Providing evidence to improve practice

Action: Water: Plant or maintain ground cover in orchards or vineyards Mediterranean Farmland

Key messages

Water use (3 studies): Two replicated, controlled studies (one randomized) from the USA found that plants used more water in plots with ground cover, compared to plots with bare soil. One replicated, randomized, controlled study from Portugal found inconsistent differences in water use (sometimes less, sometimes more) between plots with ground cover and plots with tilled soil.

  • Implementation options (2 studies): Two studies from Portugal and the USA found that plants used similar amounts of water in plots with different types of ground cover.

Water availability (17 studies)

  • Water content (13 studies): Four studies (three replicated, randomized, and controlled; one site comparison) from Spain and the USA found less water, or less available water in some comparisons, in soils with seeded cover crops, compared to tilled soils. Two replicated, randomized, controlled studies from Portugal and the USA found more water, or more available water, in soils with ground cover, compared to tilled soils, in some comparisons. Two replicated, randomized, controlled studies from France and the USA found inconsistent differences in water content (sometimes less, sometimes more) in soils with seeded cover crops, compared to bare or tilled soils. Three replicated studies (two randomized and controlled, one site comparison) from Chile, France, and Portugal found similar amounts of water in soils with or without ground cover. Three replicated, controlled studies (two randomized) from Chile and the USA found greater water infiltration or soil porosity in plots with seeded cover crops, compared to bare soil, but one replicated, controlled study from France did not.
  • Water loss (7 studies): Six replicated, controlled studies (five randomized) from Chile, France, Italy, Spain, and the USA found that less water was lost as runoff from plots with seeded cover crops, compared to bare or tilled plots, in some or all comparisons. One replicated, randomized, controlled study from Spain found inconsistent differences in runoff between plots with ground cover and plots with tilled soil.
  • Implementation options (5 studies): Three studies from vineyards in the USA found different amounts of water in soils with different types of ground cover, but two studies from Portugal and the USA did not.

Pathogens and pesticides (0 studies)

Nutrients (2 studies): One replicated, randomized, controlled study from Chile found less nitrogen, phosphorus, and dissolved organic carbon in runoff from plots with seeded cover crops, compared to plots with bare soil. One replicated, randomized, controlled study from the USA found similar amounts of nitrate, nitrogen, and phosphorus in runoff from plots with seeded cover crops, compared to bare soils.

Sediments (4 studies): Three replicated, randomized, controlled studies from Chile, Spain, and the USA found less sediment in runoff from plots with ground cover, compared to bare or tilled soil, in some or all comparisons. One replicated, controlled study from France found similar amounts of sediment in runoff from plots with seeded cover crops or bare soil.

Supporting evidence from individual studies

1 

A replicated, controlled study in 1984–1986 in two irrigated almond orchards in California, USA, found higher water use in plots with ground cover, compared to plots without ground cover. Water use: Higher water use was found in plots with ground cover, compared to control plots, in 10 of 12 comparisons (16–41 vs 14–32 inches of seasonal water use). Implementation options: Lower water use was found in plots with bromegrass, compared to clover or resident vegetation, in one of two orchards (27–32 vs 31–41 inches of seasonal water use). Methods: In two orchards (one newly planted, and one mature), plots with and without ground cover were compared (number and size of plots not reported). Bromegrass, clover, or resident vegetation were grown as ground cover, and herbicide was used in control plots. Water use (change in water content between irrigations) was measured with neutron probes (9–120 inches depth, five measurements/tree/plot on about 17 days during the growing season: 1984 and 1986 in the new orchard, and 1985–1986 in the mature orchard).

 

2 

A replicated, controlled study (years not reported) in an almond orchard in the Central Valley, California, USA, found that more water filtered into soils with ground cover, compared to bare soils. Water availability: More water filtered into soils with ground cover, compared to bare soils (2.2–2.6 vs 1.3 inches in four hours). Methods: There were four plots for each of three ground covers (Blando bromegrass, native vegetation, or strawberry clover) and one control (bare soil). Water infiltration was measured under the ground cover and the controls, after five years.

