

Evaluation of nitrogen inputs

Blue-green algal lichen nitrogen fixation

Nitrogen in rainwater

Nitrogen fixation by free-living bacteria

Nitrogen fixation by tree legumes

Water use efficiency

Estimation of biomass production from water use efficiency and nitrogen input data

Mesquite pod production

Summary and conclusions

Acknowledgement

References

Introduction

The fertility of many overgrazed semi-arid ecosystems is so low that plant water use efficiency is much below its genetically predetermined capability. As a result fertility limits plant production just as severely as water (Wight and Black, 1979). Plant productivity is still limited by water because increasing water availability increases biomass production, but the biomass increase per unit of water is lower at the lower fertility levels (Hanks and Tanner, 1952). Tree legumes probably do not experience this low fertility because they are nitrogen fixers. The genetically determined water-use efficiency of tree legumes probably is lower than surrounding grasses, because legumes are C-3 plants while grasses are C-4 plants. Nevertheless since the tree legumes are not nitrogen-limited their water-use efficiency on infertile soils under field conditions may be higher than grasses.

The planning for revegetation of semi-arid land for forage production should take into consideration the biomass production possible under nitrogen limitations, water limitations, and a combination of nitrogen-water limitations. To carry out this assessment it is necessary to know the naturally occurring nitrogen inputs and the water use efficiency of the plants on the site or projected for the site. We have reviewed the nitrogen inputs provided by: rainfall, blue-green algae-lichen crusts, non-symbiotic nitrogen fixers, and tree legumes and the ranges of water use efficiency typical of grasses and legumes. Given these nitrogen and water use inputs, the plant productivity (forage potential) is estimated for various rainfall regimes with and without exogenous nitrogen sources.

Evaluation of nitrogen inputs

An overcropped or overgrazed semi-arid ecosystem will be chosen to illustrate the nitrogen-water use efficiency interaction where it will be assumed that grazing or agricultural pressures on the ecosystem have stabilized and constant yields are obtained. Thus nutrient inputs in the form of rainfall, blue-green algal lichen crusts, and non-symbiotic nitrogen fixation are in equilibrium with nitrogen removed from the ecosystem in the form of erosion, denitrification, feed grain, forage, and livestock sold to the city. This steady state analysis removes the necessity for estimation of nutrient pool sizes in soils, plants, or animals.

Manure returned to farmers' fields has the effect of concentrating fertility in small regions, but manure causes no net nitrogen increase to the entire ecosystem as it is part of the nitrogen pool. Manure nitrogen is transformed into concentrated plant nitrogen but it cannot create a net fertility increase —only nitrogen fixers and chemical fertilizers accomplish that.

If all the nitrogen inputs into an agricultural system are known it is possible to calculate the maximum amount of nitrogen that may be removed in agricultural products. Further it is possible to estimate the total dry matter production that may be removed in agricultural products by assuming typical values for plant dry matter, protein, and nitrogen contents. To estimate the maximum dry matter production per unit of nitrogen, a low nitrogen and protein content value should be used. Inspection of food composition tables reveals that most above-ground vegetation plant parts have nitrogen contents of at least 1.5% of dry matter with corresponding protein contents (X6.25) of 9.4% (Autret, 1970).

An evaluation of the literature on each general category of nitrogen input follows.

Blue-green algal lichen nitrogen fixation

A review of blue-green algal lichen nitrogen fixation in semiarid ecosystems by Rychert et al (1978) listed five blue-green nitrogen fixation reports with nitrogen inputs in the range of 1.3 to 100 kg N ha^-1 yr^-1. Rychert et al (1978) used the maximum value for moistened and enclosed in situ acetylene reduction values extrapolated for 365 days to obtain the yearly nitrogen fixation value. In the original papers cited by Rychert et al (1978) blue-green algal crust nitrogen fixation rates were reported to be: 3–4 g N ha ^-1 ha ^-1 when corrected for actual lichen density (MacGregor and Johnson, 1971); 10 g N ha^-1 hr^-1 if the entire region was assumed to be covered by lichens (Eskew and Ting, 1978); and 10 to 80 g N ha^-1 ha^-1 by Rychert and Skujins (1974). Both and Ting (1978) and MacGregor and Johnson (1971) found low levels (1 and 0 g N ha^-1 ha^-1) for the first three hours after moistening in situ lichen crusts in contrast to Rychert and Skujins who found ethylene production five minutes after moistening the crusts. Mayland et al (1966) did not specify area measurement for their small sample assays but they calculated a nitrogen fixation of 10 kg N ha^-1 63 days^-1 which is equivalent to 14 g N ha^-1hr ^-1 assuming 12 hr days. Mayland et al (1966)'s 10 kg N/ha/63 days was based on crusts which were kept continuously moist. Crusts which were moistened every three days by Mayland et al (1966) required 80 days before the increase in total nitrogen was significantly different from the initial nitrogen content.

