Developing Guidelines for Agriculture Development in Northern Australia

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion The case of the Gilbert/Etheridge River Integrated Agriculture Project By Andrew Brooks and John Spencer Acknowledgements

Thanks to Northern Gulf NRM group for their efforts facilitating this research, in particular, Richard Musgrove, Andrew Taylor, Ricky Archer and Chairman John Bethel – who also generously showed us around his property and provided us with hospitality while in the area. Thanks also to various landholders who provided access to their properties, including Campbell Keough, Reg and Beverley Pedracini, and Russell and Cheryl Ryan.

This project received funding from the Australian Government’s National Environmental Research Program.

Cover photographs

Front cover - Andrew Brooks inspecting the erosion hazard potential of the soils within the Gilbert development area. Photo - John Spencer

Back cover - Exposed tree roots within an alluvial gully in the Gilbert catchment, highlighting the susceptibility of this landscape to erosion. Photo – Andrew Brooks© Charles Darwin University 2015

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion: the case of the Gilbert/Etheridge River Integrated Agriculture Project is published by Charles Darwin University and is for use under a Creative Commons attribution 4.0 Australia licence. For licence conditions see: https://creativecommons.org/licenses/by/4.0/

The citation for this publication is as follows:

Brooks, A. and Spencer, J. (2016). Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion: the case of the Gilbert/Etheridge River Integrated Agriculture Project. Darwin: Charles Darwin University.

For further information contact Andrew Brooks [email protected]

ISBN 978-1-925167-46-7

Printed by Uniprint

Charles Darwin University, Darwin Northern Territory, 0909, Australia. Contents

1 Executive summary �� 1 1.1 Context �� 1 1.2 Key findings �� 2 1.2.1 Gully extent �� 2 1.2.2 Gully erosion rates �� 2 1.2.3 Gully erosion sediment yields �� 2 1.2.4 The IFED proposal �� 3 1.2.5 The FGARA soil suitability analysis �� 4 1.2.6 Landscape analysis of the IFED area �� 5 1.2.7 Soil erosion hazard in the IFED area �� 6 1.2.8 Buffer design to reduce the threat of gully erosion �� 7 1.2.9 Offset opportunities �� 8 2 Background �� 10 3 Project overview �� 11 4 Project objectives �� 12 5 Structure of this report �� 12 6 Review of 2006 Griffith University gully mapping �� 13 6.1 Regional scale remote sensing gully mapping �� 13 6.2 Reassessment of 2006 alluvial gully erosion hazard mapping �� 15 6.2.1 Google Earth gully mapping �� 16 6.2.1.1 Summary of gully mapping Ground-Truthing �� 18 6.3 Resurvey of previously surveyed gullies using Differential GPS and Terrestrial LiDAR �� 19 6.3.1. Gully Erosion rate Analysis - 2006 – 2014 – Abingdon gully complex 1 (ABGC1) Bel Bel Crossing...... �� 20 6.3.2 Gully Erosion rate Analysis - 2006 – 2014 Gully complex 1 G9A GC2 – �� 21 6.3.3 Terrestrial LiDAR scanning of gully systems �� 22 6.3.4 Summary of rates data �� 25 6.4 Soil suitability and the CSIRO FGARA Report �� 26 6.4.1 Soil suitability assessment �� 26 7 The IFED proposal �� 32 7.1 Overview of proposal �� 32 7.2 Geological and topographic context for the IFED site �� 33 8 Field reconnaissance of the IFED agricultural precinct �� 37 8.1 Current gully extent �� 38 8.2 Soil chemistry and gully hazard �� 38 8.3 Identify the alluvial stream and river channel network at highest available resolution and identifying appropriate buffers �� 43 8.4 Potential offset areas �� 48 9 References �� 49

Figures

Figure 1 Example of recent clearing for agriculture in the Gilbert catchment adjacent to highly gullied land (yellow scale bar is 200m in length) Figure 2 Alluvial gully complex in the Greenhils area adjacent to the Gilbert River. Specific sediment yields more than 3000t/ha/yr have been measured from wihin this gully complex Figure 3 Terrestrial LiDAR scanner image of a highly active alluvial gully in the Gilbert catchment �� 1 Figure 4 DEM map showing the location of the proposed IFED agricultural area (outlined in yellow) within the Dismal Creek catchment above the apex of the Gilbert megafan. Note that the area is far from flat, located as it is on the tertiary age deposits of the upper Gilbert megafan, which is largely disconnected from upstream sources of “new” alluvium �� 3 Figure 5 Map of the study area and the Proposed IFED area which shows that almost the entire area encompassed by the IFED agricultural area falls within the bounds of the area “known to predict poorly” �� 4 Figure 6 DEM and cross valley sections through the Dismal Creek catchment area encompassed by the proposed IFED area showing the highly developed stream network within this area and the cross-valley topographic variability �� 6 Figure 7 Proposed stream network buffers for the Dismal Creek catchments area �� 8 Figure 8 Map showing the aggregation of alluvial gullies in the area around the confluence of the Einasleigh and Etheridge Rivers and on the Gilbert River in the Greenhills areas. Note the gully mapping shown here is the earlier remote sensing-based mapping which has been shown to over represent the extent of gullies in the areas of high gully concentration. Hence the actual extent of gullies is less than represented here, but we do know that there are some major gully complexes here that could be managed with appropriate investment �� 9 Figure 9 Map showing the catchments in northern Australia which either have mapped distributions of the extent of alluvial gully erosion or where alluvial gullies are known to exist but have yet to be mapped. Catchments shown in light grey �� 11 Figure 10 Map of the Gilbert River Project area (outlined in red), which encompasses the area likely to be developed for irrigated agriculture, and within which erosion studies have been conducted. Also shown is the coverage of the various datasets that have been used in this study �� 14 Figure 11 Image showing the focal study area (orange outline) with the location of the major infrastructure for the IFED project indicated (i.e. 2 off-stream water storage lakes; diversion channels (red); and the proposed irrigation area (yellow outline)). The yellow polygons scattered across the image are the gully areas as mapped from Aster imagery in 2006. Also shown are the centre pts of ground photos collected in 2014 as part of the ground validation exercise for the gully mapping �� 14 Figure 12 Locations of gully polygons with on ground field visits (316 photos were taken amongst 55 gully polygons) �� 15 Figure 13 Map showing the distribution of the 0.1 degree grid with the sample grids highlighted in orange �� 16 Figure 14 Graph showing the relationship between Google Earth mapped gully area (per grid square) compared with ASTER derived gully area (per 1/10th degree grid square) �� 18 Figure 15 Map showing the location of sites at which detailed gully resurveys were carried out �� 19 Figure 16 Comparison between differential GPS gully scarp surveys completed in 2006 and 2014 �� 20 Figure 17 Change detection procedure based on 2008 LiDAR data from the BelBel crossing site �� 21 Figure 18 Change detection procedure based on 2008 LiDAR data from the G9A GC2 site near Greenhills �� 21 Figure 19 DGPS gully surveys 2006 & 2014 from site G9A on the upper Gilbert River (indicated in pink on Figure 14) �� 22 Figure 20 Image of the surrounding broader alluvial gully complex at Greenhills showing the gully segment for which a high resolution 5cm DEM was derived from a Leica C10 terrestrial LiDAR. The yellow box highlights the scanned section of the gully complex �� 23 Figure 21 Gully G9BGC1 showing some of the scanned images derived from the scans of the area highlighted in Figure 18. �� 23 Figure 22 5cm pixel resolution DEM of the scanned gully at Greenhills showing the area of gully extension in the 8 year period from October 2006 to October 2014. Of the total sediment exported from this gully over this period, 44.8% was derived from headscarp extension and the remaining 55.2% from erosion of internal gully surfaces �� 24 Figure 23 Elevation change image which shows the distribution of erosion from both gully headscarp retreat and erosion of internal gully slopes. Note that there is little net change from the gully floor during this period, which reflects the transport limited nature of alluvial gullies (sensu Rose et al., 2015) �� 25 Figure 24 Ternary image of merged regional radiometrics datasets for the Flinders and Gilbert catchments (outlined as black polygons) and their surroundings. The merged radiometric datasets are presented as a ternary (red, green, blue) image, with potassium specified in the red channel, thorium in the green channel and uranium in the blue channel (Figure 3.4). This image indicates significant variability in the alluvial cover across the two catchments, reflecting a varied source for the materials present. (source CSIRO 2012 fig 3.4) �� 27 Figure 25 Map showing the distribution of field soil samples used in the digital soils mapping exercise for the FGARA digital soils mapping exercise �� 27 Figure 26 Broad geological units described in the CSIRO FGARA reporting. Note the Karumba formation depicted in yellow, which coincides with the Holroyd Plain unit mapped in the original Qld Govt geological mapping of the region (Grimes and Douch, 1978) , which is the original depositional surface of the palaeo Gilbert magefan �� 27 Figure 27 Physiographic map of the Gulf plains from Grimes and Doutch (1978) showing the location of the Holroyd Plain which is the unit that is synonymous with the Karumba Formation as represented by the CSIRO mapping (Figure 26) �� 28 Figure 28 Example of the digital soils suitability mapping produced for the FGARA �� 28 Figure 29 Maps from the FGARA digital soils mapping which provide estimates of the confidence in the mapping outputs. Note the apparent contradiction between the two maps, where map (a) indicates high to very high confidence in the map predictions in the area immediately above the fan apex, whereas map (b) indicates that because of the nature of the geological province (Karumba province or Holroyd Plain – depending on which geologic map you use) that it is likely that the radiometric data will give spurious results in this area hence leading to inaccurate soil suitability predictions �� 30 Figure 30 Blow up of the Gilbert Fan area depicted in Figure 29 which shows that the IFED agricultural precinct is located within the area that is depicted as having low predictive power using the digital soils mapping approach �� 31 Figure 31 Map of the proposed IFED development showing the water infrastructure and the location of the agricultural precinct. Shown here are the proposed off-stream dams which will extract around 40% of the mean annual high flows from the Einasleigh River at Dagworth (550,000 Ml per annum), where it will be stored in Dagworth dam and transferred to Dismal lake on Huonfles Station via an open channel. The Dismal lake storage is also fed by the local Dismal Creek catchment, which is a significant catchment in its own right. Also shown are the ~1600 x 40 ha plots that will be the focus for the agricultural precinct on Kutchera & Chadshunt Stations. Source http://www.agriculture.gov.au/ abares/outlook-2014/Documents/presentation-slides/keith-deLacy-presentation.pdf �� 32 Figure 32 Digital elevation model (DEM) of Cape York and the Southern Gulf (LHS) showing the IFED irrigation area (black rectangle) within the upper reaches of the Gilbert megafan, and within the Plio/ Miocene age Wyaaba beds (RHS) - which are some of the oldest alluvial sedimentary deposits within the Gilbert megafan. These units are actively incising and have been weathered in-situ for more than 2.6 million years �� 33 Figure 33 A DEM of a section of the upper Gilbert Fan showing the highly dissected topography encompassed within the IFED irrigation area. Also shown is the highly dissected nature of the drainage network with the irrigation area. The Irrigation area falls within the Dismal Creek catchments, which is a relatively steep, well developed catchment in its own right, set within the broader context of the Gilbert Fan. As can be ascertained from this image, this area is not a flat alluvial plain that one might expect in the typical irrigation areas found on the lower Burdekin or the Murray Darling �� 34 Figure 34 Cross sections of the Dismal Ck catchment showing the considerable relief (> 20m) between the ridges separating consecutive creek lines. Cross section locations are shown on the 30m DEM �� 35 Figure 35 Longitudinal profiles of the major creek lines running through the Dismal Ck catchment showing the relief encompassed within the irrigation area (indicated in red ), and the relatively steep gradient of the streams within the irrigation area (~ 0.14 – 0.18%) �� 36 Figure 36 Map showing the distribution of existing gullies (red polygons) within the Dismal Creek catchment. Most of these existing gullies are alluvial gullies that are located within or near to drainage lines in the area �� 37 Figure 37 Location of the sampled gully sites within the study area �� 39 Figure 38 Gully soil sample sites within the IFED irrigation area �� 39 Figure 39 Base saturation associated with sub-surface soils at each of the sampled gully sites. These data indicate that no one soil chemistry indicator can be used as a predictor of soil erodibility. High ESP is generally thought to represent erodible soils, whereas a high Ca/Mg ratio is thought to indicate a low susceptibility to erosion. These results raise questions about the validity of these assumptions in this landscape �� 40 Figure 40 Base saturation % values for sample sites �� 40 Figure 41 Pictorial representation of the soil dispersion tests carried out at selected sites within the IFED irrigation area. Some of these samples are so dispersive that the tests, which are usually run for 22 hours, were complete within 30 secs. The most erodible samples (GIL10216) come from a gully that has no outlet connected to a stream network, and yet is evidently highly active. Suspended sediment must be being removed via overland flow that passes through and over the gully...... �� 41 Figure 42 Ephemeral stream channels like this one in the Gilbert catchment can contribute significant volumes of sediment to the channel network, and further disturbance to such channels from clearing, road crossings etc., could significantly increase this sediment input. �� 43 Figure 43 Plot showing the proportion of the IFED area that would need to be left out of production to provide adequate buffers ...... 45 Figure 44 Example of two buffering options within the IFED area (20x stream order (darker colour) & 45 x stream order). Yellow box is a 40 ha plot for scale �� 46 Figure 45 Blow up showing the 20m buffer and the 45m buffer. The yellow box represents a 40 ha plot �� 46 Figure 46 Blow up of a section of the Kutchera ag area show the proposed 45m buffer (i.e. 90m for 1st and 2nd order stream lines). The box here represents the size of the 90m pixels used in the FGARA soil suitability analysis. �� 47 Figure 47 Map of the whole IFED showing the proposed minimum buffer (45m x stream order; with 1st order streams = 2nd streams). For scale a 40 ha plot is shown in the centre of the image. �� 47 Figure 48 Map showing the aggregation of alluvial gullies in the area around the confluence of the Einasleigh and Etheridge Rivers and on the Gilbert River in the Greenhills areas. Note the gully mapping shown here is the earlier remote sensing-based mapping which has been shown to over represent the extent of gullies in the areas of high gully concentration. Hence the actual extent of gullies is less than represented here, but we do know that there are some major gully complexes here that could be managed with appropriate investment. �� 48