 

3 

A replicated, randomized, controlled study in 1989–1990 in an irrigated vineyard in the San Joaquin Valley, California, USA, found that grape vines used similar amounts of water in plots with or without cover crops between the vine rows, but more water was used in total, and more water filtered into the soil, in plots with cover crops between the vine rows. Water use: Grape vines used similar amounts of water in plots with or without cover crops between the vine rows (soil water depletion within the rows: 427–531 mm, 0–180 cm depth), but more water was used in total in plots with cover crops, in one of two comparisons (plots with winter and summer cover crops: 511 vs 351 mm water/year, 0–180 cm depth). Water availability: More water filtered into the soil in plots with cover crops (cumulative infiltration after eight hours of opportunity time/irrigation event: 106–182 vs 69–74 mm/year). Methods: There were three plots (one vine row and two interrows, 183 m length) for each of two cover crops (Bromus mollis bromegrass as a winter cover crop, treated with herbicide and mulched in summer, or followed by resident vegetation as a summer cover crop), and there were three control plots (bare soil, maintained with herbicide throughout the year). The bromegrass was seeded in January and December 1989 (and reseeded in March 1989 because of poor establishment). All plots were furrow irrigated until the water had advanced to the end of the furrow (five times in March–September 1989–1990), and thus more water was given to plots with faster infiltration (plots with cover crops). Soil water was measured with a hydroprobe (23–180 cm depth, two samples/row and two samples/interrow in each plot, before irrigation and 3–5 days after irrigation). Infiltration was calculated from water advance times along the furrow.

 

4 

A replicated, randomized, controlled study in 2000–2003 in a rainfed olive orchard near Cordoba, Spain (partly the same study as (14)), found less runoff from plots with cover crops, compared to bare fallows or conventional tillage. Water availability: Less water was lost as runoff from plots with cover crops (2.5% of rainfall; 1.3 m3), compared to conventional tillage (7.4% of rainfall; 3.8 m3) or no tillage (21.5% of rainfall; 10.6 m3). Methods: There were three plots (6 x 12 m plots, with two olive trees each, on a 13% slope) for each of three treatments: cover crops (2 x 12 m barley strips, sown in October), conventional tillage (15 cm depth, 3–4 passes from September), or no tillage (with herbicide, weed-free). Plots with cover crops were tilled before the barley was sown (10 cm depth). Runoff was collected with tipping-bucket gauges, and sediment was collected in barrels, from autumn 2000.

 

5 

A replicated, randomized, controlled study in 1998–2002 in a rainfed vineyard in southern France found that planting grass between the vine rows had inconsistent effects on water availability. Water availability: More water was found in the soil, in plots with grass between the vine rows, compared to bare soil between the vine rows, in eight of 40 comparisons (0.19–0.33 vs 0.14–0.25 m3 water/m3 soil), but less water was found in one of 40 comparisons (0.22 vs 0.27). Methods: In 1998, grass seeds (Festuca arudinacea tall fescue) were sown between the vine rows in four treatment plots, and herbicide was used to control weeds between the vine rows in four control plots (12 x 15 m plots). The grass was mown three times/year, in the summer. Water was measured every three weeks, in mid-March–August 2002, in soil cores (0–150 cm depth; two cores/plot: one under the vines, one between the vines).

 

6 

A replicated, randomized, controlled study in 1996–2000 in an irrigated vineyard in the Sacramento Valley, California, USA, found that less water was available to grape leaves in plots with cover crops, compared to bare soil, between the vine rows. Water availability: Less water was available to grape leaves in plots with cover crops, compared to bare soil, between the vine rows, in three of 16 comparisons (midday water potential: –1.22 to –0.91 vs –1.11 to –0.82). Implementation options: More water was available to grape leaves in plots that were cover cropped with barley and oats, compared to other cover crops, in three of 12 comparisons (midday water potential: –1.08 to –0.82 vs –1.22 to –0.91). Methods: There were four plots for each of four cover crops (1.8 m width, between vine rows of 3.4 m width), and there were four control plots (periodically disked between the vine rows). Each plot was 10 contiguous vines and two adjacent interrows. The cover crops were Californian native grasses (not tilled, mown), annual clover (not tilled, mown), barley and oats (mown and disked), or legumes and barley (mown and disked in spring and used as a green manure). The Californian native grasses were seeded between the vine rows in autumn 1996. The others were seeded in autumn 1997–1999. All plots were drip irrigated, fertigated (20 kg N/ha/year), and the grass cover crops were also fertilized with urea (45 kg N/ha/year). Herbicide was used under the vines. Midday water potential was measured before irrigation in June and July 1998, May 1999, and June 2000 (pump-up pressure chamber, three leaves/plot).