The extrapolation of hourly acetylene reduction assays or determination of nitrogen fixation on continuously moist lichen crusts to yearly time scales are not warranted because of the sensitivity of lichen crust nitrogen fixation to water stress. Rychert and Skujins (1974) have determined that nitrogen fixation by algal crusts drops to 1% of its maximum value when the lichen reaches a water potential of approximately 1 bar. Lichens might be expected to behave similarly to saturated soils which lose water with almost the same rate as the potential evapotranspiration until a dry crust is formed on the soil surface. Assuming an average lichen crust to be 4 mm thick (Mayland et al, 1966), and assuming a generous pore volume of 30%, only 1 mm of water would be contained in the lichen crust "root zone". Potential evapotranspiration rates of 10 mm per day are common in the southern California Colorado Desert and could be expected to dry lichen crusts to less than -1 bar in a matter of hours. Eskew and Ting (1978) moistened an in situ lichen crust with 25% of the average annual rainfall (9.3 cm) in one day and observed approximately 25 nmoles of ethylene cm² (25 g N ha^-1) for the event. This determination is probably the most realistic of all determinations and indicates nitrogen fixation by blue-green algal lichen crusts generally does not exceed 1 kg N ha ^-1 yr ^-1.

Nitrogen in rainwater

Nitrogen inputs from rainwater were determined over an eleven-year period in Riverside, California (Chapman et al, 1944). The average annual rainfall for the period was 330 mm and the average nitrite, nitrate, ammonia, and total N concentrations in the rainwater were found to be 3, 88, 91, and 181 mg per litre, respectively. The average input from all nitrogen compounds for the eleven-year period was 0.59 kg ha^-1 yr^-1. This is the only literature value we are aware of that estimates nitrogen in rainfall. Probably regions with higher rainfalls will have higher nitrogen inputs but it seems unlikely that semi-arid ecosystems can derive more than 1 kg ha^-1 yr^-1 of nitrogen from rainfall.

Nitrogen fixation by free-living bacteria

There is not as much controversy over the nitrogen fixation rates for free living bacteria as there is for blue green algae. Non-symbiotic nitrogen fixation rates were examined in a semi-arid region of California (Davis) receiving 400 mm annual rainfall where the four vegetation types, irrigated turf, fertilized and irrigated wheat field, native grassland, and a fallow soil were found to have yearly nitrogen fixation rates of 4.8 kg ha^-1, 4.0 kg ha^-1, 2.1 kg ha^-1, and 3.5 kg ha^-1, respectively (Steyn and Delwiche, 1970). These results were determined from monthly samplings using both ¹⁵N enrichment techniques and acetylene reduction. The nitrogen contents of these soils ranged from 2.5 to 5.5 mg N per gram of soil. Assuming a low C/N ratio of 6, these soils would have organic carbon contents of at least 1.5 to 3.3% and would be an order of magnitude higher than expected in Sahelian Africa (Dancette and Poulain, 1969) or desert regions of south-western United States (Tiedemann and Klemmedson, 1973). Since soil organic carbon is needed as an energy source to drive non-symbiotic nitrogen fixation and since the organic matter contents reported by Steyn and Delwiche (1970) are much higher than normal semi-arid soils, their estimate should be considered to be an upper estimate. The organic matter content of the soils in the California study was estimated by Steyn and Delwiche to be more than adequate for the nitrogen fixation rates they observed but this might not be the case where soil organic matter and nitrogen contents are 5 and 0.5 mg/g, respectively.

Vlassek et al (1973) incubated soil cores over a six-month period with acetylene to stimate nitrogen fixation in a Canadian grassland where a value of 2 kg N ha^-1 yr^-1 was found. This value is in general agreement with a 1 kg N ha ^-1 yr^-1 estimate of nitrogen fixation in another Canadian grassland study (Paul et al, 1971).

Due to limitation of water and soil organic matter, non-symbiotic nitrogen fixation in the Sahelian zones of Africa, or in the Sonora, Colorado, or Mohave Deserts is probably not greater than found in the Canadian grassland study or in the grassland native vegetation site determined in the California study, i.e., 2 kg N ha^-1 yr^-1.

Nitrogen fixation by tree legumes

Nitrogen fixation values for tree legumes have been derived either from the nitrogen contents of sustained yield harvests or from differences in the soil nitrogen contents of grassland soils that are adjacent to areas cultivated with tree legumes. Unfortunately, these determinations were conducted either in a completely rainless region (Atacama desert) where Prosopis tamarugo exists solely on the groundwater or they were conducted in regions receiving 1,000 mm which is outside the semi-arid classification.

The largest values for nitrogen fixation by tree legumes comes from an analysis of sustained yield forage productivity of Leucaena leucocephala (syn. glauca) being grown in Australia. Leucaena growth for woody biomass is optimized if grown for several years without harvest. Forage quality from Leucaena is increased at the expense of woody biomass production if the trees are harvested every several months. When grown for forage Leucaena dry matter yields of forage, protein and nitrogen were found to be 12,600 kg ha^-1, 3,600 kg ha^-1, and 575 kg ha^-1 for a 9-month growing season in Australia receiving 1,000 mm annual rainfall (Hutton and Bonner, 1960). These values were observed in a trial that had been previously grazed for three years and that had a superphosphate and potassium sulfate addition but no supplemental nitrogen or irrigation. The 575 kg of nitrogen which were removed in a nine-month period from a previously grazed field represents the minimum value for nitrogen fixation by the Leucaena.

