Trees, water and salt An Australian guide to using trees for healthy

catchments and productive farms

Edited by Richard Stirzaker, Rob Vertessy and Alastair Sarre

Trees, water and salt: an Australian guide to using trees for healthy catchments and productive farms

Edited by Richard Stirzaker, Rob Vertessy and Alastair Sarre Production editor: Martin Field Design and layout: Design One Solutions

©Copyright Joint Venture Agroforestry Program 2002

This work is copyright. Except for the Joint Venture Agroforestry Program and CSIRO logos, graphical and textual information in this publication may be reproduced in whole or in part, provided that it is not sold or put to commercial use and its source is acknowledged.

Such reproduction includes fair dealing for the purpose of private study, research, criticism or review as permitted under the Copyright Act 1968. Reproduction for other purposes is prohibited without the written permission of the Joint Venture Agroforestry Program or the Rural Industries Research and Development Corporation.

Photographs supplied by the authors, as well as Sharon Davis, Murray-Darling Basin

Commission, CSIRO Land & Water, State Forests ofNSW Images on p.l12 & p.l28 copyright© State Forests ofNSW Images on the cover, and p.6, p.7, p.ll, p.l4, p.27, p.28, p.36, p.37, p.lOl, p.140 & p.l41 copyright© CSIRO Land and Water.

This report presents the results of a project funded by the Joint Venture Agroforestry Program. However, the Joint Venture does not necessarily endorse or support the findings or

recommendations presented herein unless expressly stated by the Joint Venture in writing.

ISBN 0 642 58308 0 RIRDC publication number: 01/086

What this chapter is about Woodlots are generally established on land previously used for agriculture,

and following the harvest of trees, the land may be returned to agriculture.

Land used for agriculture generally has more stored soil water and higher

nutrient status. Thus, the initial productivity of a woodlot established on

such land may be relatively high, particularly in the lower rainfall zones

where water would otherwise be limiting. However, productivity may

decline as the stand develops and consumes these resources, even to the

point of tree mortality due to drought.

This chapter explores the performance of woodlots in areas considered

too dry for conventional forestry. It examines the implications of climate,

soil, salt and groundwater for the length of rotation required to achieve a

balance between groundwater control and tree productivity. And it

suggests a strategy in which woodlots can be moved around the farm to

'mine' soil water, thereby increasing the impact on recharge.

Setting the scene The rotation of crops with perennial pastures such as lucerne - phase farming -is sometimes promoted as a means of reducing leakage. Phase farming with trees

is conceptually very similar, except that trees can extract water from subsoils that are chemically and physically very hostile. Lucerne, for example, does not survive in acidic subsoils and is only moderately tolerant of salty soils.

This chapter addresses the questions:

1) how long will woodlot productivity be supplemented by stored soil water from the preceding agricultural system?

2) how long will it take after woodlot establishment for groundwater recharge to be minimised? and

3) how long following woodlot harvest will recharge control persist without

replanting trees?

The answers to these questions will help establish the optimum lengths of each phase and the risk of drought-induced mortality. However, we have few or no

observations of real systems in the many geographic regions for which we might desire this kind of land use, so this chapter relies on simulation modelling.

Woodlots are defined in this book as any tree-planting configuration for which the edge effects on the supply of water and light are minimal when averaged over the whole stand. The productivity of a woodlot is determined largely by rainfall and the water-holding capacity of the soil (assuming that the most appropriate species for the site have been selected). Figure 4.1 gives an example of how the productivity of a woodlot decreases with declining rainfall. The economic viability of woodlots depends on a range of factors , including productivity, but as rainfall declines, the growing of woodlots becomes an increasingly risky venture. Drought deaths observed in blue gum (E. globulus)

woodlots in Western Australia bear testimony to this.

When a site is marginal for woodlots, the store of water and nutrients left over from prior agriculture may have a large impact on productivity. The implications for a second rotation immediately following the harvesting of a woodlot are clear: the benefits inherited from agriculture and reflected in the first rotation's productivity will not carry over into the second rotation. The challenge is to devise an agroforestry system that makes best use of the growth enhancement phase, optimises recharge control and is economically viable.

500

-"' E 400 -0 .. Ill

300 Gl >. .... Ill Gl E 200 :I 0 >

"'1:1 Q 100 Q

3=

0

400

•

600

•• • • •

800

Average annual rainfall (mm)

• •

1000 Figure 4. 1: The productivity of woodlots decreases along a rainfall gradient. At some point, the woodlot becomes financially (and possibly physically) unviable. Redrawn from Wong et al. (2000).