 

7 

A study in 1998–2002 in an irrigated vineyard in the Sacramento Valley, California, USA, found more water in soil that was cover cropped with legumes, compared to grasses, in summer, but found less water in winter. Implementation options: More water was found in soil that was cover cropped with legumes, compared to grasses, in the dry season (13% vs 6% water content), but less water was found in the wet season, after a flood (28% vs 33%). Methods: A leguminous cover crop (Trifolium fragiferum perennial strawberry clover) was planted in the southern half of the vineyard, and three native Californian, perennial, summer-dormant grasses (Elymus glaucus blue wildrye, Hordeum brachyantherum meadow barley, and Bromus carinatus California brome) were planted in the northern half. These cover crops were planted between every other vine row. They were mown 4–5 times/year and their residues were retained. The vineyard was fertigated with drip lines. Soil samples were collected in five sub-plots, in one 10 x 15 m plot, in each cover crop (0–10 cm depth, 3 cm diameter, nine times in July 2001–October 2002).

 

8 

A replicated, randomized, controlled study in 2002–2004 in a rainfed vineyard in central Portugal found similar amounts of water use in plots with or without cover crops between the vine rows. Water use: Similar amounts of water were used by vines and other vegetation in plots with or without cover crops (226–383 vs 222–357 mm/year). More water was used in plots with cover crops in one of three time-periods (budbreak–bloom: 2.1–3.3 vs 1.6–2.9 mm/day), but less was used in one of three time-periods, in one of two years (veraison–harvest: 0.83–0.89 vs 1.2 mm/day). Implementation options: Similar amounts of water were used by vines and other vegetation in plots with different types of cover crops (resident vegetation or sown grasses and legumes) between the vine rows (226–372 vs 241–383 mm/year). Methods: There were four plots for each of two cover-cropping treatments (resident vegetation or sown cover crops, both without tillage between the vine rows), and there were four control plots (with tillage between the vine rows; depth not reported). The plots were four vine rows each (100 vines/row). The sown cover crops were 60% grasses (Lolium and Festuca spp.) and 40% legumes (Trifolium spp.), sown in March 2002. The interrows of all plots were mown (treatments: twice/year, in February and May–June; controls: once/year, in February, height not reported). All plots were fertilized, and herbicide was used under the vines. Soil water content was measured between budbreak (early February) and harvest (capacitance probes, 10–100 cm depth, three samples/plot). Water use was estimated from water content and rainfall.

 

9 

A replicated, randomized, controlled study in 2003–2005 in an irrigated vineyard in Lake County, California, USA, found that less water was available in plots with cover crops, compared to tilled soil, between the vine rows. Water availability: Less water was available in plots with cover crops, compared to tilled soil, in some comparisons, in one of two years (maximum difference in leaf water potential in 2004: –1.27 vs –0.93 mPa). Methods: There were 12–22 plots (20 feet length) with cover crops (5 feet width, seeded in October 2003, each with a different species) between the vine rows (8 feet width) and 12–22 plots with tilled soil between the vine rows (2004: 22 plots; 2005: 12 plots). Leaf water potential was measured once or twice a week (11 am–1 pm, pressure bomb, one vine/plot, June–July 2004 and July–August 2005). All plots were drip irrigated (weekly for 10–12 weeks from July, 40–48 gallons/vine/year).

 