Acacia mearnsii (syn. mollisima) trees are grown in South Africa in 8-year cycles for tannin production. When the first cycle of trees was grown on infertile soils the nodules were very abundant, but in the second and third cycles the nodule frequency decreased (Orchard and Darb, 1956). Presumably the high soil nitrogen levels present in later cycles repressed further nodule development. The soil nitrogen content was determined in the 30-year old. A. mearnsii tannin plantation and in an adjacent grassland field. The soil in the A. mearnsii plantation was found to have a soil nitrogen content 6,050 kg ha ^-1 greater than the nearby grassland which equals a 200 kg N ha^-1 yr l soil nitrogen increased (Orchard and Darb, 1956). The A. mearnsii contribution to the soil nitrogen was probably much less than the nitrogen going into plant production, and therefore a nitrogen fixation rate of 200 kg N ha ^-1 yr ^-1 by A. mearnsii is conservative. The rainfall was not stated but A. mearnsii generally is grown where the annual precipitation exceeds 900 mm.

Nitrogen fixation can be estimated for Prosopis tamarugo growing in the rainless (0.7 mm yr^-1) Chilean salt flats known as the "Pampa del tamarugal". These trees exist solely on a 3 to 6 m deep groundwater aquifer which is recharged from nearby mountains. The Chilean government has planted thousands of hectares of P. tamarugo in these salt flats for livestock fodder production. The median pod and leaf yield for 30-year-old trees are both 6,000 kg ha ^-1 yr^-1 (Salinas and Sanchez, 1971). The nitrogen content of the pods and leaves of these "tamarugo" trees have been reported to be 1.5 and 1.8%, respectively, by Pak et al (1977) and the nitrogen content for the yearly production is 198 kg ha ^-1 yr^-1. Since this plant material is fed to livestock which are later sold in the city, this value of 198 kg ha ^-1 yr^-1 is a good estimate of nitrogen fixed by P. tamarugo.

It may seem unreasonable to quote nitrogen fixation values of tree legumes at 1,000 mm annual rainfall after being so critical of effect of moisture on nitrogen fixation by algallichen crusts. The difference is that the 4 mm thick lichen crust root can only hold 1 mm of water, whereas tree legume roots may penetrate to 10 m (Meinzer, 1927) where the stored soil moisture could amount to several thousand millimetres of water. Stored soil moisture 10 cm or more below the soil surface would be much less affected by evaporative losses from solar insulation than would soil moisture in the 4-mm thick lichen crust "root zone".

It is difficult to extrapolate the nitrogen fixation that would be possible in rainfed semi-arid regions receiving 200–500 mm annual rainfall from ecosystems receiving more than 1,000 mm annually or from ecosystems with unlimited groundwater. However, an upp er estimate of nitrogen fixation can be obtained using empirically derived coefficients. If the annual rainfall was 500 mm, the soil water recharge 75% of rainfall, the water use efficiency 1,000 kg of water per kg dry matter, the dry matter required per gram of nitrogen is 7 g (Herridge and Pate, 1977), and if 50% of the dry matter were used for nitrogen fixation, then 267 kg of nitrogen and 1,875 kg of dry matter could be produced per hectare per year. Most plant dry matter has nitrogen contents in the range of 1.5% and therefore only 28 kg of N would be required for the 1,875 kg of dry matter.

Nitrogen fixation values for tree legumes in one area with groundwater and in two areas with precipitation over 1,000 mm were at least 200, 200, and 550 kg N ha ^-1 yr^-1, respectively. We predict nitrogen fixation rates on the order of 100 kg N ha^-1 yr^-1 probably are possible in areas receiving 500 mm annual rainfall.

Water use efficiency

Two important features distinguish the water use efficiency of tree legumes from other annual crops.

For annual crops the distribution of the rainfall throughout the growing season may be much more important than the average annual rainfall. Tree legumes have deeper root systems which tend to integrate rainfall throughout the season, thus making short-term fluctuations relatively unimportant. We are conducting a water use efficiency study with four varieties of Prosopis in the California Imperial Valley that illustrates the independence of tree legumes over rainfall distribution patterns. This site has a potential evapotranspiration of 2400 mm for the 9-month Prosopis growing season and maximum temperatures of 37.7°C and 43.3°C for 102 and 13 days, respectively. A 450 mm preirrigation was applied 45 days before transplant, 100 mm was applied with the transplanting operation, and two 280 mm irrigations were applied the rest of the season. The dry matter productivity for the best variety was 12,600 kg ha^-1 (Felker and Cannell, unpublished). The second year's water application is now being carried out with a single irrigation applied before the onset of hot weather to maximize water infiltration. We expect this single irrigation to adequately provide for the second year's growth.

A second feature which distinguishes trees receiving infrequent irrigation or rainfall from well-irrigated annual plants is that transpiration from the trees cannot be estimated either from potential evapotranspiration or from meteorological data. These trees use the water as it becomes available and then go into a stage of limited metabolic activity until more water becomes available. These plants may have xylem water potentials as low as -45 bars (Mooney, 1977) and have been termed arido-active plants by Fisher and Turber (1978).

Since the tree legume dry matter production is not correlated with a summation of daily potential evapotranspiration, but is related to the yearly annual precipitation input, it is justifiable to estimate yearly dry matter production using empirical water use efficiency coefficients and the annual precipitation.