The modelled systems Most observations on plantation performance are for sites receiving more than 600 mm annual rainfall, while data on tree performance and water use in drier country are mostly for silvicultural systems other than woodlots. Thus, we apply a

numerical ecohydrological model (WAVES, incorporating carbon and allometric algorithms from another model called 3-PG) to predict the hydrological and growth responses to a range of soil and climatic scenarios. We have some confidence that this model captures the essential behaviour and processes under consideration. This approach allows an analysis of the sensitivity of the phased system to a variety of soil and climatic conditions as well as for periods of simulation that greatly exceed the normal length of experimental observation.

The U04.VES model was used to examine four scenarios (see box below) using the climate record of the town of Merredin, a dry location about 250 km east of Perth . The climate there is mediterranean, with 70% of the 320 mm average annual rainfall falling in the six months from April to September. The average

potential evaporation is nearly 1 800 mm/yr. The scenarios contrast high and low water-holding soils and the impact of fresh and saline groundwater.

The modelled scenarios

• Scenario 1: a weakly duplex soil witl1 sand to 2 m over a sandy loam 3 m deep.

• Scenario 2: a strongly duplex soil with 1 m of sand over 9 m of clay.

• Scenario 3: 1m of sand over clay with a fresh watertable at 2m.

• Scenario 4: same as in Scenario 3, except that when the trees use groundwater it is replaced by saline groundwater of 2 7 000 EC (approximately half the salinity of seawater).

For each simulation:

• the initial soil water and salt profile was established by simulating a 30-year agricultural phase;

• the winter wheat crop/summer pasture phase was modelled for 30 years, followed by a 30-year tree rotation, which was followed in turn by 30 years of crop/pasture; and

• the same 30-year climate record was used for each phase.

For the first two scenarios, the soil was allowed to drain to a very deep groundwater system beyond the reach of tree roots. For scenarios 3 and 4 the watertable stayed at a depth of 2 m regardless of rainfall and groundwater consumption.

The crop and pasture physiological parameterisations were based on extensive testing at the modelled sites and we are confident in their calibrations. The tree

physiological parameterisation was based on data from a site growing E. globulus in Victoria under dry conditions; we are less confident in the absolute values of wood production in our simulations.

The water balance results are intended to reflect the general sensitivity of hydrological responses to the influences of soil type and climate. The modelling is not meant to be a precise or accurate estimate of the performance of woodlot

systems at specific localities. We assumed no chemical or physical constraints to rooting that would limit the biological potential of the plants. We also assumed that the entire simulated soil profile was potentially available to tree roots. The

possible impacts of external risks such as pests, disease or fire were not considered.

Modelling resu Its

Scenario 1: 2 m sand over 3 m sandy loam

The average leakage over the initial

3 0 years of crops and pastur es was 21 mm/yr. H owever, there was consid­erable variability between years, with leakage of 10 mm or less in eight years

and greater than 50 mm in five years (Figure 4.2). Such variability was driven by the amount of annual rainfall (in the wettest year this was 481 mm, in the

driest it was 178 mm) and by its timing, with heavy winter rains contributing most to leakage.

Trees effectively prevented leakage after one year of growth and the LAI came

into equilibrium with contemporary rainfall within 2-5 years (Figure 4.3). The

tree biomass was 68 t/ha after 30 years, giving a mean annual increment (MAl) of 2.3 t/ha/yr. Following a return to the

crop/pasture phase, leakage was back to pre-woodlot levels within three years.

Figure 4.3 shows that there is no clear period under this scenario in which the mining of the soil water allows the tree to develop a higher LAI than we would expect from rainfall alone. In fact, the LAI responds to annual variation in rainfall from as early as year 4. Thus, a phased woodlot-crop/pasture system in this environment would have a very short rotation time - about three years of cropping followed by three years of trees.

Crops/pasture Trees Crops/pasture 800 1.0

0.8

'E 600 >:: .. !. ~

"' 0.6 E

bO ; !. .. 400 - Storag~ ... 0 - Leakag$ "' ..... 0.4 bO

"' .. .... ·c; ..

"' I;>

200 ..... 0.2

0 0.0

0 30 60 90

Tim e (years)

Figure 4.2: The storage of water in the soil (blue line) and the daily recharge (red line) under Scenario I. A crop/pasture system was in place for the first 30 years, followed by a 30-year woodlot and a return to the crop/pasture system in year 60. This soil can only store about 120 mm of water and frequent leal<age events occur during the crop/pasture phase. Leal<age occurs only during the first year of the tree phase. The soil store is rapidly drawn down to the lower limit (500 mm) by the trees and is filled again within three years of the tree harvest.