Tables

Table 1 Areas of mapped (minimum) active gully erosion across key northern Australian catchments 10 Table 2 Datasets that have been reviewed during this project and which have formed the basis for subsequent analyses of erosion potential undertaken during this study �� 13 Table 3 Summary statistics of the “ground-truthed” gully polygons �� 16 Table 4 Summary statistics for the Google earth gully mapping test data �� 17 Table 5 Table showing the extent of gullies in the test blocks represented in both the ASTER derived gully layer and the Google Earth mapped gully layer �� 17 Table 6 Table showing extent of over/under representation of gully area ASTER derived gully layer and the Google Earth mapped gully layer �� 17 Table 7 Change in total are of the gully at BelBel Crossing on the Einasleigh River 2006-2014 �� 21 Table 8 Change in total area of the gully at site G9A 2006-2014 �� 22 Table 9 Summary statistics showing the volume of erosion �� 24 Table 10 Suitability classification applied in the FGARA �� 29 Table 11 The four soil stability categories based on K-factor used in this study (Rosewell and Loch, 2002) �� 29 Table 12 Effect of different buffer widths on the proportion of land taken out of production. �� 45

NERP Emerging Priorities Project – Final Milestone Report – August 2015

Project: Developing Guidelines for Agriculture Development in Northern Australia to avoid and reduce accelerated erosion: the case of the Gilbert/Etheridge River Integrated Agriculture Project

Host Institution: Charles Darwin University

Project Leader: Andrew Brooks, Griffith University

Project Team: John Spencer, Griffith University

Purpose: This project will use the Gilbert-Einasleigh region of Queensland as a case study to demonstrate how new irrigation developments can be planned and implemented to minimise the threat of developing new alluvial gully erosion or exacerbating existing erosion (both associated with alluvial gullies and stream bank erosion).

Figure 1 Example of recent clearing for agriculture in the Gilbert catchment adjacent to highly gullied land (yellow scale bar is 200mFigure in length) 1 Example of recent clearing for agriculture in the Gilbert catchment adjacent to highly gullied land (yellow scale bar is 200m in length).

Figure 2 Alluvial gully complex in the Greenhils area adjacent to the Gilbert River. Specific sediment yields more than 3000t/ha/ yr haveFigure been 2 Alluvial measured gully fromcomplex wihin in thethis Greenhills gully complex area adjacent to the Gilbert River. Specific sediment yields of more than 3000 t/ha/yr have been measured from within this gully complex

Report Citation:

Brooks, A.P., and Spencer, J. (2015). “Developing Guidelines for Agriculture Development in Northern Australia to avoid and reduce accelerated erosion: the case of the Gilbert/Etheridge River Integrated Agriculture Project”, Final report for National Environmental Research Programme (NERP) Emerging Priorities Theme, Griffith University, pp. 58.

Acknowledgements: Thanks to Northern Gulf NRM group for their efforts facilitating this research, in particular, Richard Musgrove, Andrew Taylor, Ricky Archer and Chairman John Bethel – who also generously showed us around his property and provided us with hospitality while in the area. Thanks also to various landholders who provided access to their properties, including Campbell Keough, Reg & Beverley Pedracini and Russell & Cheryl Ryan.

1 Executive Summary

1.1 Context

There has been considerable interest in recent years as to the potential for irrigated agriculture 1 development within the Gilbert River region. To date two proposals for establishing irrigated agriculture in the Gilbert catchment have been proposed; the Gulf Savannah Development proposal focused around the pockets of good1 qualityExecutive alluvial soils summary along the Gilbert River between Greenhills and Chadshunt Stations; and a second larger proposal known as the Integrated Food and Energy Development proposal focused on Kutchera and1.1 Chadshunt Context Stations, in the Dismal Creek catchment bounded by the Gilbert, Etheridge and Einasleigh Rivers.There hasTo beendate considerable more attention interest hasin recent been years directed as to the at potential IFED due for irrigatedto the scaleagriculture of the proposal and the high profile naturedevelopment of its within proponents. the Gilbert TheRiver IFEDregion. project To date has two alsoproposals attracted for establishing more criticism irrigated from the local agriculture in the Gilbert catchment have been proposed; the Gulf Savannah Development proposal community becausefocused around of its thescale pockets and ofits good radical quality departure alluvial soils from along accepted the Gilbert practiceRiver between by focusing Greenhills on areas of low quality soils supplementedand Chadshunt Stations; by large and scale a second fertigation larger proposal (i.e. knownirrigation as the with Integrated all plant Food nutr and ientsEnergy provided within Development proposal focused on Kutchera and Chadshunt Stations, in the Dismal Creek catchment the supplied irrigationbounded by water).the Gilbert, For Etheridge this reason and Einasleigh we have Rivers. chosen To date to morefocus attention on the has IFED been case directed study to explore some of the landat IFED management due to the scale challenges of the proposal faced and throughthe high profile implementing nature of its suchproponents. a project The IFEDin this landscape. The project has also attracted more criticism from the local community because of its scale and its radical study focal area,departure boun fromded accepted by the practiceEinasleigh, by focusing Gilbert on andareas Etheridge of low quality Rivers soils supplemented was also the by focallarge scale area for previous synoptic scalefertigation erosion (i.e. hazard irrigation mapping with all undertakenplant nutrients byprovided Griffith within University the supplied in irrigation2006. water). For this reason we have chosen to focus on the IFED case study to explore some of the land management challenges faced through implementing such a project in this landscape. The study focal area, The Flinders andbounded Gilbert by the River Einasleigh, Agricultural Gilbert and Resource Etheridge Assessment Rivers was also (FGARA) the focal area published for previous by synopticCSIRO in 2013 assessed a broadscale erosionrange hazardof issues mapping around undertaken the suitability by Griffith of University the soils, in 2006.climate and access to water and markets in the GilbertThe region Flinders for and a broad Gilbert rangeRiver Agricultural of potential Resource crops. Assessment Despite (FGARA) the comprehensive published by CSIRO analysis in 2013 undertaken by CSIRO in the FGARA,assessed a one broad issue range not of issues canvassed around thewas suitability the potential of the soils, risks climate of initiating and access major to water sub-surface and erosion markets in the Gilbert region for a broad range of potential crops. Despite the comprehensive analysis in the form ofundertaken gully and by channel CSIRO in erosionthe FGARA, in onethe issue area. not In canvassed the adjacent was the Flinderspotential risksand of Mitchell initiating Rivermajor catchments, sub-surface erosionsub-surface has erosion been in found the form to ofcontribute gully and channel between erosion 97% in the -100% area. ofIn thethe adjacent sediment Flinders load and to these rivers, hence it is criticalMitchell that River any catchments, development sub-surface in this erosion landscape has been befound assessed to contribute for its between potential 97% to-100% accelerate sub- of the sediment load to these rivers, hence it is critical that any development in this landscape be surface erosionassessed processes. for its potential To address to accelerate this gap, sub-surface the Federal erosion processes.Government To address commissioned this gap, the thisFederal report as an emerging priorityGovernment through commissioned the National this report Environmental as an emerging Research priority through Programme the National (NERP). Environmental Research Programme (NERP).

Figure 3 Terrestrial LiDAR scanner image of a highly active alluvial gully in the Gilbert catchment

1.2 Key findings

1.2.1 Gully extent

A remote sensing mapping programme undertaken by Griffith University in 2006 in conjunction with the Northern Gulf Natural Resource Management Group (NGNRMG), found that gully erosion was widespread through the region, both in the form of hillslope gullies on the Georgetown granite soils and alluvial gullies on the areas of Quaternary alluvium along the main rivers in the region. A total of around 10,000 ha of active gullies were mapped using ASTER satellite imagery in the initial synoptic scale mapping exercise. As part of this study we revisited the earlier mapping both on the ground and through manually digitising gullies using the much higher resolution imagery that is now available in Google Earth compared to what was available when the original mapping was undertaken. This analysis found that the original mapping over-estimated the extent of highly active gullies, probably by a factor of around two (giving a total area of active gullies of around 5000 ha).

The over-estimation was, however, not uniform. Areas with relatively low gully density per unit area (the large majority of the landscape) tended to significantly under-estimate the total extent of gullies, by as much as a factor of 10; while areas with high gully densities (smaller areas with extensive gullying) tended to over-estimate by a factor of about 4.

Given the inherent error that clearly exists in the remote sensing mapping approach, we can conclude that, providing satellite imagery of sufficiently high resolution is available (2.5m pan sharpened Spot 5 data as a minimum), manual digitisation based on visual observation is the only satisfactory method for mapping gullies in this landscape at a level that is sufficient for management and sediment budget quantification. This assumes that LiDAR data is not available, which would be the preferred data source if it was widely available.

1.2.2 Gully erosion rates

As part of this study we re-surveyed several gullies using differential GPS of gully scarp boundaries and Terrestrial LiDAR scanning of gully sections at sites for which we had airborne LiDAR data captured in 2006. The results from these surveys indicate that fairly typical alluvial gullies are delivering sediment yields of 260 - 450 t/ha/yr of sediment to the stream network (from the active gully area). More active gullies, such as the site G9GC1 at Greenhills are producing specific yields of 2160 t/ha/yr.