Water use efficiency in terms of grammes of water required to produce a gramme of dry matter has been examined under conditions of high fertility by various workers. In a classic study, Briggs and Shantz (1914) measured water use efficiency in 15 kg sealed earthen pots of various cultivars of corn, wheat, oats, sorghum, 15 species of legumes, and grasses, forbs and shrubs native to Colorado. Water use efficiency for cultivars of the C-4 plants corn and sorghum, ranged from 220 to 400 kg H₂O per kg dry matter and from 571 to 935 kg H₂O per kg dry matter for 14 legume species. Cowpea (Vigna sinensis) was the most efficient (571 kg H₂O per kg dry matter) while purple vetch (Vicia atropurpurea) was the worst with 935 kg H20 per kg dry matter. More recent work using gas exchange measurements under Australian field conditions confirmed this with a value of 305 to 340 kg H₂O/kg DM for fertilized grass production and 700 kg H₂O/kg dry matter for legumes (Ludlowand Wilson, 1972). Field studies which measured dry matter accumulation per unit of water applied also found water use efficiencies of 238 kg H₂O/kg DM for maize (Alessi and Power, 1976).

Under non-fertilized range conditions a very different picture emerges because of the dependence of water use efficiency upon plant nutritional status. Hanks and Tanner (1952) examined the water use efficiency on fertilized and unfertilized Wisconsin soils. The water consumed was measured by determining the difference in stored soil moisture between harvest and planting after correction for precipitation and deep drainage. The water use efficiency for fertilized plots expressed in kg H₂O/kg DM averaged 143% higher for corn and 174% higher for oats than the unfertilized control in a 2year field trial. A ten-year range fertilization study in Montana (Wight and Black, 1979) determined the water use efficiency by measuring precipitation and soil moisture with neutron probe equipment. The 10-year average dry matter yields for unfertilized and fertilized plots were 1057 and 2325 kg ha^-1 and the average water use efficiency for unfertilized and fertilized range was 3846 kg H₂O/kg DM and 1721 kg H₂O/kg DM. These workers observed a yield plateau of 1250 kg ha ^-1 which was obtained in some of the wettest and driest years, e.g., 439 and 263 mm plant available water, respectively. This suggests that water was not limiting plant productivity above 1250 kg ha^-1. These water use efficiencies are still much below that determined for many of the dominant plants of the Montana site such as Bouteloua gracilis, Artemesia fiigida, and Agropyron smithii for which Briggs and Shantz (1914) reported water use efficiencies of 377, 474, and 1076 kg H₂O/kg DM, respectively, under conditions of good fertility. The increase in water use efficiency observed by Wight and Black (1979) probably would have been greater if the Montana soils had lower fertility levels comparable to Sahelian type soils. For example, prior to fertilization, the Montana soil had organic matter and nitrogen contents of 2.5 and 0.3%, respectively, whereas typical Senegalese (West African) millet farming regions receiving 500 mm annual rainfall have organic carbon and nitrogen contents of 0.27 and 0.03%, respectively (Dancette and Poulain, 1969).

Two Prosopis water use efficiency studies, Dwyer and DeGarmo (1970) and McGinnies and Arnold (1939) found low water use efficiencies for Prosopis. Dwyer and DeGarmo (1970) reported a water use efficiency for shoot production of 4,400 kg H₂O kg DM while McGinnies and Arnold reported a value of 1700 kg H₂O/kg DM. The Prosopis used by these workers probably came from New Mexico and Arizona where the studies were conducted. Under 2 uniformly irrigated field trials at the University of California, Riverside, with 30 Prosopis accessions, accessions from New Mexico and Arizona had 1/6 to 1/30 the biomass productivity of Prosopis accessions from South America (Felker et al, submitted). All these Prosopis had the same water availability, and therefore the accessions with the greater biomass productivity could be expected to have substantially greater water use efficiency. We believe advanced South American Prosopis strains will have the same range in water use efficiency as that determined by Briggs and Shantz (1914) for annual legumes (e.g., 570 to 935 kg H₂O/kg DM).

Estimation of biomass production from water use efficiency and nitrogen input data

Having reviewed inputs of nitrogen and water use efficiency it is possible to calculate (estimate!) biomass productivities and pod yields that are possible with and without the use of tree legumes. The values that will be used for these calculations have been discussed earlier in this paper and are listed in Table 1.

Table 1.Summary of coefficients used for biomass production calculation

(Substantiations for these coefficients are provided in the text.)

Assuming that 25% of the mean annual precipitation is lost by runoff and soil evaporation (Doorenbos and Pruit, 1975) for a precipitation efficiency of 75%, 188 to 375 mm of water is available to support plant growth in the 250 to 500 mm annual rainfall regime. At 250 mm annual rainfall under conditions of high fertility C-4 plants should be able to produce 8545 kg DM ha ^-1 yr^-l to 4700 kg DM ha ^-1 yr^-l and legumes should be able to produce 3298 to 2010 kg DM ha^-l yr^-1 depending upon the water use efficiencies reported in Table 1. If nitrogen inputs from all sources other than tree legumes are considered, i.e., 4 kg N ha ^-1 yr^-1 only 267 kg DM would be possible at any rainfall using the coefficient of 1.5% N in dry matter. If a nitrogen fixation value of 100 kg N ha^-1 yr^-1 by tree legumes is used 6667 kg DM ha^-1 yr^-l would be possible and this would be adequate for plant dry matter production in regions receiving up to 500 mm annual rainfall using the highest water use efficiency for legumes, i.e., 570 kg H₂O kg DM^-1. Under the scenario of no nitrogen fixing legumes, water is not nearly as limiting as is nitrogen, and C-4 plants with their higher photosynthesis capacities and water use efficiencies would have lower biomass productivities than legumes.