10 

A replicated, randomized, controlled study in 2000–2005 in an irrigated vineyard in the Salinas Valley, California, USA, found less water in soils with cover crops compared to bare soils, but also found that less water was lost as runoff, and runoff water had less sediment, in plots with cover crops. Water availability: In winter, less water was lost as runoff from plots with cover crops, compared to bare soils (38–96 vs 177 gallons/plot). However, less water was found in plots with cover crops, compared to bare soils, in one of two comparisons (in plots that were cover cropped with rye: 21.5% vs 23.5% soil moisture in mid-February 2003, for example, when measured between vine rows; number of significantly different comparisons not clearly reported). In the growing season, less water was found in soils with cover crops, compared to bare soils (in all years, when measured between vine rows: 17% vs 15% soil moisture in mid-May 2004, for example; or, in two of three years, when measured in vine rows: 19% vs 18% in mid-July 2004, for example; number of significantly different comparisons not clearly reported). Nutrients: In winter, similar amounts of nutrients were found in runoff from plots with cover crops or bare soils (nitrate: 1.2–2 vs 1.7 ppm; total nitrogen: 4.5–6.4 vs 5.6 ppm; total phosphorus: 1.6–2.5 vs 2.6 ppm). Sediments: In winter, less sediment was found in runoff from plots with cover crops, compared to bare soils, in one of two comparisons (with triticale as the cover crop: 508 vs 1,735 mg/litre). Implementation options: In winter, more water was found in plots that were cover cropped with triticale, compared to rye (23% vs 21.5% soil moisture, for example, in mid-February 2003; number of significantly different comparisons not clearly reported). Methods: There were nine plots for each of two treatments and one control. The treatments were triticale (X Triticosecale) or Secale cereale Merced rye, planted in November 2000–2004 as cover crops (32 inches width) between the vine rows (8 feet width), mown in spring, and disked into the soil in the following November. Bare soils were maintained in the controls through disking in spring and summer (depth not reported). Each plot had 100 vines and the adjacent areas between the vine rows. All plots were drip-irrigated in April–October. Runoff was measured with sumps (16 inches diameter, 5 feet depth) at the lower end of each plot. Soil moisture was measured with a neutron probe (3.5 feet depth). It was not clear whether these results were a direct effect of cover crops or tillage.

 

11 

A replicated, randomized, controlled study in 2001–2006 in a vineyard in the Central Coast, California, USA (same study as (12)), found more water in soils with cover crops between the vine rows, compared to tilled soils without cover crops. Water availability: More water was found in soils with cover crops, compared to tilled soils, in some comparisons (e.g., in spring: 19–21% vs 17% water), but similar amounts of water were found in most comparisons. Methods: There were six plots (84.3 x 2.4 m interrows between vines) for each of two cover crops, and there were six control plots (cultivated every two months to control weeds). The cover crops (1.8 m width) were Triticale x Triticosecale Trios or Secale cereale rye, seeded in November 2001–2005 (interrows disked before seeding), and mown in April 2002–2006. Soil samples were collected every 2–3 weeks in December 2005–November 2006 (19 samples/plot, two cores/sample, 0–15 cm depth).

 

12 

A replicated, randomized, controlled study in 2001–2006 in an irrigated vineyard in the Central Coast, California, USA (same study as (11)), found that cover crops had inconsistent effects on soil water content. Water availability: More water was found in soils with cover crops, compared to tilled soils, in some comparisons (in early spring: 20–21% vs 17% water), but less water was found in other comparisons (in late spring: vs 8–10% vs 12–14%). Methods: There were six plots for each of two cover crops (Secale cereale rye or Triticale x Triosecale Trios, sown between the vine rows in autumn, mown in spring), and there were six control plots (tilled between the vine rows every two months; depth not reported). All plots were tilled in autumn. The plots were each 84 x 1.8 m, between two vine rows. Soil samples were collected every 2–3 weeks in November 2005–2006 (two samples/plot, 0–15 cm depth).

 

13 

A replicated, controlled study in 1999 in a vineyard in southern France found less runoff from plots with grass, compared to bare soil, between the vine rows. Water availability: Less water was lost as runoff from plots with grass between the vine rows (17–45 vs 26–60 mm runoff/100 mm simulated rainfall). Similar amounts of water infiltration were found in plots with grass or bare soil between the vine rows (9 vs 10 mm). Sediments: Similar amounts of sediment were found in runoff from plots with grass or bare soil between the vine rows (2.7–4.9 vs 3.8–5.7 g soil/litre water). Methods: One interrow was cultivated (10 cm depth) and planted with grasses (without herbicide), and another interrow was chemically weeded (with herbicide: conventional management), for four months each. Rainfall was simulated in three plots (1 x 1 m plots) in each interrow (1 x 1 m plots, 60 mm water/hour, for 60 minutes). Soil samples were collected in each plot (200 observation points/m2; 5 topsoil samples/plot, 0–5 cm depth).