1.2.3 Gully erosion sediment yields

Assuming the sediment yields from the moderately active gullies are representative of the yields more generally, mean annual sediment contribution from gullies within the upper Gilbert fan would fall within the range 1.3 – 2.3 Mt/year, most of which can be regarded as land-use accelerated erosion. If the observed specific yields in the order of 2000 t/ha/yr are representative of a significant proportion of the mapped gullies, the mean annual sediment yields from gullies could be several times more than this conservative range.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 3

1.2.4 The IFED proposal

Proposal Overview: The proposed IFED irrigated agriculture area is located in the Dismal Creek catchment (encompassing two properties currently used as grazing leases - Chadshunt and Kutchera), which have a combined area of around 103,500 hectares. The area of the proposed development within these two properties is 65,000 ha., consisting of around 1600 x 40 ha plots, which would be leased in a share-farming type arrangement. Under this model lessees are responsible for clearing and laser levelling the land in preparation for irrigated agriculture and are contracted to provide the agricultural products to the processing plants on site. The IFED developers would provide the water infrastructure, the sugar mill for processing the sugar cane that is proposed as the primary crop, and a guar gum processing plant, to process the guar beans that are proposed as the other major crop to be grown on the site. The water supply infrastructure will consist of two major off-river water storages, on Dagworth and Huonfels Stations, that will receive 550,000 Ml/yr (around 1 Sydharb) in flow diverted from the Einasleigh and Etheridge Rivers each year, plus an unspecified volume from local runoff in the dam catchments. In total the two storages have a capacity of 2,100,000 Ml (or around 4 Sydharbs). The high flow diversion from the Einasleigh River equates to around 40% of the mean annual flow at Dagworth, which means that in low flow years that it will be a much higher proportion. Soil fertility was not regarded as an important criteria for site selection given that most nutrients required by the crops will be supplied in the water supply via sub-surface trickle irrigation in the process known as fertigation. While soil fertility may not be regarded as important, soil structure, stability and drainage capacity are critical if irrigation is to be sustainable in the long term and land degradation is not to ensue in the short to medium term.

Figure 4 DEM map showing the location of the proposed IFED agricultural area (outlined in yellow) within the Dismal Creek catchment above the apex of the Gilbert megafan. Note that the area is far from flat, located as it is on the tertiary age deposits of the upper Gilbert megafan, which is largely disconnected from upstream sources of “new” alluvium 4

1.2.5 The FGARA soil suitability analysis 1.2.5 The FGARA soil suitability analysis

a. TheThe CSIRO CSIRO FGARA FGARA soil soil suitability suitability analysis assessed assessed a awide wide range range of of derived derived parameters parameters using using a Digital a Digital SoilsSoils Mapping Mapping (DSM) (DSM) approachapproach to assessassess the the suitability suitability of of 74 74 different different crop/irrigation crop/irrigation scenarios scenarios in this in this region.region. The The DSM DSM approach approach reliedrelied heavilyheavily on on remotely remotely sensed sensed radiometric radiometric data, data, coupled coupled with with 30m 30m topographictopographic data data andand limitedlimited ground surveysurvey data. data. Hence, Hence, systematic systematic error error in the in the radiometric radiometric data data could introducecould introduce substantive substantive error errorinto theinto mapping. the mapping. b. A major caveat in the CSIRO FGARA report was that much of the area falling within the Karumba geologic formationA major caveat (also knownin the CSIRO as the FGARA Holroyd report Plain was and that the much Wyaaba of the Beds area in fallingother withingeologic the maps) Karumba is know n to havegeologic poor formation predictive (also power known with as regard the Holroyd to surface Plain soiland characteristics. the Wyaaba Beds The in otherKarumba geologic formation, maps) which is known to have poor predictive power with regard to surface soil characteristics. The Karumba consists of the Tertiary age remnants of the upper part of the Gilbert alluvial megafan deposited formation, which consists of the Tertiary age remnants of the upper part of the Gilbert alluvial between 2.6 and 23 Ma, was acknowledged to contain extensive areas of well-developed fericrete and megafan deposited between 2.6 and 23 Ma, was acknowledged to contain extensive areas of well- otherdeveloped lateritic fericrete deposits, and whichother lateriticare known deposits, to interfere which are with known the radiometric to interfere withsignatures, the radiometric potentially leading tosignatures, spurious resultspotentially in the leading Digital to Soilsspurious Mapping results approach. in the Digital As Soilsshown Mapping in Figure approach. 4, virtually As shownthe entire in area ofFigure the proposed 4, virtually IFED the agriculturalentire area of precinct the proposed falls within IFED agricultural the area flagged precinct by falls CSIRO within as beingthe area ‘known flagged to predictby CSIRO poorly’. as being ‘known to predict poorly’.

FigureFigure 4 Map 5 Map of theof the study study area area and and the the Proposed Proposed IFEDIFED area area which which shows shows that that almost almost the theentire entire area encompassedarea encompassed by the IFEDby the IFED agricultural area falls withinagricultural the bounds area of falls the areawithin “known the bounds to predict of poorly”the area “known to predict poorly”

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion

With a 90m pixel resolution the analysis is also blind to complex spatial heterogeneity, and so variability at finer resolution was necessarily averaged to the minimum resolution of the analysis. Our initial assessment is that there is considerable spatial variability in the soils within this area at a much finer resolution than is likely to have been captured by the 90m resolution radiometrics (irrespective of the issue outlined above associated with lateritic soils).

The CSIRO FGARA soil suitability analysis classified most of the soils in the IFED area as class 3 or 4 soils for spray, trickle and furrow irrigation. However, as highlighted above, these results must be viewed with considerable scepticism given the known limitations of the survey techniques in this area. This is acknowledged in the FGARA documentation, which states: “The Assessment is designed to inform consideration of development, not to enable particular development activities. As such, the Assessment informs – but does not seek to replace – existing planning processes.” (Bartley 2013, Preface).

In many areas it appears likely that the assessment over-estimated the suitability of the soils for irrigated agriculture, and it is likely that a much greater proportion of the soils in this area are class 5 soils. In particular the fact that sub-surface dispersibility was not systematically assessed, may render the assessment of the area as predominantly class 3 soils for trickle irrigation as overly optimistic. If the majority of the soils are indeed class 4 or class 5, and irrespective of whether fertigation is used, as a first order site selection criterion, this would typically preclude such an area being selected as the focus of a major irrigated agriculture development.

While the soil suitability analysis considers the susceptibility of surface erosion as a function of surface soil K factor and slope, and concludes that there is little erosion risk in this area, such an assessment does not assess the risk posed by sub-surface erosion – which is likely the source of >97% of all sediment delivered to streams in the region (based on evidence from the adjacent Mitchell catchment). Our preliminary assessment would suggest that sodic dispersible sub-soils are widespread within this area, and that much of the landscape is susceptible to sub-surface erosion, should these sub-soils be exposed.

1.2.6 Landscape analysis of the IFED area

The defining characteristic of the IFED area is that it is situated on an extremely old landscape surface, which was deposited as part of the Gilbert megafan between 2.6 – 23 million years ago. As the Etheridge and Gilbert Rivers have incised into the upper megafan the Dismal Creek catchment has effectively become isolated from the upper Gilbert catchment, starving it from any ‘new alluvium’ from the upper catchment. Consequently, the soils in this area have been weathering in-situ since they were deposited, giving rise to extremely poor, low nutrient soils.

A topographic analysis of the IFED area highlights the fact that this area is highly dissected by the well- developed Dismal Creek catchment stream network (Figure 6). The landscape is relatively steep for irrigated agriculture with a typical cross-valley gradients in the range of 0.3-0.6% and a typical long- valley gradient of 0.14-0.18%. These slopes are around three times the down-valley gradients, and up to 10 times the cross valley gradients found within the Burdekin irrigation area. 6

Figure 6 DEM and cross valley sections through the Dismal Creek catchment area encompassed by the proposed IFED area showing the highly developed stream network within this area and the cross-valley topographic variability

1.2.7 Soil erosion hazard in the IFED area

Our mapping indicated that there are 106ha of active gullies within the IFED agricultural precinct, which are fairly evenly distributed across the whole site, but particularly associated with drainage lines (i.e. they are predominantly alluvial gullying). There are also numerous examples of small gullies which have been initiated by road drainage lines. This suggests that much of the landscape is prone to gully erosion if disturbed.

Random field samples indicate that highly dispersible sodic sub-soils are common.

The soil chemistry data indicates a complex relationship between gully erosion and soil chemistry, and the traditional assumptions about soil stability (e.g. high Ca/Mg ratios = stable soils) may not necessarily apply in these soils. It is likely that there is a multi-variate, non-linear relationship between soil chemistry and erodibility. It is recommended that much more work needs to be done to understand this relationship.

Given that most gullies are initiated in, or adjacent to, drainage lines, the density of the drainage network in this area highlights the need to be extremely cautious and endeavour to exceed best management practices in the design and layout of the road network and areas that will be cleared and laser levelled.

Given the micro topography of the area, the development of paddocks for irrigation, particularly sub-surface trickle irrigation, will inevitably mean that sodic soils are exposed, which will need to be stabilised with large quantities of gypsum. The costs of gypsum required to stabilise the soils will need to be factored into the economic analysis of the overall development strategy.

Given the apparent susceptibility of this landscape to gully erosion, none the least due to the antiquity of the landscape, the presence of sodic soils, the gradient of the landscape and the high drainage density, generous buffers will be required around all drainage lines/depressions; and around all existing gullied areas to ensure that gully erosion is not exacerbated in this area.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 7

1.2.8 Buffer design to reduce the threat of gully erosion

Given the constraints identified within this landscape, and the high risk of initiating gully erosion under an intensified land-use regime, we recommend that generous buffers are left around all drainage lines and existing gullies that vary in width according to the stream order.

We modelled a range of buffer widths as a function of stream order according to the equation:

Where S0 > 1: Bw (m) = n x S0,

Where S0 = 1: Bw (m) = 2n x S0

Bw = buffer width (each side of stream), n = buffer multiplier (10 – 45m ); and S0 = stream order (Strahler)

All mapped gullies also include a buffer of the same extent as the 1st and 2nd order streams.

Based on the modelling, a buffer multiplier of 45m is achievable and would leave around 75,000 ha of land that could potentially be developed, while minimising the threat of gully erosion initiation in and around the drainage network.

Further work would need to be done to refine the stream network given that this analysis was based on the available 1:100,000 scale stream network, and there are some drainage lines not captured in this analysis. However, the total area would not differ significantly from that represented here.

If it did transpire that there was a significant length of stream network missed in this analysis it may be necessary to reduce the buffer multiplier, but we would not recommend that it be reduced to less the 30m.

The aim with these buffers would be to completely eliminate all land use pressure from these zones, including clearing, roads and tracks and disturbances associated with water infrastructure.

These buffers should be the bare minimum area within the agricultural precinct left uncleared. Additional wildlife corridors should be superimposed on this network to provide wildlife movement across the drainage network (i.e. in a NW/SE direction).

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 8

Figure 7 Proposed stream network buffers for the Dismal Creek catchments area

1.2.9 Offset opportunities

Despite the limitations of the earlier remote sensing gully mapping it is clear that there are some areas of intensive gully activity which should be the focus of remedial activities irrespective of whether further developments proceed within the Gilbert region. In the event that large scale developments do proceed, significant contributions towards the rehabilitation of these areas should be built into the development process. Costs of intensive gully rehabilitation have been shown to be in the order of $30,000 per hectare of active gully (Shellberg and Brooks, 2013).