Biomass productivities of legumes can be estimated to be in the 2,000 to 3,300 kg DM ha^-1 yr^-1 range at 250 mm annual rainfall and 4,000 to 6,600 kg DM ha 1 yr 1 at 500 mm annual rainfall. Fruit or seed yield can be estimated from total dry matter production in annual crops using empirically determined harvest index coefficients. Harvest index can be defined as percentage of dry matter in the fruit divided by the total yearly dry matter production. Values of 20-40% are common in annual plants but little data is available for trees. Wallace et al (1951) determined the dry matter partitioning into fruits, blossoms, leaves, and twigs for 14-year-old citrus trees. Over a 2-year period encompassing a light and heavy alternate bearing citrus crop, the total reproductive effort into blossoms, fruit loss, and fruit accounted for 74.5% of total dry matter into fruit, leaves, and twigs. Growth of the main trunk and large branches were not included in the Wallace et al (1951) study but are not expected to be large in a mature (14-year-old) citrus tree. The 75% figure is not comparable to annuals because it does not include all above-ground productivity, but it does indicate large dry matter partitioning into fruiting structures. In order to achieve a 4,000 kg ha 1 pod yield goal at 500 mm annual rainfall, a 61% harvest index would be required if dry matter accumulations of 6,600 kg ha^-1 yr^-1 were achieved. In light of the high partitioning of citrus dry matter into reproductive structures this does not seem unreasonable.

To ensure these rates of nitrogen fixation, water use efficiency, and dry matter accumulation, other mineral nutrients are also needed. The potassium content of mesquite pods have been reported to vary between 1.0 to 1.2% (Becker and Grosjean, 1980) requiring an addition of 48 kg K ha^-1 yr^-1 to replace that removed in 4,000 kg of pods. Phosphorus levels of mesquite pods are not known but are probably one-tenth of the nitrogen content. Since mesquite pods are 12.5% protein (2% N) the phosphorus content of the pods would be 8 kg. Soil sulphur levels below 5 ppm have been shown to limit legume growth in Australian field trials (Probert and Jones, 1977). An amendment of 50 kg of sulphur ha^-1 would achieve a 5 ppm concentration to a 1 m depth and probably adequately provide mesquite sulphur needs. The minerals P, K, and S are stable, not readily leached, and probably could be provided in one application every 10 years. Assuming a slight safety margin, 100 kg P, 500 kg K and 50 kg S would be required every 10 years for a U.S. cost (1979 dollars) of $63, $119, and $5, respectively.

Mesquite pod production

Having concluded a hypothetical prediction of mesquite biomass production, pod yield data will be examined for three field plots in southern California. The oldest field plot located in Riverside has 25 randomly replicated accessions and was established in June 1977. This plantation was initially conceived as a nursery so that the trees are widely spaced (4.5 × 6.1 m) and in the summer the trees are irrigated every three weeks by furrow irrigation.

A second planting, established in July 1978 at Riverside included 32 accessions representing 12 species. Twelve trees of each accession were grown in three different basins where water was supplied when the soil water potential reached 0.6, 2 or 5 bars. A third planting was established in March 1979 in the California Imperial Valley to screen for biomass production, pod production, and to conduct a water use efficiency study. The Riverside climate is cooler (daily July maximum of 34.6°C) than the Imperial Valley climate (daily July maximum of 41.7°C) with the result that mesquite growth is nearly half as rapid in Riverside as it is in Imperial Valley (Felker et al, 1980).

In the summer of 1978 one pod with a mass of 2.5 grams was produced from the Prosopis planted in Riverside. In 1979 Prosopis in their third season produced 30.8 kg of pods and those in their second season produced 2.4 kg. In Imperial Valley, 15 of the 1300 six-month-old trees flowered and at least six produced pods. This early pod production is extremely important because it shows potential for achieving economic pod yields at early ages. Early flowering would also reduce the generation time in breeding and crossing studies and greatly facilitate transfer of genetic characters between accessions.

Pod yield data for the 21/2 year old Riverside trees, for the 11/2 year old Riverside trees and for the 1/2 year old Imperial Valley trees are presented in Tables 2, 3 and 4. The consistently high pod production for several mesquite selections from Arizona, and nearby Mexico, e.g., P. velutina (0020), P. spp. (0032) and P. spp. (0025) are particularly noteworthy. As exists with most mesquite characters propagated from seed, the pod production among the early pod producing lines is quite variable. In the oldest plots the accession with highest average pod production (P. velutina (0020), had trees with no pods as well as those with 4.8 kg (10.61b). In fact, four of the 110 trees in the oldest plots at Riverside produced over 50% of the pod production for the plot. There is no obvious positive correlation between plant size and pod production since the three heavy pod producing accessions are considerably smaller than the large South American accessions. P. chilensis and P. alba.