 

14 

A replicated, randomized, controlled study in 2000–2006 in a rainfed olive orchard near Cordoba, Spain (partly the same study as (4)), found less runoff from plots with cover crops, compared to bare fallows or conventional tillage. Water availability: Less water was lost as runoff from plots with cover crops, compared to bare fallows, in six of seven years (0.1–6% vs 3–36% of rainfall), and compared to conventional tillage, in three of seven years (0.1–0.2% vs 0.5–2.1%). Methods: There were three plots (6 x 12 m plots, with two olive trees each, on a 13% slope) for each of three treatments: cover crops (2 x 12 m barley strips, sown in October), conventional tillage (15 cm depth, 3–4 passes from September), or no tillage (with herbicide, bare fallows). Plots with cover crops were tilled before the barley was sown (10 cm depth). Runoff was collected with tipping-bucket gauges, and sediment was collected in barrels, from autumn 2000.

 

15 

A replicated, randomized, controlled study in 2001–2006 in a chestnut orchard in northeast Portugal found that more water was available to chestnut trees in plots with ground cover (without tillage), compared to plots with conventional tillage, in the driest year. Water availability: More water was available to chestnut trees in plots with ground cover, in one of four years (2005, the driest year: data reported as higher predawn water potential in chestnut leaves). Similar amounts of water were found in soils with or without ground cover (0.1–0.2 cm3 water/cm3 soil, at most depths, on most dates). Implementation options: Similar amounts of water were available to chestnut trees in plots with seeded cover crops, compared to resident vegetation (data reported as predawn water potential in chestnut leaves). Similar amounts of water were found in soils with seeded cover crops, compared to resident vegetation (0.1–0.2 cm3 water/cm3 soil, at most depths, on most dates). Methods: There were three plots for each of two treatments (no tillage with resident vegetation or grasses and legumes, sown in 2001), and there were three control plots (conventional tillage, 15–20 cm depth, thrice/year). Each plot (600 m2) had six chestnut trees (40 years old in 2001) and was fertilized but not irrigated. Soil water content was measured weekly with time-domain reflectometer probes (0–15 and 0–30 cm depth: four samples/plot; 45 and 75 cm: 2 samples/plot), in 2003–2006. Water potential was measured in June–September 2003–2006 (August–September in 2005) with gas exchangers (12 leaves/plot, south facing, up to 3 m high, 7:00–13:00 hours).

 

16 

A site comparison in 2006 in two rainfed almond orchards near Granada, Spain, found less water in soils with cover crops, compared to conventional tillage. Water availability: Less water was found in soils with cover crops (2–5 vs 5–9 g water/100 g soil). Methods: Conventional tillage (chisel plough, 20–25 cm depth, 3–4 times/year in 2001–2005, October 2005, and April and June 2006) was used in one orchard, and no tillage was used in another orchard with two cover crops (oats and vetch or oats only, sown in January 2006 on one 1 ha plot each). Both orchards were fertilized (30 t compost/ha), but the orchard with cover crops got more fertilizer (1,500 kg organic fertilizer/ha on one-third of each plot, 250 kg mineral fertilizer/ha on one-third). The orchard with cover crops had cereal-fallow rotations before the cover crops, and it was tilled in November. Soil samples were collected on 7 June and 18 July 2006 (0–20 cm depth). It was not clear whether these results were a direct effect of cover crops (and tillage), fertilizer, or site.

 

17 

A replicated, randomized, controlled study in 2005–2007 in irrigated vineyards in Sicily, Italy, found that less water was lost as runoff from plots with cover crops, compared to plots with conventional tillage (without cover crops), between the vine rows. Water availability: Less water was lost as runoff from plots with cover crops, compared to plots with conventional tillage (35–48 vs 57 mm). Methods: There were three plots (three vine interrows/plot; 2.2 x 3 m interrows) for each of four temporary cover-crop treatments (V. faba; V. faba and V. sativa; Triticum durum; or T. durum and V. sativa), each of two permanent cover-crop treatments (T. subterraneum, F. rubra, and Lolium perenne, or T. subterraneum, F. rubra and F. ovina), and each of three control plots (conventional tillage in the interrows, 3–4 times/year, 15 cm depth). Cover crops were sown in October. Temporary cover crops were tilled into the soil in April, but permanent cover crops were not tilled. The slope of the vineyard was 16%. Runoff was measured after each significant rainfall event (15 events in November 2005–October 2007) with sediment traps (Gerlach traps: 1 m diameter, 40 litres).