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 9

Figure 8 Map showing the aggregation of alluvial gullies in the area around the confluence of the Einasleigh and Etheridge Rivers and on the Gilbert River in the Greenhills areas. Note the gully mapping shown here is the earlier remote sensing-based mapping which has been shown to over represent the extent of gullies in the areas of high gully concentration. Hence the actual extent of gullies is less than represented here, but we do know that there are some major gully complexes here that could be managed with appropriate investment

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion

2 Background With the current focus on developing the agricultural potential of northern Australia, it is critical to evaluate potential risks that characteristics of this landscape pose to the viability of these new developments, as well as to the existing natural assets. A major investigation undertaken by CSIRO in the Gilbert catchment (Petheram et al., 2013), assessed the suitability of the catchments for irrigated agriculture from a broad range of 10perspectives, including, climate, soils, crop agronomic suitability, the potential for the location of dams, access to markets and broader macro-economic factors, amongst other factors. The one factor that was not considered in the CSIRO analysis, however, was the suitability of the soils from the perspective of their potential susceptibility2 Backgr toound alluvial gully erosion, or the likely threats of the initiation of off-site erosion in areas immediatelyWith the adjacent current focus to areason developing likely theto beagricultural developed. potential Alluvial of northern gully Australia, erosion it is hascritical been shown to be widespread acrossto evaluate northern potential Australia, risks that and characteristics is prevalent of this in landscape some ofpose the to areasthe viability that of are these currently new under active developments, as well as to the existing natural assets. A major investigation undertaken by CSIRO in consideration forthe intensive Gilbert catchment agricultural (Petheram development et al., 2013), assessed (Shellberg the suitability and Brooks, of the catchments 2012). Research for irrigated undertaken through the Tropicalagriculture River from and a broad Coastal range Knowledge of perspectives, Programme including, climate, (TRaCK) soils, crophas agronomic highlighted suitability, the extent of current the potential for the location of dams, access to markets and broader macro-economic factors, alluvial gully erosionamongst across other factors.large areas The one of factorthe northern that was not Australia considered (Table in the CSIRO 1, Figure analysis, 8 ),however, which was by definition occurs on the alluvialthe suitability land ofthat the issoils potentially from the perspective most prospective of their potential for susceptibility agricultural to alluvial development. gully erosion, Whilst only a or the likely threats of the initiation of off-site erosion in areas immediately adjacent to areas likely small proportionto of be thedeveloped. total land Alluvial area, gully theerosion extremely has been shown high toerosion be widespread rates acrossthat occurnorthern within Australia, alluvial gullies, which can oftenand have is prevalent erosion in ratessome of that the areasexceed that 500are currently t/ha/year under from active individualconsideration gully for intensive complexes (Brooks., et agricultural development (Shellberg and Brooks, 2012). Research undertaken through the Tropical al., 2008; Shellberg,River andet. Coastalal., 2012), Knowledge with Programmesediment (TRaCK) concentrations has highlighted at thegully extent outlets of current that alluvial exceed gully 100,000 mg/l. (Shellberg et al.,erosion 2013). across The large catchments areas of the draining northern Australia to the (TableGulf 1,of Figure Carpentaria 9), which byand definition indeed occurs most large floodplain on the alluvial land that is potentially most prospective for agricultural development. Whilst only a rivers in northernsmall Australia, proportion inclu of theding total theland largerarea, the ones extremely draining high erosion to the rates Great that occurBarrier within Reef, alluvial are highly prone to this type of alluvialgullies, gully which erosion can often (Brooks have erosion et al., rates 2009; that exceed Shellberg 500 t/ha/year & Brooks, from individual2012, 2013 gully ),complexes and hence this is a (Brooks., et al., 2008; Shellberg, et al., 2012), with sediment concentrations at gully outlets that critical issue in anyexceed development 100,000 mg/l. (Shellbergscenario. et al., 2013). The catchments draining to the Gulf of Carpentaria and indeed most large floodplain rivers in northern Australia, including the larger ones draining to the Research has shownGreat Barrierthat notReef, only are highly does prone this to erosion this type processof alluvial representgully erosion (Brooksthe dominant et al., 2009; sediment Shellberg source to & Brooks, 2012, 2013), and hence this is a critical issue in any development scenario. most large rivers in northern Australia (Rustomji et al., 2010; Shellberg and Brooks, 2012), but it also represents a keyResearch threat has to showninfrastructure that not only such does asthis roads, erosion processdams, representirrigation the paddocksdominant sediment and other source infrastructure to most large rivers in northern Australia (Rustomji et al., 2010; Shellberg and Brooks, 2012), but across large areasit also of representsthe north. a key threat to infrastructure such as roads, dams, irrigation paddocks and other infrastructure across large areas of the north.

Table 1 Areas of mappedTable 1 Areas (minimum) of mapped active(minimum) gully active erosion gully erosion across across key key northern northern Australian Australian catchments catchments

This form of erosionThis form represents of erosion representsa dominant a dominant source source of accelerated of accelerated erosion in manyin many tropical tropical rivers and rivers and as such is disproportionatelyas such responsible is disproportionately (compared responsible with (compared the area with of the land area ofinvolved) land involved) for forthe the elevated elevated sediment load sediment load in many rivers in Northern Australia (Shellberg and Brooks, 2012). in many rivers in Northern Australia (Shellberg and Brooks, 2012).

Alluvial gully erosion can be very difficult and expensive to manage and so once initiated it represents a threat to the viabilityDeveloping of guidelines the somefor agriculture of developmentthe most in northern valuable Australia alluvial to avoid and land reduce acceleratedassets erosionas well as to infrastructure and to downstream river reaches and receiving waterbodies. Consequently, avoiding the initiation of new alluvial

Alluvial gully erosion can be very difficult and expensive to manage and so once initiated it represents a threat to the viability of the some of the most valuable alluvial land assets as well as to infrastructure and to downstream river reaches and receiving waterbodies. Consequently, avoiding the initiation of new alluvial gully erosion should be given the highest priority in natural resources management planning, even ahead of any actions to address existing gully erosion problem areas.

Figure 9 Map showing the catchments in northern Australia which either have mapped distributions of the extent of alluvial gully erosion or where alluvial gullies are known to exist but have yet to be mapped. Catchments shown in light grey

3 Project overview

Given the high degree of interest in agricultural development across northern Australia and the known threat posed by alluvial gully erosion, there is a pressing need to look closely at the extent to which this inherent characteristic of the northern Australian landscape poses a risk to investment in the north. In this project we use the Gilbert/Einasleigh area as a case study to assess how an irrigation development might be planned and implemented in such a way as it minimises the risk of initiating new alluvial gully erosion or exacerbating existing erosion problem areas (both alluvial gullies and stream bank erosion). To date two proposals for establishing irrigated agriculture in the Gilbert catchment have been put forward; the Gulf Savannah Development proposal (2009), and the Integrated Food and Energy Development proposal (IFED, 2012). To date most attention has been directed at IFED due to the scale of the proposal and the high profile nature of its proponents. It is also fair to say that the IFED project has attracted more criticism from the local community because of its scale, its radical departure from accepted practice, and the greater potential impact/benefit of the development on the local economy, infrastructure and environment. For this reason we have chosen to focus on the IFED case study to explore some of the land management challenges faced through implementing such a project in this landscape. We have chosen an area bounded by the Einasleigh and Gilbert Rivers as the focus for this study (Figure 10, Figure 11), as this encompasses the areas in which irrigated agriculture development is proposed, and includes the focal areas for previous synoptic scale erosion hazard remote sensing mapping undertaken by Griffith University (Brooks et al., 2006) Gully mapping work undertaken as part of this previous remote sensing mapping programme was not ground truthed to the extent that would have been preferred at the time, and so one of the first tasks of this project was to revisit the areas previously mapped and undertake further ground truthing of the mapping beyond that undertaken in the original synoptic scale mapping exercise.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 12

4 Project objectives

1. This project will improve the knowledge base for planning and assessing potential erosion risks associated with new agriculture developments in northern Australia, particularly off site gully and stream bank erosion that may be inadvertently initiated as part of the development process. 2. The project will raise awareness of issues associated with gully and streambank erosion (and elevated sediment loads in rivers) within the community so that unintended consequences associated with agricultural development might be avoided. 3. Within the pilot study area, the project will also identify the areas that are currently highly degraded under broad-scale grazing and the data developed within the project should inform land-use planning processes.

5 Structure of this report

This report has been broken into three parts:

1. Review of relevant background data, including: a. the 2006-07 gully mapping project undertaken by Griffith for the Northern Gulf NRM Group and Federal Govt; b. the CSIRO FGARA report; c. other research undertaken through the Tropical Rivers and Coastal Knowledge (TRaCK) Programme. 2. Undertake a preliminary soil erosion risk assessment of the area proposed for development within the Integrated Food and Energy Development (IFED) in the Gilbert catchment, which includes analysis of the landscape setting and susceptibility of the soils to erosion during the development process and over the longer term operation of the agricultural area. 3. Provide management recommendations regarding: a. approaches for minimising the impacts of development within the development area – should development’s such as the IFED proceed b. How to avoid development in inappropriate (high risk) areas c. The identification of hotspots of gully activity for potential emedialr actions as offsets to land- use intensification in low risk, high value land.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 13

6 Review of 2006 Griffith University gully mapping 6.1 Regional scale remote sensing gully mapping

In 2006 alluvial gully erosion was identified as a key threat to water quality across large areas of the northern and southern Gulf regions, and a collaborative project between the Northern Gulf NRM Group, Land and Water Australia and Griffith University (Brooks et al., 2006) was set up to undertake synoptic scale mapping across the region to gauge the scale of the issue. This mapping exercise, which was the first of its kind in Queensland, put the issue of gullies in alluvium firmly on the agenda for land and water managers across northern Australia. While extensive gully erosion mapping had previously been undertaken in the Victoria River District in the Northern Territory, it had not been articulated that this erosion process differs fundamentally from the standard hillslope erosion model fundamentally from the standard hillslope erosion model that was widely recognised in southern Australia, that was widely recognised in southern Australia, and was assumed to be similar across northern and was assumed to be similar across northern Australia (e.g. Prosser et al., 2001). While hillslope gully Australia (e.g. Prosser et al., 2001). While hillslope gully erosion does occur in northern Australia, alluvialerosion gully does erosion occur in is northern far more Australia, widespread alluvial and gully represents erosion a is greater far more threat widespread to water and quality, represents aquatic a ecosystemsgreater threat and to to water infrastructure quality, aquatic than doesecosystems hillslope and gullying. to infrastructure than does hillslope gullying.

InIn thisthis partpart of thethe project,project, wewe revisit revisit a a number number of of sites sites that that were were originally originally surveyed surveyed in 2006in 2006 and and undertake undertakefollow-up surveysfollow-up to gaugesurveys the to rates gauge of alluvialthe rates gully of activityalluvial overgully the activity intervening over the 8 yearintervening period since8 year the last periodsurveys. since We the also last revisit surveys. a sample We (0.63%) also revisit of the a samplegullies mapped(0.63%) in of the the Gilbert gullies catchment mapped inin the2006 Gilbert from ASTER catchmentsatellite imagery. in 2006 We from drew ASTER upon satellite a range imagery.of datasets We that drew were upon collected a range during of datasets the previous that researchwere as collectedoutline in during Table 2 the. Key previous datasets research that are as being outline used in as Table base 2. data Key for datasets assessing that change are being since theused original as base work datawas undertakenfor assessing in change2006/07 since include the the original LiDAR work data, wasground undertaken GPS surveys in 2006/07 of selected include gullies; the gully LiDAR distribution data, ground GPS surveys of selected gullies; gully distribution derived from ASTER imagery, aerial imagery. derived from ASTER imagery, aerial imagery.

TableTable 2 2 Datasets Datasets that that have have been been reviewed reviewed during during thisthis project andand whichwhich have formed the basis for subsequent analysesanalyses ofof erosion erosionpotential potential undertaken undertaken during thisduring study. this study Dataset Comments Aerial video 2004 Video covering major channels flowing to the Gulf of Carpentaria (film using digital video camera, with incorporated GPS system, fixed to underside of single engine fix wing plane). Airborne data 2006 Lidar (1m resolution) blocks covering ~128km2 within the Gilbert (Airborne Research catchment. Australia, small fixed Trispectral laser scanner (NDVI) covering LIDAR blocks. wing plane with High resolution aerial photos approximately positioned from aircraft GPS. underwing instrumentation pods) On-ground field work Tracklog record of all investigated sites and routes in between. 2006 Extensive ground photo catalogue (co-ordinates extracted from tracklog). GPS survey of selected gullies Aster Mosaic of one year of the Northern and Southern Gulf regions Selection of individual scenes Landsat Library of SLATS yearly mosaic and individual scenes. Selection of historic Various resolutions aerial photos Final Gully Mapping 15m pixel resolution & 1km grid showing proportion of grid square Data polygons occupied by gully pixels

The focal area for this study is shown in Figure 9, along with the GPS track logs showing the ground survey undertaken in this study. This builds on the pre-existing field and aerial surveys completed in, 2004, 2006 and 2008. The study focal area encompasses the areas of greatest gully erosion density as well as the areas Developingthat are guidelinesproposed for for agriculture irrigated development agriculture in northern development Australia to (avoidFigure and 10 reduce). accelerated erosion

Figure 9 Map of the Gilbert River Project area (outlined in red), which encompasses the area likely to be developed for irrigated agriculture, and within which erosion studies have been conducted. Also shown is the coverage of the various datasets that have been used in this study.

The focal area for this study is shown in Figure 10, along with the GPS track logs showing the ground survey undertaken in this study. This builds on the pre-existing field and aerial surveys completed in, 2004, 2006 and 2008. The study focal area encompasses the areas of greatest gully erosion density as well as the areas that are proposed for irrigated agriculture development (Figure 11).