Table 2. Mesquite pod yield for trees at end of third growing season at Riverside

During the summer months, these accessions are irrigated every 3 weeks by furrow irrigation. Plot yield previous year was 1 pod or 2.5 g. Moisture content is approximately 10%.

Table 3. Mesquite pod production for second growing season at Riverside

Only 5 of 27 varieties at Riverside produced pods second growing season. There are 12 possible pod producing trees per moisture treatment on a 4 × 4 ft spacing. The basin size is 12 × 16 ft (5.2 × 10^–3 acre).

Table 4. Mesquite flowering and pod set six months after transplant in Imperial Valley

These measurements were made 23 August, 1979, when most of the pods were quite immature. As these trees are a 6 hr round trip drive from Riverside, it was not possible to collect mature pods because they fall off at maturity and the rabbits eat them. There were 16 possible trees/accession that could have produced pods in the biomass section and 8 in the pod character section.

A germplasm collection trip was made in 1979 seeking to find pods high in sugar which have potential for fermentation to ethanol. Pods were evaluated in the field by tasting them for sugar. In one location of wild trees in the southern California Coachella Valley, an exceedingly sour and a very sweet tree were collected within 100 m of each other. These contrasting tasting pods were sent to the USDA Western Regional Research Center for analysis where the sugar contents for the pods minus seeds were found to be 40.8% and 13.4% for the sweet and sour pods, respectively. The sugar content for entire sweet pod was 36.5%. We now have the capability to mechanically separate the seeds from the pods (see below) and it is possible to obtain a dry 40% sugar flour after passage through a seed thresher and cleaner. This flour is easily transportable as long as it is kept dry.

Manual separation of mesquite seeds from pods is a long, arduous task. We have constructed a device reported by Texan workers (Flynt and Morton, 1969) to be capable of separating mesquite seeds from pods. The starting unit costs $300 to purchase and required $130 of shop work to complete. This device does separate seeds from pods, but approximately 30% of the seeds are destroyed.

We have found that "heavy duty" home meat grinders with coarse blades separate seeds from pods with less seed damage, but unfortunately they have burned-out motors, broken-off gears, etc. We are having a stronger meat grinder fabricated from a commercial meat grinder head (Hobart No. 12) and an industrial 1/3 hp, 170 rpm electric gear motor.

A six-month-old small tree (75 cm tall) was observed in the Imperial Valley plots with 30 blossoms that eventually had at least three pods. The wild parent of this tree was rather small in being approximately 3.3 m tall and 6 m wide, but was heavy with pods two years in a row. Because of the small size of these trees and the early age at which they produce pods, it seems reasonable that a high plant density (about 670 trees/ha) of small trees (3 m tall and 3 m wide) should be capable of producing larger yields in shorter time than larger more widely-spaced trees. For example, the 3 m tall and wide UCR trees which produced 4.8 kg of pods would have produced 3,201 kg/ha^-1 if planted 3 m apart in the row and 5 m between rows (667 trees/hectare). A 5 m between row spacing would allow 1.5 m for each of the trees to protrude into the row, and 2 m for small vehicular traffic between rows. While the 4.8 kg of pods/tree were obtained in the third growing season under frequently irrigated conditions, data from the differential irrigation treatment plots (Table 3) indicate that approximately half that yield might have been obtained under near dry land conditions. Accordingly, it seems that a yield of 6 kg/tree/year might be obtained in 5 years under near dry land conditions and this would result in 4,000 kg/ha^–1/ year at an age of 5 years.

As previously reported (Felker, 1979), we have obtained as much as 50 kg of dry pods from an 8 m tall, 10 m wide mesquite tree with multiple 25 cm diameter trunks. If trees were planted to achieve this ultimate size they would have at least 10 m spacings for a density of 100 trees ha^-1. An early pod yield of 6 kg/tree would only produce 600 kg ha^-1 at the 10 × 10 m spacing whereas the 3 × 5 m spacing would produce 4,000 ha^-1. This illustrates the advantage of high-density, small, early-bearing trees.

Our concept of pod harvest technology is that the pods would be allowed to fall to the ground when they reached maturity. If labour were as inexpensive and as abundant as exists in many developing countries, the fallen pods could be raked into piles and loaded into bags, hoppers, or trucks. If mechanical harvesting is desirable the pods probably could be windrowed with a modified hayrake and then picked up with a mechanical device. The major difficulty we expect in mechanical harvesting is development of devices to prune the trees all the way to the trunk. We have hand-pruned 200 trees that are being examined for pod production.

Prosopis pod production is seasonal with the pods ripening and falling off the tree in approximately one month. This will require pod collection, storage, and rationing of pods to animals throughout the year. This requires more effort than grazing animals but as we have seen the greater pod productivity should outweigh the difficulties in collecting and feeding pods to livestock. In connection with preparation of a "State of the art on Acacia albida", (Felker, 1978) the senior author interviewed ten peanut farmers in Senegal regarding Acacia albida and found that 9 out of 10 farmers collected and stored

Acacia albida pods for use later in the year. Two of these farmers stored 500 kg of pods to feed to beef cattle. Mesquite pods are presently collected in Mexico where they are ground and mixed with other rations for use as a livestock feed (Lorence, 1970). In 1965 mesquite pod sales in Mexico's commercial sector were estimated to be 14 million pesos (1.1 million dollars) (Lorence, 1970). This collection, storage, and rationing of tree legume pods in West Africa and Mexico illustrates that it is an acceptable social concept for developing country farmers.