600 - Rain (mm)

- Tree leaf area index

e 4 >< "' !.

"C 400 ·=

] .. "' ...

c .. ·;;; -... 2

.. .S! -;;; 200 "' :I

c "' ... c 1-<C

o+-- ----.- - ---.- ---.--- -.-------.- - -t o 0 10 . IS 20 25 30

Time (years)

Figure 4.3: Woodlot LAI and rainfall over the middle 30 years under Scenario I. There is no clear period of enhanced growth as a consequence of mining the soil water; LAI responds to fluctuations in annual rainfall.

Scenario 2: 1 m of sand over 9 m of clay at Merredin

Average leakage under crops and pastures on this deep, heavy-textured soil was 10 mm/yr, half that predicted under Scenario 1. Leakage didn't respond to individual wet and dry years, but showed a slow response to a sequence of wetter or drier-than-average years (Figure 4.4).

Leakage ceased within two years of tree­planting as the large soil store of water was drawn down . As the trees used this water, tree LAI - and hence productivity - developed quickly (Figure 4.5). However, LAI had returned to a level dictated by annual rainfall after 5-7 years. Tree biomass was 3 3. 9 t!ha after 3 0 years (MAl = 1.2 t!ha/yr).

T he soil profile took 15 years to refill after woodlot harvest . Leakage started very slowly and had not reached the pre-woodlot rate by year 90.

Tree growth in this environment 1s relatively poor, but the thick clay layer creates effective drought protection. A rotation length of 15-2 0 years for crop/pasture and five years for trees would minimise r echarge and take advantage of the most productive period of tree growth.

Crops/pasture Trees Crops/pasture

2600 , r 0.10 - Storage

e 2400 Leakage

0.08 >:: !. "' .,

Ji! 2200 j 0.06 E c !. ·;;; :.. ..

2000 0.04 1).0 -;;; "' :I .... c "' .. c ... cc 1800 0.02

1600 0.00 0 30 60 90

Time (years)

Figure 4.4: The storage of water in the soil (blue line) and the daily recharge (red line) under Scenario 2. A crop/pasture system is in place for the first 30 years, followed by a 30-year woodlot and a return to the crop/pasture system in year 60. This soil can store about 300 mm of water and leakage responds slowly to changes in vegetation. Leakage is prevented within two years of woodlot establishment. After a return to crops and pasture in year 60, the soil store slowly refills. Leakage commences after about 15 years but has not reached pre-woodlot levels by year 90.

600 ' - Annual rainfall

- Tree leaf area index

4 )( e .. ., E ·= ._. 400 3 "' .. Ji! :..

"' c -·;;;

"' :.. 2 ~ -;;; .. 200 1\ .. :I

:.. c 1-

c cc

0 0

0 10 20 30

Time (years)

Figure 4.5: Woodlot LAI and rainfall over the middle 30 years under Scenario 2. There is a clear period of enhanced growth as the trees mine the soil water, after which LAI comes into equilibrium with annual rainfall.

Scenario 3: 1 m of sand over clay with a fresh watertable at 2 m

Leakage on this soil type averaged

34 mm/yr, higher than in scenanos 1 and 2 because there was only 2 m of unsaturated soil above the watertable.

The short periods in which leakage fell below zero (Figure 4.6) occurred when water moved from the groundwater into the root zone of the crops or pastures after they had dried out the unsaturated zone. However, the net flow of water was

downwards.

Trees effectively prevented groundwater recharge after 1-2 years of growth (Figure 4. 7), after which water was taken in large amounts (up to 300 mm/yr)

from the watertable, which dropped by about 0.4 m per year. The fresh groundwater beneath the woodlot allowed the trees to continue growing at rates much faster than would be possible on

contemporary rainfall alone, reaching an LAI of about 4 after ten years. Tree biomass was 160.5 t/ha after 30 years

(MAl= 5.4 t!halyr).

The use of groundwater by the woodlot resulted in a local drop in the watertable. However, the model assumed good

connection between local and regional water levels, so the impact oflocal changes in leakage was influenced by an area much larger than the plot. The result is that when the woodlot was harvested at

year 60 the watertable returned to a depth of 2 m within two years.

With less well-connected groundwater

systems, local rises would be expected under cropping, leading to possible declines in production due to high water-

tables . With trees, the local drawdown of watertables could be more substantial and both extraction and tree

growth may decline as a result. Thus, the transmissivity of the aquifer system would ultimately determine the optimal lengths of the tree and crop phases.