 

18 

A replicated, randomized, controlled study in 2008–2011 in an irrigated avocado orchard in Chile found less runoff, less sediment and nutrient in runoff, and more soil pores in plots with cover crops, compared to bare soil. Water availability: Less water was lost as runoff from plots with cover crops, compared to bare soil (0 vs 3–4 mm). No difference in water retention was found between soils with cover crops or bare soils (9–13 vs 8–13 m3 water/m3 soil), but soils with cover crops had a higher percentage of large pores (4–6% vs 3–4% macroporosity). Nutrients: Less nitrogen (0–5 vs 42–68 g/ha), phosphorus (0 vs 20–24 g/ha), and dissolved organic carbon (0–3 vs 345–637 g/ha) was found in runoff from plots with cover crops, compared to bare soil. Sediments: Less sediment was found in runoff from plots with cover crops, compared to bare soil (0 vs 1,000–3,400 kg soil/ha). Methods: Cover crops were grown in five treatment plots, and bare soil was maintained with herbicide in five control plots, in an avocado orchard, on a 47% slope (10 x 50 m plots). The groundcover (Lolium rigidum ryegrass and a legume, Medicago polymorpha) was sown in August 2008 and mown in February 2009–2010 (residues were retained). All plots were fertilized and irrigated. Runoff was collected in buried barrels downslope of each plot.

 

19 

A replicated, randomized, controlled study in 2003–2005 in eight rainfed olive orchards in southern Spain found that cover crops had inconsistent effects on runoff, but less sediment was found in runoff from plots with cover crops, compared to tilled plots. Water availability: Less water was lost as runoff from plots with cover crops, compared to tilled plots, on four of eight farms (19–56% less runoff), but more water was lost on one of eight farms (15% increase in runoff). Sediments: Less soil was found in runoff from plots with cover crops, compared to tilled plots, on seven of eight farms (63–89% less sediment). Methods: On each of eight farms, cover crops were grown (two of eight farms) or weeds were not controlled (six of eight farms) on three plots, but weeds were controlled by conventional tillage (depths not reported) on three plots (1 m2 microplots). Plots were surrounded by steel sheets, which routed the runoff into plastic containers. Runoff and sediments were measured after each rainfall event.

 

20 

A replicated site comparison in 2009 in rainfed vineyards in southern France found similar water retention in soils with or without ground cover. Water availability: Similar water retention was found in soils with or without cover crops (data on water content at field capacity not reported). Methods: In 146 plots of three soil types, there was permanent vegetation (4–22% of plots in each soil type), temporary vegetation (48–68%), or bare soil (16–42%) between the vine rows, for at least five years before soil sampling. Soil samples were collected from the interrows in March–May 2009 (10 homogenized samples/plot, 0–15 cm depth).

 

21 

A replicated, randomized, controlled study in 2008–2010 in an irrigated vineyard in the San Joaquin Valley, California, USA, found that grape vines used similar amounts of water, and soils had similar water contents, in plots with cover crops or resident vegetation between the vine rows. Implementation options: Similar amounts of water were used by grape vines in plots with cover crops or resident vegetation between the vine rows (midday stem water potential: –1.6 to –0.6 MPa). Similar amounts of water were found in soils between the vine rows with cover crops or resident vegetation, in most comparisons (soil water content: 15–34%). Methods: Cover crops were grown in the alleys (2.5 m width) between the vine rows (3.1 m width) on 16 plots (two alleys/plot, 190 vines/plot), and resident vegetation was allowed to grow on 8 plots, over the winter. There were two combinations of cover crops (oats only, or oats and legumes, seeded in November, on 8 plots each). All plots were mown in spring and tilled (15–20 cm depth) in spring, summer, and autumn. Herbicide was used to control weeds in the vine rows (50 cm width). Vines were drip-irrigated (60–70% of evapotranspiration). Soil water content was measured every 1–2 weeks, and stem water potential was measured every 2–3 weeks, during the growing season in 2008–2009 (frequency domain reflectometry, 0–110 cm depth).

 

Referenced papers

Please cite as:

Shackelford, G. E., Kelsey, R., Robertson, R. J., Williams, D. R. & Dicks, L. V. (2017) Sustainable Agriculture in California and Mediterranean Climates: Evidence for the effects of selected interventions. Synopses of Conservation Evidence Series. University of Cambridge, Cambridge, UK.