Summary and conclusions

All sources of nitrogen input into semi-arid ecosystems have been evaluated and it is concluded that all non-legume sources combined contribute no more than 4 kg N ha^-1 yr^-1. Leguminous trees have been shown to fix from 200 to 550 kg N ha^-l yr^-1 and we postulate that a nitrogen fixation rate of 100 kg N ha^-l yr^-l is possible at 500 mm annual rainfall. Water use efficiencies by legumes and non-legumes have been reviewed where ranges of 220-400 kg H₂O kg DM^–1 and 570-930 kg H₂O kg DM^-1 have been found for C-4 plants and legumes, respectively. By taking these water use efficiencies, nitrogen inputs, and the low fertility of semi-arid soils into consideration, we predict that a sustained dry matter removal no greater than 300 kg N ha^-l yr^-1 is possible without legumes because of nitrogen limitations. We also predict that sustained dry matter removals in the 2,000 to 6,000 kg N ha^-1 yr^-1 range are possible with nitrogen fixing tree legumes. Assuming a high harvest index of 60%, a 4,000 kg N ha^-l pod yield goal is predicted for Prosopis.

The sustained dry matter yield of 300 kg N ha^-1 yr^-1 without legumes is in good agreement with continuously cropped dry matter yields in Senegal. Gillier (1960) found the yield of continuously cropped peanuts without fallowing or fertilizer amendments was 355 kg/ha and Dancette and Poulain (1969) found sorghum yield outside A. albida canopy cover to average 460 kg ha^-1 at 500 mm annual rainfall. Dancette and Poulain (1969) also demonstrated that the yield of an annual legume such as peanuts could be increased if grown beneath a tree legume. As discussed elsewhere (Felker, 1979) we believe the peanut yield increases under Acacia albida trees illustrate the advantages deeply rooted tree legumes have in overcoming inhibitory effects of transient moisture stress on nodulation and nitrogen fixation. If non-nitrogen fixing trees, shrubs or grasses are to be grown without nitrogen fertilization they should be grown in mixed plantations with nitrogen fixing trees such as Prosopis or Acacia.

Field data on pod productivity shows great variability between and within accessions probably because of outcrossing in Prosopis. A maximum of 4.8 kg of pods has been produced from a single 2.5 m tall tree in the third growing season at Riverside. Some accessions, which have produced pods as early as six months after transplant display remarkable promise for earliness in pod production. Dwarf Prosopis trees in 3 × 5 m spacings yielding 667 trees per hectare are suggested to achieve high yields at an early age (5 years). Prosopis orchards will require more attention than pastureland, as the pods from Prosopis will have to be collected, stored and rationed to animals for most effective utilization. We predict the greater effort required for tree legumes will raise the sustained capability for dry matter production from 300 kg N ha ^-l yr^-l to 3,000 kg N ha^-1 yr^-l and greatly increase the livestock producing capability of the land.

Acknowledgement

The financial support of the U.S. Department of Energy Grant No. ET-78-G-01-3074 is gratefully acknowledged.

References

Alessi, J., and Power, J.F. (1976). 'Water use by dry land corn as affected by maturity class and plant spacing'. Agron. J. 68: 547–550.

Autret, M. (1970). 'Amino acid content of foods'. Rome, FAO Nutr. Studies 24.

Becker, B. and Grosjean, O.K.K. (1980). 'A composition study of pods of two varieties of mesquite (Prosopis glandulosa, P. velutina) '. J. Agri. Food Chem. 28: 22–25.

Briggs, L.J. and Shantz. H.L. (1914). 'Relative water requirements of plants'. J. Agric. Research 3(1): 1–64.

Chapman, H.D., Liebig, G.F. and Rayner, G.F. (1944). 'Nitrogen gains and losses'. Hilgardia 19: 57–128.

Dancette, C. and Poulain, J.F. (1969). 'Influence of Acacia albida on pedoclimatic factors and crop yields'. African Soils 14: 143–184.

Doorenbos, J. and Pruitt, W.O. (1975). 'Crop water requirements'. FAO Irrigation and Drainage Paper 24.

Dwyer, D.D. and DeGarmo, H.C. (1970). 'Greenhouse productivity and water use efficiency of selected desert shrubs and grasses under four soil-moisture levels'. New Mexico State Univ Agric. Exp. Sta. Bull. 570.

Eskew, D.L. and Ting, I.P. (1978). 'Nitrogen fixation by legumes and blue-green algal lichen crusts in a Colorado desert environment'. Amer. J. Bot. 65:850–856.

Felker, P. (1978). 'State of the Art: Acacia albida as a complementary permanent intercrop with annual crops'. AID

Publications, Washington, D.C. Free of charge to developing countries through their Embassy.