1.0 Crops/pasture Trees Crops/pasture

0.5

~· · · ·· · >: "' "C

E 0.0

!. .. t>C -0.5 "' ~ "' .. i ,,~j.o .....

-1.0

·~ i

-1.5 0 30 60 90

Time (years)

Figure 4.6: Under Scenario 3, water generally leaks to the groundwater during the two cropping periods (years 0-30 and 60-90) , although there are short periods of upflow from the groundwater into the unsaturated soil. The trees rapidly prevent recharge after year 30 and then start using water from the watertable. After the trees are harvested the watertable rises quickly because the aquifer is assumed to be highly transmissive. Note: negative leakage values indicate groundwater uptake.

- Annual rainfall

500 - Tree leaf area index

'E 400 4

!. 300

~ c ·;;

200 ... 2

"iii = c 100 c

c:c

0 +----,~---.----.-----,----.-----+0

0 5 10 IS 20 25 30

Time (years)

Figure 4.7: Woodlot LAI and rainfall over the middle 30 years under Scenario 3. The LAI continues to increase over the first ten years as the trees make use of the fresh groundwater.

Scenario 4: 1 m of sand over clay with a saline watertable at 2 m

The first 30-year crop/pasture phase yielded the same leakage result -34 mm/yr - achieved under Scenario 3 (Figure 4.8). Leakage to groundwater was prevented within two years of tree­planting. Water was initially taken up in large amounts from the watertable, which dropped by about 1 m over the first five years.

Under this scenario, the initially fresh water consumed by trees was replaced by salty water at 27 000 EC. LAI increased rapidly as the trees consumed the store of fresh groundwater, but the subsequent build-up of salt in the profile began to limit tree growth (Figure 4.9; Figure 4.10 shows the dynamics of salt build-up under the woodlot). LAI reached its peak around the fifth year but by the twelfth had come back into equilibrium with contemporary rainfall, with 0.5 mm/yr leakage. Tree biomass was 66.6 t/ha after 30 years (MAl = 2.2 t/ha/yr).

The salt build-up reduced the amount of groundwater used by the trees, so the watertable eventually rose to its original level. This affected the subsequent cropping phase, which now had to grow in a more saline environment. Average leakage consequently rose from 34 to 43 mm/yr.

1.0

0.5 >:: "' ::!;! E 0.0

.§_ .. b.O -0.5 "' ~ .. ~~~

..... -1.0

- 1.5 0 30 60 90

Time (years)

Figure 4.8: Leakage over the first 30-year crop/pasture phase under Scenario 4 is the same as for Scenario 3 (Figure 4.6). Trees initially use a large amount of groundwater but this declines after about five years as salt builds up in the profile. Leakage in the second rotation is about 25% greater than in the first because of salt in the subsoil. Note: negative leakage values indicate groundwater uptake.

- Annual rainfall

500 - Tree leaf area index

400 4 >< 'E ..

"C

.§_ ·= 300 3 "' Ji! .. ... "' c ...

·~ 200 2 "' ... ~ ii .. :I .. c ... c 100 ..... <

0 0

0 5 10 IS 20 25 30

Time (years)

Figure 4.9: Woodlot LAI and rainfall over the middle 30 years under Scenario 4. The enhancement of LAI is shortlived because salt builds up in the profile.

Figure 4.10: (a) the build-up of salt in the soil profile under trees responsible for woodlot decline; and (b) the subsequent leaching of this salt back towards the groundwater under cropping/pasture after woodlot harvest.

Figure 4.10 shows the dynamics of salt build-up under the plantation and the subsequent leaching of salt during the cropping phase. Salt builds up in the soil above the watertable because the trees transpire fresh water and prevent most of the salt from entering their roots. This process is discussed in more detail in Chapter 7. After ten years there is so much salt in the first metre of soil above the watertable that further groundwater uptake is very low.

Or-----------------~W~o-od~lo't

I .c 'Q. 2

"' Q

0 10 20 30 40 50

Soil water salinity (dS/m)

0 ,---------------~~~-, 2nd cropping

I 1 2 "' Q

10 20 30 40 so

Soil water salinity (dS/m)

The salt slowly leached out during the subsequent cropping phase, but even after 30 years a considerable amount remained. As a consequence, the next rotation of trees would not enjoy the same initial growth rates.