Figure 10 Map of the Gilbert River Project area (outlined in red), which encompasses the area likely to be developed for irrigated agriculture, and within which erosion studies have been conducted. Also shown is the coverage of the various datasets that have been used in this study

Figure 11 Image showing the focal study area (orange outline) with the location of the major infrastructure for the IFED project indicated (i.e. 2 off-stream water storage lakes; diversion channels (red); and the proposed irrigation area (yellow outline). The yellow polygons scattered across the image are the gully areas as mapped from Aster imagery in 2006. Also shown are the centre pts of ground photos collected in 2014 as part of the ground validation exercise for the gully mapping Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 15

6.2 Reassessment of 2006 alluvial gully erosion hazard mapping

Initial ground-truthing was undertaken by visiting a sample of 55 (or a sample of 0.63 %) of the mapped gully polygons on the ground (Figure 12) and ascertaining whether the mapped features were accurately mapped. Results from this exercise (Table 3) indicated a high proportion of false positives from the ASTER gully mapping exercise. The most common features that were misrepresented were light coloured rocky outcrops or patches of annual grasses located on floodplains in approximately the locations that alluvial gullies would be expected. Based on these results we then carried out another test using Google Earth imagery, in which we digitised all visible gullies. Such an analysis was not possible when the original mapping was done due to the limited availability of high resolution imagery covering the majority of the area within Google Earth back in 2006, where the best available imagery was the ~30m Landsat imagery. In part, the misrepresentation of gully polygons in the original mapping is a data resolution issue, given that the ASTER analysis was largely detecting areas of “bare ground”, which are generally closely associated with gully distribution, but are not an exclusive indicator. Often the actual gullies are a smaller proportion of the total area that is mapped as bare ground using the ASTER remote sensing technique. There were also some registration issues with the original mapping that explain some of the difference between mapped and observed gullies, given that there are often gullies in close proximity to the mapped gully pixels, but not at the exact location.

Figure 12 Locations of gully polygons with on ground field visits (316 photos were taken amongst 55 gully polygons)

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion conducted. Also shown is the coverage of the various datasets that have been used in this study.

Figure 10 Image showing the focal study area (orange outline) with the location of the major infrastructure for the IFED project indicated (i.e. 2 off-stream water storage lakes; diversion channels (red); and the proposed irrigation area (yellow outline)). The yellow polygons scattered across the image are the gully areas as mapped from Aster imagery in 2006. Also shown are the centre pts of ground photos collected in 2014 as part of the ground validation exercise for the gully mapping.

6.2 Reassessment of 2006 alluvial gully erosion hazard mapping

Initial ground-truthing was undertaken by visiting a sample of 55 (or a sample of 0.63 %) of the mapped gully polygons on the ground (Figure 11) and ascertaining whether the mapped features were accurately mapped. Results from this exercise (Table 3) indicated a high proportion of false positives from the ASTER gully mapping exercise. The most common features that were misrepresented were light coloured rocky outcrops or patches of annual grasses located on floodplains in approximately the locations that alluvial gullies would be expected. Based on these results we then carried out another test using Google Earth imagery, in which we digitised all visible gullies. Such an analysis was not possible when the original mapping was done due to the limited availability of high resolution imagery covering the majority of the area within Google Earth back in 2006, where the best available imagery was the ~30m Landsat imagery. In part, the misrepresentation of gully polygons in the original mapping is a data resolution issue, given that the ASTER analysis was largely detecting areas of “bare ground”, which are generally closely associated with gully distribution, but are not an exclusive indicator. Often the actual gullies are a smaller proportion of the total area that is mapped as bare ground using the ASTER remote sensing technique. There were also some registration issues with the original mapping that explain some of the difference between mapped and observed gullies, given that there are often gullies in close proximity to the mapped gully pixels, but not at the exact location.

16 . Figure 11 Locations of gully polygons with on ground field visits (316 photos were taken amongst 55 gully polygons)

Table 3 SummaryTable statistics 3 of Summary the “ground-truthed” statistics gully of polygons the “ground-truthed” gully polygons

Number of Mapped Gullies (2006) 8671 Number of 2008 mapped Gully Polygons Visited (2014) 55 (0.63%) True Positives for the 55 Gully Polygons 16 (29%) False Positives for the 55 Gully Polygons 39 (71%)

6.2.1 Google Earth gully mapping 6.2.1 Google Earth Gully Mapping TheThe existing existing ASTER ASTER derived derived gully gully mapping mapping has has been been reassessed reassessed both both on on the the ground ground and and using using the the higherhigher resolution resolution imagery imagery now now available available across across the the entire entire catchment. catchment. At At the the time time that that the the mapping mapping was undertaken in 2006/07 the 15m resolution ASTER imagery was the highest resolution imagery was undertaken in 2006/07 the 15m resolution ASTER imagery was the highest resolution imagery available across the region. As a minimum, there is now 5m pan-sharpened Spot imagery available available across the region. As a minimum, there is now 5m pan-sharpened Spot imagery available across the whole region, with selected areas having <1m Quickbird and other image sources. With theseacross data the wewhole have region, sampled with a portionselected of areas the havingarea originally <1m Quickbird mapped andand otherre-mapped image itsources. via manual With digitistheseation data ofwe visible have sampledgully and a scald portion areas. of the A gridarea of originally 1/10th of mapped a degree and squares re-mapped was overlainit via manual on the gullydigitisation mapping of undertakenvisible gully inand 2006 scald using areas. Aster A grid satellite of 1/10 imagery.th of a Thirteen degree squaresgrid squares was overlainwere selected on the whereingully mapping gullies thatundertaken are discernible in 2006 in using Google Aster Earth satellite imag imagery.ery were Thirteenmanually grid mapped. squares Additional were selected gullies wherein gullies that are discernible in Google Earth imagery were manually mapped. Additional gullies were manually mapped to complete the coverage within11 the IFED irrigation area. A total of 473 gully polygons were manually digitised. As shown in Table 4 there were was a similarly high number of false positives indicated using the manual digitisation method from Google Earth as the field checking.

Figure 13 Map showing the distribution of the 0.1 degree grid with the sample grids highlighted in orange

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion werewere manually manually mapped mapped to to complete complete thethe coveragecoverage within thethe IFEDIFED irrigation irrigation area. area. A A total total of of 473 473 gully gully polygonspolygons were were manually manually digitised. digitised. AsAs shownshown inin Table 44 therethere werewere was was a a similarly similarly high high number number of offalse false positivespositives indicated indicated using using the the manual manual digitisationdigitisation method fromfrom GoogleGoogle Earth Earth as as the the field field checking. checking.

17 FigureFigure 12 12 Map Map showing showing the the distributiondistribution of the 0.10.1 degreedegree grid grid with with the the sample sample grids grids highlightedhighlighted in orangeorange

TableTableTable 4 4 Summary 4 Summary Summary statistics statistics statistics for the Google for for the theearth GoogleGoogle gully mapping earth test gullygully data mappingmapping test test data data

TotalTotal area area of of Google Google Earth Earth mapped mapped gullies.gullies. 5.65.6 km km2 2 AreaArea of of Go Googleogle Earth Earth mapped mapped gulliesgullies with Aster mappingmapping Area Area 5.25.2 km km2 2 TotalTotal area area of of overlap overlap between between GoogleGoogle Earth mapped gulliesgullies and and Aster Aster derived derived 2 0.870.87 km km 2 gullygully polygons polygons (2006). (2006). PercentagePercentage of of Google Google Earth Earth mapped mapped gullygully area capturedcaptured inin Aster Aster derived derived gully gully 16.8%16.8% polygonspolygons (2 (2006)006) area. area. RatioRatio of of the the number number of of Google Google EarthEarth mappedmapped gullies intersectingintersecting and and not not 143/420143/420 intersecting Aster derived gully polygons (2006). (34%) intersecting Aster derived gully polygons (2006). (34%)

TableTable 5 Table 5 Table showing showing the extent the of gullies extent in theof testgullies blocks in represented the test blocksin both the represented ASTER derived in gully both layer the and ASTER the Google Earth mapped gully layer Tablederived 5 Table gully showing layer and the the extent Google of Earth gullies mapped in the gully test layerblocks represented in both the ASTER derived gully layer and the Google Earth mapped gully layer Google Earth Area of Aster Derived Gully Google Earth Area of Aster Derived Gully Percentage of Google Earth Grid Mapped Gully Polygons Intersecting Percentage of Google Earth Grid Mapped Gully Polygons Intersecting Mapped Gullies Captured in Square Area Google Earth Mapped Mapped Gullies Captured in Square Area2 Google Earth Mapped2 Aster Derived Gully Polygons (km ) Gullies (km ) Aster Derived Gully Polygons 53 (km0.09712) Gullies0.0034 (km 2) 3.5% 5354 0.09710.1627 0.00340.0019 1.2%3.5% 5466 0.16270.9259 0.00190.1822 19.7%1.2% 6668 0.92590.1029 0.18220.0090 8.7%19.7% 6869 0.10290.0753 0.00900.0051 6.8%8.7% 6970 0.07530.0640 0.00510.0000 0.0%6.8% 7082 0.06401.3882 0.00000.5629 40.5%0.0% 8284 1.38820.1178 0.56290.0173 14.7%40.5% 8485 0.11780.1223 0.01730.0074 6.1%14.7% 8586 0.12230.2725 0.00740.0044 1.6%6.1% 86114 0.27250.0863 0.00440.0089 10.3%1.6% 114118 0.08630.4328 0.00890.0018 0.4%10.3% 135 0.9391 0.0472 5.0% 118 0.4328 0.0018 0.4% 135Total 0.93914.7869 0.04720.8516 5.0%

TableTotal 6 Table showing4.7869 extent of over/under0.8516 representation of gully area ASTER derived gully Tablelayer 6 Table and showing the Google extent ofEarth over/under mapped representation gully layer of gully area ASTER derived gully layer and the Google Earth mappedTable gully6 Table layer showing extent of over/under representation of gully area ASTER derived gully layer and the GoogleGoogle EarthEarth mapped gully layer Ratio of Google Earth Mapped TableGrid 6 Table showingMapped Gullyextent of over/underAster Derived representation Gully of gullyGully areaArea ASTERand Aster derived Derived gully layerSquare and the GoogleGoogleArea Earth Earth mappedPolygons gully layer Area (km2) GullyRatio Polygonof Google Area Earth Within Mapped Grid Grid Mapped(km2 Gully) Aster Derived Gully Gully Area Squareand Aster Derived Square53 Google0.0971Area Earth Polygons0.0702 Area (km2) GRatioully P olygonof Google138% Area Earth Within M appedGrid Grid54 Mapped0.1627(km Gully2) Aster Derived0.0227 Gully Gully AreaSquare716% and Aster Derived 2 Square6653 Area0.92590.0971 Polygons0.76940.0702 Area (km ) Gully Polygon120%138% Area Within Grid 2 6854 (km0.10290.1627) 0.03470.0227 297%716%Square 5369 0.09710.0753 0.07020.0718 105%138% 5470 0.16270.0640 0.02270.011512 558%716% 82 1.3882 5.4733 25% 84 0.1178 0.036712 321% 85 0.1223 0.0995 123% 86 0.2725 0.1305 209% 114 0.0863 0.2980 29% 118 0.4328 0.3496 124% 135 0.9391 1.1583 81% Total 4.7869 8.5260

6.2.1.1 Summary of Gully Mapping Ground-Truthing

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion The analysis above (Table 5, Table 6) shows a poor relationship between Google Earth mapped gullies and the Aster derived gully polygons. Figure 13 shows that the number of false positive gullies (from ASTER) increases exponentially as the number of Google Earth mapped gullies per grid square increases. This suggests that the ASTER-based method is fairly good at low gully densities, picking out isolated gullies from the surrounding landscape, providing it is of a sufficient scale to be detected. In areas of low gully density, however, the ASTER imagery tends to under predict gully area, whereas in areas with high gully density the ASTER mapping tends to both overestimate the total area of gullies, and have a poor ability to predict the actual location of gullies.