Felker, P. (1979). 'Mesquite: An all purpose leguminous arid land tree'. In: New Agricultural Crops, G.A. Ritchie (ed.), AAAS Symp., Vol. 38, Westview Press, Boulder, Colorado, pp. 89–132.

Fischer, R.A. and Turner, N.C. (1978). 'Plant productivity in the arid and semi-arid zones'. Ann. Rev. Plant Physiol. 29: 277–317.

Flynt, T.O. and Morton, H.L. (1969). 'A device for threshing mesquite seeds'. Weed Sci. 17: 302–303.

Gillier, P. (1960). 'La reconstitution et le maintien de la fertilité des sols du Sénégal et le probleme des jachéres'. Oleagineux 15: 637–704.

Hanks, B.J. and Tanner C.B. (1952). 'Water consumption by plants as influenced by soil fertility'. Agron. 44: 98–100.

Herridge, D.F. and Pate, J.S. (1977). 'Utilization of net photosynthate for nitrogen fixation and protein production in an annual legume'. Plant Physiol. 60: 759–764.

Hutton, E.M. and Bonner, I.A. (1960). `Dry matter and protein yields in four strains of Leucaena glauca (Benth)'. J. Austral. Instit. Agric. Sci. 26: 276–277.

Lorence, F.G. (1970). `Importancia economia de los mezquites (Prosopis spp.) en algunos estados de la Republica Mexicana'. In: Mezquites y Huizaches, Ediciones del Instituto Mexicano de Recursos Naturales Renovable A.C. Mexico D.F., pp. 1–69.

Ludlow, M.M. and Wilson, G.L. (1972). 'Photosynthesis of tropical pasture plants. IV. Basis and consequences of differences between grasses and legumes'. Aust. J. Biol. Sci. 25: 1133–1145.

MacGregor, A.N. and Johnson, D.E. (1971). 'Capacity of desert algal crusts to fix atmospheric nitrogen'. Soil Sci. Amer. Proc.35:843–844.

Mayland, H.F.; McIntosh, T.H. and Fuller, W.H. (1966). `Fixation of isotopic nitrogen on a semi-arid soil by algal crust organisms'. Soil Sci. Soc. Amer. Proc. 30: 56–60.

McGinnies, W.G. and Arnold, J.F. (1939). 'Relative water requirements of Arizona range plants'. Technical Bull. No. 80, Arizona Agric. Exp. Sta.

Meinzer, O.E. (1927). 'Plants as indicators of groundwater'. U.S. Geological Survey, Water Supply Paper 577, 95 pp.

Mooney, H.A. (1977). In: Mesquite-its biology in two desert ecosystems. B.B. Simpson (ed.) Stroudsburg, Dowden, Hutchinson and Ross, Inc. PA, pp. 30–38.

Orchard, E.R. and Darb, G.D. (1956). 'Fertility changes under continued wattle culture with special reference to nitrogen fixation and base status of the soil'. 6th Int. Soil Sci. Cong., Paris 4: 305–310.

Pak, N. et al. (1977). 'Analytical study of tamarugo (Prosopis tamarugo) an autochthonous Chilean feed'. J. Sci. Fd. Agric. 28: 59–62.

Paul, E.A.; Meyers, R.J.K, and Rice, W.A. (1971). 'Nitrogen fixation in grasslands'. In Biological nitrogen fixation in natural and agricultural habitats. Plant and Soil, Special Volume, pp. 495–507.

Probert, M.E. and Jones, R.K. (1977). 'The use of soil analysis for predicting the response to sulphur of pasture legumes in the Australian tropics'. Aust. J. Soil Res. 15: 137–146.

Rychert, R.C. and Skujins, J. (1974) 'Nitrogen fixation by blue-green algae-lichen crusts in the Great Basin desert'. Soil Sci. Soc. Amer. Proc. 38: 768–771.

Rychert, R. et al (1978). 'Nitrogen fixation by lichens and free living microorganisms in deserts'. In: Nitrogen in Desert Ecosystems, N.E. West and J.J. Skujins (eds.) Stroudsburg, PA, Dowden Hutchinson and Ross, Publ., pp. 20–30.

Salinas, H.E. and Sanchez, S.C. (1971) 'Informe Tecnico' No. 38, Santiago, Chile, Instituto Forestal, Seccion Sivicultura, 35 pp.

Steyne, P.L. and Delwiche, C.C. (1970) 'Nitrogen fixation by nonsymbiotic microorganisms in some California soils'. Environ. Sci. & Technol. 4(12): 1122–1128.

Tiedemann, A.R., and Klemmedson, J.G. (1973). 'Nutrient availability in desert grassland soils under mesquite (Prosopis juliflora) trees and adjacent open areas'. Soil Sci. Soc. Amer. Proc. 37: 107–111.

Vlassak, K., Paul E.A. and Harris, R.E. (1973). 'Assessment of biological nitrogen fixation in grassland and associated sites'. Plant and soil 38: 737–644.

Wallace, A., Cameron, S.H. and Mueller, R.T. (1951). 'Seasonal use and loss of nutrients and dry matter in orange trees'. Amer. Soc. Hort. Sci. 58: 5–10.

Wight, J. R. and Black, A.L. (1979) 'Range fertilization: Plant response and water Use'. J. Range Manage. 32: 345–349.