The conditions contained in Scenario 4 are far more common in Australia than those in Scenario 3. In our modelling, a salty profile reduced crop yields by 5-25% and increased recharge by 25%, while the tree rotation suffered a significant productivity decline 5-10 years after establishment. Without an effective drainage strategy incorporating a leaching fraction under trees, woodlots over shallow, saline watertables are unlikely to have a sustainable long-term role. Cropping is also not the answer: a leaky system doesn't cause the local build-up of salt but it does lead to rising groundwater and hence to an increase in the area of salt-affected and waterlogged soils. On the other hand, salt build-up will be much less of a problem when the woodlot phase is kept as short as possible- say, about five years.

General principles The key to phase farming is to get the lengths of the tree and crop phases roughly right. The length of the enhanced tree phase is largely determined by the depth to which tree roots can penetrate as well as the water-holding capacity of the soil. In Scenario 1, the trees rooted to 5 m in a soil of moderate water-holding capacity.

There was no pronounced increase in tree growth and thus using woodlots in rotation with agriculture is probably not justified. In contrast, tree productivity was greatly enhanced over the deep, heavy-textured soil in Scenario 2. However, it is unlikely that the enhancement phase would last more than a couple of years, and the contribution this would make to the overall profitability of the enterprise is unknown.

The optimal length of the cropping phase is determined by the amount of water the tree phase has removed from the subsoil and the rate at which the subsoil refills under cropping. Since the cropping phase is likely to be more profitable than the tree phase, it is desirable to have it extend for as long as possible. Again, the water storage capacity is lower in shallower soil and the leakage rate under

cropping greater in soils with lower water-holding capacity. The cropping phase could be less than three years (eg Scenario 1) or greater than 30 years

(eg Scenario 2).

Thus, there are two potential benefits from short-rotation woodlots. First, the stored water may provide enhanced growth or at least reduce the mortality of trees in the early years. More importantly, the trees are removed once their

hydrological task has been completed and cropping can continue for some time

with minimal leakage.

4 >< ..

"C

·= OS .. ... OS - 2 OS .. ...

2 4 6

Time (years)

- Crop/pasture

Tree

8 10

Figure 4.11: The LAI of crops/pasture and trees. Crops attain a much higher LAI in winter and produce more product for existing markets, but .trees use more water.

The simulations reveal some interesting

insights into the different growth patterns of annual and perennial plants. Figure 4.11 shows the simulated LAI for the

crop/pasture system and the first ten years of a tree rotation for the sanie climate sequence used in the scenarios described earlier. Compared to trees,

the annuals attained a much higher LAI in winter, while the reverse occurred in summer. The consequence of these different patterns of growth is that it is

almost impossible for annual plants to use all the rainfall. Paradoxically, the perennial that uses more water is less productive, at least in economic terms. This will only be changed by the

development of new markets for short­rotation tree products; some, such as reconstituted wood, biomass and charcoal, are beginning to emerge. Machinery

for efficiently harvesting and removing the stumps of such crops must also be developed.

From our simulations, we can distill some general principles for woodlot design in aid of groundwater recharge control:

1) very shallow or very sandy profiles require almost continuous protection

by trees. In such areas, the benefits of initial tree productivity following a period of cropping/pasture are minimal and not worth the risk to

groundwater associated with any period of cropping;

2) deeper profiles with heavy-textured

subsoil offer the opportunity to alternate woodlots with crops and pasture to achieve groundwater

recharge control and improved tree productivity. The ultimate length of woodlot rotations beyond the enhancement period devolves into a decision based on economic and silvicultural considerations, given that the hydrological benefits accrue early in the woodlot phase. The risk of drought death

increases once woodlot LAI reaches equilibrium with contemporary rainfall and should be considered when determining rotation length; and

3) woodlots over fresh watertables can be very productive and sustainable with or without a crop/pasture phase. Woodlots on saline watertables are much more problematic and risky, with the likelihood of decreasing productivity and even death as salt builds up in the root zone. The time required for salt

profiles to recede after woodlot harvest may be long. Under these conditions, the woodlot phase must be kept short.

How to use this knowledge Phase farming with woodlots is a strategy that can potentially achieve hydrological control through the incorporation of trees in traditional farming enterprises. The key biophysical design considerations are:

a) the degree to which recharge must be controlled to protect the land or stream (see Chapter 3);

b) the specific considerations of soil texture, depth and groundwater quality; and

c) the minimum length of the tree phase of the rotation required to achieve a forest product.

From these considerations, one can calculate the total area of land within a farm

or other unit of land that should be occupied by trees at any one time to achieve the desired hydrological outcome. Woodlots can then be moved around the land unit in sequence. The key to this agroforestry strategy is the ability of woodlots to mine soil water and thereby to have a lingering hydrological impact after

harvest. With suitable soils, such a movable feast may require significantly less land to achieve the equivalent protection offered by continuous forestry.