The general conclusion we can draw from this exercise is that while the ASTER based mapping can give a general idea of the relative distribution of alluvial gullies within these savannah environments (i.e. it can distinguish areas of high and low gully density) the technique is not sufficiently accurate for quantitative predictions of gully area or potential sediment loads derived from gully erosion. Manual digitisation of gullies is a far superior method, providing image resolution equivalent to 2.5m pan- sharpened Spot data (or better) is available within Google Earth. The ASTER derived mapping from the 2006 study does give a reasonable idea of the hotspots of gully activity, but the mapping requires further detailed work if it is to be used as a basis for property-scale rehabilitation planning. It should also be noted that, where available, high resolution (1m) lidar data is far superior for gully mapping than any remotely sensed imagery.

Figure 13 Graph showing the relationship between Google Earth mapped gully area (per grid square) compared with ASTER derived gully area (per 1/10th degree grid square).

6.2.1.1 Summary of Gully Mapping Ground-Truthing

6.2.1.1 Summary of gully mapping Ground-Truthing The analysis above ( The analysis above (Table 5, Table 6) shows a poor relationship between Google Earth mapped Table 5, Table 6) shows a poor relationship between Google Earth mapped gullies and the Aster derived gully gullies and the Aster derived gully polygons. Figure 14 shows that the number of false positive gullies (frompolygons. ASTER) Figure increases 13 shows exponentially that the number as the of number false positive of Google gullies Earth (from mapped ASTER) gulliesincreases per exponential grid squarely as increases.the number This of suggests Google Earth that mappedthe ASTER-based gullies per method grid square is fairly increases. good atThis low suggests gully densities, that the ASTER-based picking out isolatedmethod gullies is fairly from good the at lowsurrounding gully densities, landscape, picking providing out isolated it is gulliesof a sufficient from the surroundingscale to be detected.landscape, In areasproviding of low it isgully of a density,sufficient however, scale to bethe detected. ASTER imagery In areas tends of low to gully under density, predict however, gully area, the whereasASTER imagery in areastends with to under high predictgully density gully area, the ASTERwhereas mapping in areas tendswith high to both gully overestimate density the ASTER the total mapping area tendsof gullies, to both andove haverestimate a poor the ability total areato predict of gullies, the andactual have location a poor ofability gullies. to predict the actual location of gullies.

TheThe general general conclusionconclusion we cancan drawdraw from from this this exercise exercise is isthat that while while the the ASTER ASTER based based mapping mapping can cangive a givegeneral a general idea of idea the ofrelative the relative distribution distribution of alluvial of alluvialgullies within gullies these within savannah these savannah environments environments (i.e. it can (i.e.distinguish it can distinguish areas of high areas and of low high gully and density) low gully the density)technique the is techniquenot sufficiently is not accurate sufficiently for quantitative accurate for quantitative predictions of gully area or potential sediment loads derived from gully erosion. Manual predictions of gully area or potential sediment loads derived from gully erosion. Manual digitisation of digitisation of gullies is a far superior method, providing image resolution equivalent to 2.5m pan- gullies is a far superior method, providing image resolution equivalent to 2.5m pan-sharpened Spot data (or sharpened Spot data (or better) is available within Google Earth. The ASTER derived mapping from thebetter) 2006 is studyavailable does within give Googlea reasonable Earth. idea The ofASTER the derivedhotspots mapping of gully from activity, the but2006 the study mapping does give requires a furtherreasonable detailed idea work of the if hotspots it is to be of used gully asactivity, a basis but for the property-scale mapping requires rehabilitation further detailed planning. work It ifshould it is to be alsoused be as noted a basis that, for property-scalewhere available, rehabilitation high resolution planning. (1m) It lidar should data also is befar notedsuperior that, for where gully available,mapping high thanresolution any remotely (1m) lidar sensed data isimagery. far superior for gully mapping than any remotely sensed imagery.

Figure 13 Graph showing the relationship between Google Earth mapped gully area (per grid square) compared with ASTER Figurederived 14 gullyGraph area showing (per 1/10 theth relationship degree grid between square). Google Earth mapped gully area (per grid square) compared with ASTER derived gully area (per 1/10th degree grid square)

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 19

6.3 Resurvey of previously surveyed gullies using Differential GPS and Terrestrial LiDAR

To ascertain the rates of sediment production associated with gully erosion in the Gilbert area we undertook resurveys of a number of gullies that had originally been surveyed in 2006. One gully complex was resurveyed using terrestrial LiDAR and two using differential GPS (Figure 16, Figure 19). The results from these resurveys indicate that the gullies have typically been expanding at a rate of 1 – 2% of their 2006 area per annum over the last 8 years. Together, the two gullies surveyed have contributed a minimum of around 15,000 tons of sediment over the 8 year period (2000 t/yr) between surveys from headscarp retreat alone. From similar analysis elsewhere (Shellberg., et al., 2013) , we know that the contribution from new gully headscarp retreat to the total sediment yield from the gully complex is only a minor proportion of the total sediment yield at the gully outlet – perhaps only 25% (Shellberg et al., 2013) – and hence the total yield from the gullies is likely to be much greater than this figure.

Figure 15 Map showing the location of sites at which detailed gully resurveys were carried out.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 20

6.3.1 Gully Erosion rate Analysis - 2006 – 2014 – Abingdon gully complex 1 (ABGC1) – Bel Bel Crossing

6.3.1.1 Erosion Volume Estimate

The area of erosion for the surveyed gullies was determined from the areal difference between the 2006 and 2014 ground survey, and the volume of change estimated by extrapolation from the LiDAR DEM acquired in 2006. A representative area of gully present in the 2006 LiDAR was selected and was then broken into relatively homogenous sections (Erosion Polygons) (Figure 17). An area proximal to each erosion polygon was then digitised and the pixels values of the DEM within the Estimate Polygons (Figure 17) extracted and an average elevation calculated. The average elevation of an Estimate Polygon was the subtracted from each pixel elevation of the corresponding Erosion Polygon. The total volume of erosion was then determined from the sum of the differences between all erosion polygon pixel elevations and the average polygon elevation for the adjacent polygon. The total volume of erosion for the gully complex (ABGC1) was determined to be 6450.99 m3 (+/- 1612 m3), or 224 t/ ha/yr +/- 56 t).

Figure 16 Comparison between differential GPS gully scarp surveys completed in 2006 and 2014

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion The results from these resurveys indicate that the gullies have typically been expanding at a rate of 1 – 2% of their 2006 area per annum over the last 8 years. Together, the two gullies surveyed have contributed a minimum of around 15,000 tons of sediment over the 8 year period (2000 t/yr) between surveys from headscarp retreat alone. From similar analysis elsewhere (Shellberg., et al., 2013) , we know that the contribution from new gully headscarp retreat to the total sediment yield from the gully complex is only a minor proportion of the total sediment yield at the gully outlet – perhaps only 25% (Shellberg et al., 2013) – and hence the total yield from the gullies is likely to be much greater than this figure.

Figure 14 Map showing the location of sites at which detailed gully resurveys were carried out.

6.3.1 Gully Erosion rate Analysis - 2006 – 2014 – Abingdon gully complex 1 (ABGC1) – Bel Bel Crossing

6.3.1.1 Erosion Volume Estimate The area of erosion for the surveyed gullies was determined from the areal difference between the

21 2006 and 2014 ground survey, and the volume of change estimated by extrapolation from the LiDAR DEM acquired in 2006. A representative area of gully present in the 2006 LiDAR was selected and was then broken into relatively homogenous sections (Erosion Polygons) (Figure 16). An area proximal to each erosion polygon was then digitised and the pixels values of the DEM within the Estimate Polygons (Figure 16) extracted and an average elevation calculated. The average elevation of an Estimate Polygon was the subtracted from each pixel elevation of the corresponding Erosion Polygon. The total volume of erosion was then determined from the sum of the differences between all erosion polygon pixel elevations and the average polygon elevation for the adjacent polygon. The total volume of erosion for the gully complex (ABGC1) was determined to be 6450.99 m3 (+/- 1612 m3), or 224 t/ha/yr +/- 56 t).

Figure 15 Comparison between differential GPS gully scarp surveys completed in 2006 and 2014

Figure 17 Change detection procedure based on 2008 LiDAR data from the BelBel crossing site Figure 16 Change detection procedure based on 2008 LiDAR data from the BelBel crossing site Table 7 Change in total are of the gully at BelBel Crossing on the Einasleigh River 2006-2014 Table 7 Change in total are of the gully at BelBel Crossing on the Einasleigh River 2006-2014

ABGC1 Headscarp retreat rate and sediment yield Gully Area 2014 72110.05 m2 Gully Area 2006 64686.91 m2 Area Increase 7471.87 m2 (11.6%) Volume Estimate 6450.99 m3 Sediment yield (BD 1.8) 224 t/ha/yr (head scarp only) (+/- 56)

6.3.2 Gully Erosion rate Analysis - 2006 – 2014 Gully complex 1 G9A GC2 –

6.3.2.1 Erosion volume estimate 14 A similar process to that described above for ABGC1 was also undertaken for G9AGC2 and the total sediment yield from this gully complex determined to be 1746 m3 (+/- 436 m3) or 137 t/ha/yr. As with the other site, this figure is from the headscarp retreat only, and as such the total sediment yield from the gully is likely to be 2 – 3 times higher than this.

Figure 18 Change detection procedure based on 2008 LiDAR data from the G9A GC2 site near Greenhills

22 Figure 14 Map showing the location of sites at which detailed gully resurveys were carried out.

6.3.1 Gully Erosion rate Analysis - 2006 – 2014 – Abingdon gully complex 1 (ABGC1) – Bel Bel Crossing

6.3.1.1 Erosion Volume Estimate The area of erosion for the surveyed gullies was determined from the areal difference between the 2006 and 2014 ground survey, and the volume of change estimated by extrapolation from the LiDAR DEM acquired in 2006. A representative area of gully present in the 2006 LiDAR was selected and was then broken into relatively homogenous sections (Erosion Polygons) (Figure 16). An area proximal to each erosion polygon was then digitised and the pixels values of the DEM within the Estimate Polygons (Figure 16) extracted and an average elevation calculated. The average elevation of an Estimate Polygon was the subtracted from each pixel elevation of the corresponding Erosion Polygon. The total volume of erosion was then determined from the sum of the differences between all erosion polygon pixel elevations and the average polygon elevation for the adjacent polygon. The total volume of erosion for the gully complex (ABGC1) was determined to be 6450.99 m3 (+/- 1612 m3), or 224 t/ha/yr +/- 56 t).

Figure 15 Comparison between differential GPS gully scarp surveys completed in 2006 and 2014

Figure 19 DGPS gully surveys 2006 & 2014 from site G9A on the upper Gilbert River (indicated in pink on Figure 15) Figure 16 Change detection procedure based on 2008 LiDAR data from the BelBel crossing site Table 8 Change in total area of the gully at site G9A 2006-2014 Table 7 Change in total are of the gully at BelBel Crossing on the Einasleigh River 2006-2014

6.3.3 Terrestrial LiDAR scanning of gully systems

High resolution scans of alluvial gullies were undertaken14 at selected sites previously surveyed, in order to both quantify the extent of erosion, but also to better understand the relative contributions from different parts of the gully complex. The following example from the Greenhills area on the Gilbert River (see Figure 15) highlights the extent of sediment production from a large alluvial gully complex, and enables us to differentiate the relative contribution from the gully head and from within the gully complex. The average specific yield from this section of the much larger gully complex (Figure 20) is 2164 t/ha/yr, with 44.8% of the sediment yield from the past 8 years derived from head-scarp extension, and 55.2% derived from the internal portion of the pre-existing gully. The data indicates (Table 9) that in this highly active system the gully is expanding at a rate of around 5% per annum. While the sampled gully is only around 0.2ha in size, it is a part of a larger gully complex with a total area of around 15 ha. Hence, even accounting for the fact that the sampled section of this gully may be the most active part of the gully, and hence the overall average sediment yield might only be around 1000 t/ha (which is in line with similar gullies in the Mitchell catchment – see Shellberg et al., 2013) it is likely that this gully complex is delivering, on average, more than 15,000 t/ year to the channel network.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion

23 sediment yield might only be around 1000 t/ha (which is in line with similar gullies in the Mitchell catchment – see Shellberg et al., 2013) it is likely that this gully complex is delivering, on average, more than 15,000 t/ year to the channel network.

Figure 20 Image of the surrounding broader alluvial gully complex at Greenhills showing the gully segment for which a high resolution 5cmFigure DEM was 19 derivedImage of from the asurrounding Leica C10 terrestrial broader LiDAR. alluvial The gully yellow complex box highlightsat Greenhills the showingscanned sectionthe gully of segmentthe gully complexfor which a high resolution 5cm DEM was derived from a Leica C10 terrestrial LiDAR. The yellow box highlights the scanned section of the gully complex

FigureFigure 21 Gully 20 G9BGC1Gully G9BGC1showing some showing of the scanned some images of the derived scanned from the images scans of derivedthe area highlighted from the in Figurescans 20 of the area highlighted in Figure 19. Figure 20 Gully G9BGC1 showing some of the scanned images derived from the scans of the area highlighted in Figure 19.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion

Table 9Table Summary 9 Summary statistics statistics showing showing the volume the volume of erosion of erosion Gully Extension Area Volume Eroded 1027 m3 (44.8%) area 597 m2 (28.2%) Specific yield 3,440 t/ha/yr (BD = 1.6) Gully Internal Area Volume Eroded 1267 m3 (55.2%) (55.2%) area 1523 m2 (71.8%) Specific yield 1,664 t/ha/yr (BD = 1.6) Total area of gully change (sum of two above) Volume Eroded 2294 m3 (100%) area 2120 m2 (100%) Specific yield 2164 t/ha/yr (BD = 1.6)

Figure 22Figure 5cm pixel 21 resolution 5cm pixel DEM resolution of the scanned DEM gully atof Greenhills the scanned showing gully the area at of Greenhills gully extension showing in the 8 year the period area of from Octobergully 2006 extensionto October 2014. in the Of the 8 totalyear sediment period exported from Oct fromober this gully 2006 over to this October period, 44.8% 2014 was. Of derived the from total headscarp extension and the remaining 55.2% from erosion of internal gully surfaces

34 Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 25

Figure 23 Elevation change image which shows the distribution of erosion from both gully headscarp retreat and erosion of internal gully slopes. Note that there is little net change from the gully floor during this period, which reflects the transport limited nature of alluvial gullies (sensu Rose et al., 2015).

6.3.4 Summary of rates data From these data it is apparent that alluvial gully systems within the central Gilbert are consistently producing significant volumes of sediment. Minimum estimates of the sediment yield from these gully complexes puts the rates at between 130 and 225 t/ha/yr, largely based on estimates from the headscarp retreat alone. As demonstrated elsewhere (Shellberg et al, 2013) the total yield from a large alluvial gully complex such as this is likely to be as much as 2 – 3 times higher than that from headscarp retreat alone, so if we take the lower estimate to account for the fact that some of the internal erosion has already been accounted for, the typical rates should fall somewhere between 260 - 450 t/ha/yr of sediment to the stream network (from the active gully area). Rates at the upper end of the range, as demonstrated by the data from G9GC1 at Greenhills, are producing specific yields of 2160 t/ha/yr. As yet it is unclear how representative these rates are of the broader region, but as indicated in Figure 20, there are much greater areas in this general area of which these rates would appear to be typical. Based on the gully mapping and these updated erosion rate data, it is possible to estimate the total sediment yield from gullies in the mid Gilbert region. From the 2006 gully mapping (Brooks., et al., 2006) a total of 9980 ha of gullies were mapped from the areas covered by the ASTER imagery – an area extending from the Gilbert fan apex to around Forsyth in the upper catchment. If we assume that the mapping on average may be over estimating the total area of active gullies by a factor of two (i.e. where the over-estimation (4 fold) in some areas with many gullies is offset by under-estimation in others (10 fold)), giving a conservative estimate of around 5000 ha of active gully area within the catchment. Based on these figures, mean annual sediment contribution from gullies would fall within the range 1.3 – 2.3 Mt/year, most of which can be regarded as land-use accelerated erosion. If the observed specific yields in the order of 2000 t/ha/yr are representative of a significant proportion of the mapped gullies, the mean annual sediment yields from gullies could be several times more than this conservative range.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 26

6.4 Soil suitability and the CSIRO FGARA Report

The major series of studies and associated reports completed by CSIRO (Petheram et al., 2013) into the agricultural potential of the Flinders and Gilbert River systems (The Flinders and Gilbert Agricultural Resource Assessment - or FGARA) was tasked with assessing the capacity of this landscape to support economically and environmentally sustainable intensive agricultural development. The assessment focused primarily on the four key attributes of the region and the potential constraints on, or issues around, irrigated agriculture:

• The suitability of soils to support a range of intensive agricultural activities • The availability and reliability of water resources to support the development, and the potential viability of various irrigation development schemes • The economic viability of any development given the constraints on market access and associated constraints • Potential negative impacts on the environment and on existing industry (e.g. the commercial fisheries in the Gulf of Carpentaria).

Given that this report is focused on avoiding potential risks associated with development on unsuitable soils, for the purposes of this report we will focus on summarising just those aspects of the FGARA study that dealt with soils in the Gilbert catchment. This will form a useful background for further analysis of the suitability of soils in the proposed IFED area. The reader is referred to the FGARA reports for further information on the other three aspects of agricultural development in the region. The key focus in this review will centre on the area outlined for prospective development under the IFED scheme in the area bounded by the Einasleigh, Etheridge and Gilbert Rivers.

6.4.1 Soil suitability assessment

Given the vast area of the Gilbert catchment and the limited time available for undertaking the soils suitability analysis within the FGARA, the analysis by necessity was required to adopt what is referred to as a digital soil mapping approach (Petheram, 2013). This approach uses a spatial statistical modelling approach and relies heavily on remotely sensed data, particularly aerial radiometric survey (Figure 24), coupled with the 1 arcsec Dem (~30m) and limited field survey (Figure 25). Within the context of a simplified regional geology (Figure 26, Figure 27), a picture of the region’s soils and their appropriateness or otherwise for various cropping scenarios is derived using a spatial geo-statistical modelling approach. The final product comes in the form of a 90m resolution raster map which assesses the suitability of a vast array of potential crops under different irrigation scenarios (e.g. furrow, trickle, spray) (Figure 28). The derived maps are classified into 5 suitability classes, as outlined in Table 10.

The suitability framework makes the following assumptions for irrigated crops: (i) unlimited irrigation water is available to the cropping location; (ii) spray irrigation systems can deliver irrigation rates of up to 16-17 mm/day during periods of peak demand; (iii) spray irrigation systems can deliver irrigation rates of 24-25 mm/day for sugarcane crops during periods of peak demand; (iv) the applied irrigation water infiltrates sufficiently well to maintain soil water in the rooting profile sufficient for crop growth (crops not stressed); (v) spray irrigation scheduling allows sufficient time for regular maintenance; (vi) the amount of water applied does not drain below the rooting depth (CSIRO, 2012).

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 27

the materials present. (source CSIRO 2012 fig 3.4)

Figure 24 Ternary image of merged regional radiometrics Figuredatasets for23 the Ternary Flinders andimage Gilbert of catchments merged (outlined regionalas black polygons) radiometrics and their surroundings. datasets for The the merged Figure 25 Map showing the distribution of field soil samples used radiometric datasets are presented as a ternary (red, green, in the digital soils mapping exercise for the FGARA digital soils Flindersblue) image, and with Gilbert potassium catchments specified in the (outlined red channel, mapping exercise asthorium black in thepolygons) green channel and and their uranium surroundings. in the blue Thechannel. merged This image radiometric indicates significant datasets variability are in the alluvial cover across the two catchments, reflecting a varied presentedsource for the materialsas a ternary present (red, (source green, CSIRO 2012 blue) fig 3.4). image, with potassium specified in the red channel, thorium in the green channel and Figure 24 Map showing the distribution of uranium in the blue channel (Figure 3.4). field soil samples used in the digital soils This image indicates significant variability mapping exercise for the FGARA digital in the alluvial cover across the two soils mapping exercise. catchments, reflecting a varied source for

FigureFigure 26 25 Broad Broad geological geological units described units in describedthe CSIRO FGARA in the reporting. CSIRO Note FGARA the Karumba reporting. formation Note depicted the in Karumbayellow, formationwhich coincides depicted with the Holroyd in yellow, Plain unit which mapped coincides in the original with Qld the Govt Holroyd geological Plainmapping unit of the mapped region (Grimes in the and originalDouch, 1978) Qld , which Govt is thegeological original depositional mapping surface of the of the region palaeo (GrimesGilbert magefan and Douch, 1978) , which is the original depositional surface of the palaeo Gilbert37 magefan.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion

Figure 26 Physiographic map of the Gulf plains from Grimes and Doutch (1978) showing the Figure 27 Physiographic map of the Gulf plains from Grimes and Doutch (1978) showing the location of the Holroyd Plain which Figureis thelocation unit 26 that Physiographic of is synonymousthe Holroyd with map Plain the of Karumba whichthe Gulf Formation is theplains unitas representedfrom that Grimes is bysynonymous the and CSIRO Doutch mapping with (1978)(Figure the 26)Karumba showing the locationFormation of the as Holroyd represented Plain bywhich the CSIROis the unitmapping that is (Figure synonymous 25). with the Karumba Formation as represented by the CSIRO mapping (Figure 25).

Figure 27 Example of the digital soils suitability mapping produced for the FGARA.

FigureFigure 28 27 Example Example of the of digital the soils digital suitability soils mapping suitability produced mapping for the FGARA produced for the FGARA.

39 39 Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion

TableTable 10 Suitability 10 Suitability classification classification applied in the appliedFGARA in the FGARA

6.4.1.16.4.1.1 SoilSoil Suitability Suitability Analysis Analysis - Overview - Overview AA broad broad range range of of parameters parameters were were considered considered in in the the soil soil suitability suitability analysis, analysis, including: including: • furrow irrigation efficiency (which is a proxy for soil infiltration characteristics), • furrow irrigation efficiency (which is a proxy for soil infiltration characteristics), • microrelief, which is primarily intended to represent the presence of gilgai which can lead to • microrelief, which is primarily intended to represent the presence of gilgai which can lead to pondingponding and and poor poor paddock paddock drainage, drainage, as aswell well as asbeing being an an indicator indicator for for the the potential potential presence presence of of sodic sodic soils soils.. A A threshold threshold of of 30cm 30cm of of micro micro-relief-relief was was adopted adopted as as a a threshold threshold for for consideration, consideration, althoughalthough it isit notis not clear clear how how this this was was observed observed from from the the 1 sec1 sec DEM, DEM, given given that that this this is wellis well below below the the resolutionresolution of oftopographic topographic variability variability that that can can be be detected detected from from such such data. data. There There would would appear appear to tobe be a contradictiona contradiction between between being being able able to todetect detect the the presence presence of ofgilgai gilgai but but not not to tobe be able able to tomeasure measure landscapelandscape complexity complexity (below). (below). • •NutrientNutrient balance balance - pH - pH • Rockiness (e.g. the presence of gravels, stones and boulders) • Rockiness (e.g. the presence of gravels, stones and boulders) • Soil depth • •SoilSoil depth physical condition • •SoilSoil physical moisture condition availability • Surface soil erodibility (as represented by the USLE K factor ; Wischmeir and Smith, 1975; Rosewell • Soiland moistur Loch,e 2002) availability (Table 11) • •SurfaceWind soilerosion erodibility (as represented by the USLE K factor; Wischmeir and Smith, 1975; Rosewell and Loch, 2002) (Table 11) Factors not assessed include, flooding, salinity, secondary salinization and landscape complexity. • Wind erosion. FactorsTable not11 The assessed four soil include, stability flooding, categories salinity, based secondary on K -salinizationfactor used and in landscapethis study complexity. (Rosewell and Loch, 2002) Table 11 The four soil stability categories based on K-factor used in this study (Rosewell and Loch, 2002)

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion Most of these soil characteristics upon which the analysis is based can only be reliably determined from field survey, and hence the region-wide extrapolation of these parameters is highly dependent on the statistical representation between the soil survey data and the remote sensing data. Hence, even with a perfect relationship between the site data and the remotely sensed data at a site (which is extremely unlikely), any misrepresentation or inaccuracies in the way that the remotely sensed data represents the ground conditions will lead to compounding errors in the final representation of the soil suitability predictions.

Within the IFED focal area it would appear that there is potential for significant misrepresentation of the soil suitability classification due to some of the inherent characteristics of this part of the landscape, in large part due to the geological province upon which the site is located. As shown in Figure 28 and Figure 29 it is apparent that due to the presence of fericrete and other laterites within these Plio- Miocene age (2.58 – 23.03 Ma) fan depositional units (variously described as the Karumba Province, the Holroyd Plain or the Wyaaba Beds), that the radiometric data and hence the soil suitability mapping, are likely to give spurious results in this area. Hence, without considerable investment in extensive field mapping of the soils within the Karumba province (i.e. outside of the areas that have had detailed field sampling - Figure 24), considerable caution should be taken in assuming that the digital soils mapping predictions provide a useful guide for planning irrigated agricultural development. Even within these constraints (which have a tendency to paint this particular area in the most positive light possible), it is apparent (e.g. Figure 27) that for traditional furrow irrigation of crops like sugar cane, there are virtually no soils which fall into suitability classes 1 – 3.

Figure 28 Maps from the FGARA digital soils mapping which provide estimates of the confidence in the mapping outputs. Note the apparent contradiction between the two maps,

Most of these soil characteristics upon which the analysis is based can only be reliably determined from fieldMost survey, of these and soil hence characteristics the region- uponwide whichextrapolation the analysis of these is based parameters can only is behighly reliably dependent determined on the statisticalfrom field representation survey, and hencebetween the theregion-wide soil survey extrapolation data and the of remote these sensingparameters data. is highlyHence, dependent even with a perfecton the relationshipstatistical representation between the betweensite data andthe soilthe surveyremotely data sensed and the data remote at a site sensing (which data. is extremely Hence, unlikely)even with, any a perfectmisrepresentation relationship or between inaccuracies the site in thedata way and that the the remotely remotely sensed sensed data data at a represents site (which the groundis extremely conditions unlikely), will leadany misrepresentationto compounding errorsor inaccuracies in the final in representationthe way that the of remotelythe soil suitability sensed data represents the ground conditions will lead to compounding errors in the final representation of the soil predictions. suitability predictions. Within the IFED focal area it would appear that there is potential for significant misrepresentation of Within the IFED focal area it would appear that there is potential for significant misrepresentation the soil suitability classification due to some of the inherent characteristics of this part of the landscape, of the soil suitability classification due to some of the inherent characteristics of this part of the in large part due to the geological province upon which the site is located. As shown in Figure 28 and landscape, in large part due to the geological province upon which the site is located. As shown Figure 29 it is apparent that due to the presence of fericrete and other laterites within these Plio- in Figure 29 and Figure 30 it is apparent that due to the presence of fericrete and other laterites Miocene age (2.58 – 23.03 Ma) fan depositional units (variously described as the Karumba Province, the within these Plio-Miocene age (2.58 – 23.03 Ma) fan depositional units (variously described as the HolroydKarumba Plain Province, or the Wyaabathe Holroyd Beds) Plain, that or thethe radioWyaabametric Beds), data that and the hence radiometric the soil suitabilitydata and hencemapping, the are likelysoil suitabilityto give spurious mapping, results are likelyin this to area. give spuriousHence, without results inconsiderable this area. Hence,investment without in extensive considerable field mappinginvestment of thein extensive soils within field the Karumbamapping of province the soils (i.e. within outside the ofKarumba the areas province that have (i.e. hadoutside detai ofled the field samplingareas that - Figurehave had 24), detailed considerable field cautionsampling should - Figure be 26),taken considerable in assuming caution that the should digital be soils taken mapping in predictionsassuming thatprovide the adigital useful soils guide mapping for planning predictions irrigated provide agricultural a useful development. guide for planning Even withinirrigated these constraintsagricultural (which development. have a tendency Even within to paint these this constraints particular (which area in have the amost tendency positive to lightpaint possible), this particular it is apparentarea in the (e.g. most Figure positive 27) that light for possible), traditional it is furrow apparent irrigation (e.g. Figure of crops 28) likethat sugar for traditional cane, there furrow are virtually nirrigationo soils which of crops fall into like suitabilitysugar cane, classes there 1 are – 3. virtually no soils which fall into suitability classes 1 – 3.

Figure 28 Maps from the FGARA digital soils mapping which provide estimates of the Figure 29 Maps from the FGARA digital soils mapping which provide estimates of the confidence in the mapping outputs. Note confidencethe apparent contradiction in the mapping between outputs. the two maps, Note where the map apparent (a) indicates contradiction high to very high confidencebetween in the the twomap predictions maps, in the area immediately above the fan apex, whereas map (b) indicates that because of the nature of the geological province (Karumba province or Holroyd Plain – depending on which geologic map you use) that it is likely that the radiometric data will give spurious results in this area hence leading to inaccurate soil41 suitability predictions

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 31

It is also worth pointing out that the way in which soil erodibility is assessed within the soil suitability assessment only considers the role of surface erosion, as a function of the USLE K factor and the land slope. The K factors assessed only represents the erodibility of the surface soils, not the sub-surface soils, which are the primary determinant of whether the landscape is susceptible to gully erosion. As highlighted above, and explored further below, gully erosion is the dominant erosion process in this landscape, particularly erosion into alluvial surfaces deposited within Quaternary age floodplains, or into the old fan unit depositional surfaces. Sediment tracing within the adjacent Mitchell and Flinders catchments (Caitcheon et al., 2012) demonstrated that sub-surface sediment sources (gully and stream bank erosion) represented 97% and 100% respectively of the sediment load entrained within the main channels of these two catchments. While the equivalent tracing study has not been undertaken in the Gilbert, there are many similarities between the Gilbert megafan and the Mitchell megafan (see Brooks et al., 2009), and hence it is highly likely that a similar ratio of surface/sub-surface erosion as observed in the Mitchell occurs in the Gilbert. Hence, to understand the susceptibility of this landscape to erosion, one has to understand the primary drivers of gully and channel erosion, which can only be done by placing the contemporary landscape within the context of the long term evolution of the entire alluvial megafan, as further detailed in Section 7.1.2.

Figure 30 Blow up of the Gilbert Fan area depicted in Figure 29 which shows that the IFED agricultural precinct is located within the area that is depicted as having low predictive power using the digital soils mapping approach.

Figure 29 Blow up of the Gilbert Fan area depicted in Figure 28 which shows that the IFED agricultural precinct is located within the area that is depicted as having low predictive power Developing guidelines for agriculture developmentusing inthe northern digital Australia soils to avoidmapping and reduce approach. accelerated erosion

7 The IFED proposal 7 The IFED Proposal 7.1 Overview of proposal 7.1.1 Overview of Proposal TheThe IFEDIFED proposal proposal is is an an integrated integrated food food and and energy energy production production development development which which has as has its ascentre its centre a majora major water water resource resource infrastructure infrastructure development development coupled coupled to a proposed to a proposed 65,000 65,000 ha irrigated ha irrigated agriculturalagricultural precinct. precinct. The The irrigation irrigation area area is proposedis proposed to supportedto supported extensive extensive areas areas of sugar of sugar and guar and guar beansbeans (for(for guar guar gum). gum). The The precinct precinct will will include include a sugar a sugar mill mill and and an electricity an electricity co-generation co-generation plant plantwhich willwhich be fuelledwill be withfuelled the with waste the products waste productsfrom the sugarfrom themill. sugar The proposalmill. The is proposala fairly radical is a fairly departure radical fromdeparture usual frompractice usual in major practice irrigated in major agriculture irrigated developments agriculture ,developments, which normally which identify normally areas of identify highly suitableareas of soils highly and suitable design thesoils infrastructure and design thearound infrastructure the supply ofaround water the to thesesupply areas of water of productive to these areas of soils.productive In this soils. case the In thisprimary case focus the primary was on focusthe design was ofon the the water design harvesting of the water and storageharvesting and storage infrastructureinfrastructure such such that that it itwould would be be able able to togravity gravity feed feed the thewater water to the to agriculturalthe agricultural precinct. precinct. The The proposalproposal inc incorporatesorporates a apractice practice called called fertigation, fertigation, a process a process where where all plant all plant nutrients nutrients will be will delivered be delivered toto thethe cropscrops within within the the water water supply, supply, delivered delivered to tothe the crops crops via viasub -sub-surfacesurface trickle trickle irrigation. irrigation. The The locationlocation of of the the agricultural agricultural area area was was thus thus dictated dictated by theby thelocation location of the of waterthe water infrastructure, infrastructure, with soilwith soil fertilityfertility andand produc productivitytivity a alow low order order consideration consideration.. In essenceIn essence this thisis a processis a process similar similar to large to largescale scale hydroponics,hydroponics, where where it itis isassumed assumed that that the the only only purpose purpose of theof thesoil soilis to is provide to provide the physical the physical substrate substrate to supportto support the theplants. plants. Such Such an approach, an approach, however, however, requires requires soils that soils have that uniform have uniform slopes (i.e.slopes requiring (i.e. laserrequiring levelling) laser and levelling) that will and not that wash will away not underwash intenseaway under monsoonal intense rains. monsoonal In the following rains. In section the following we willsection assess we whether will assess a sufficient whether extent a sufficient of such soilsextent exist of withinsuch soils this area,exist andwithin what this some area, of and the what some of managementthe management concerns concerns may be may within be within this area. this area.

FigureFigure 31 30 Map Map of the of proposed the proposed IFED development IFED development showing the water showing infrastructure the water and the infrastructure location of the agricultural and the locationprecinct. Shown of the here agricultural are the proposed precinct. off-stream Shown dams which here willare extract the proposed around 40% off of the-stream mean annual dams high which flows from willthe Einasleigh extract Riveraround at Dagworth 40% of (550,000 the mean Ml per annual annum), highwhere flows it will be fro storedm the in DagworthEinasleigh dam River and transferred at to Dismal lake on Huonfles Station via an open channel. The Dismal lake storage is also fed by the local Dismal Creek catchment, which Dagworthis a significant (550,000 catchment Ml in itsper own annum) right. ,Also where shown it arewill the be ~1600 stored x 40 hain plotsDagworth that will dambe the andfocus transferredfor the agricultural toprecinct Dismal on Kutchera lake on & HuonflesChadshunt Stations. Station Source via an http://www.agriculture.gov.au/abares/outlook-2014/Documents/presentation-slides/ open channel. The Dismal lake storage is also fed bykeith-deLacy-presentation.pdf the local Dismal Creek catchment, which is a significant catchment in its own right. Also shown are the ~1600 x 40 ha plots that will be the focus for the agricultural precinct on

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 33

7.2 Geological and topographic context for the IFED site

In order to assess the suitability of the soils within the IFED area, we firstly need to understand the landscape evolutionary history of the site. Placing the site within its landscape evolutionary context can provide understanding of whether the site is comprised of recently deposited alluvium (the typical location of irrigated agriculture, other than volcanic soils), and if not what the dominant geology is and what it’s weathering and erosional history has been.

Perhaps the defining characteristic of the Gulf Rivers between the Holroyd and the Norman Rivers is that this entire landscape consists of a sequence of merged alluvial megafans (sensu Leier et al., 2005), dominated by the Mitchell and the Gilbert megafans. As can be seen in Figure 32 the IFED agricultural precinct is located on the upper (and oldest) depositional units of the Gilbert megafan.

Figure 32 Digital elevation model (DEM) of Cape York and the Southern Gulf (LHS) showing the IFED irrigation area (black rectangle) within the upper reaches of the Gilbert megafan, and within the Plio/Miocene age Wyaaba beds (RHS) - which are some of the oldest alluvial sedimentary deposits within the Gilbert megafan. These units are actively incising and have been weathered in-situ for more than 2.6 million years

Looking more closely at this region of the Gilbert megafan (Figure 33) it is evident that the three major river systems in the Gilbert megafan (the Gilbert, Etheridge and Einasleigh Rivers) have incised into this upper portion of the megafan since it was deposited in the Miocene, isolating the Dismal Creek catchment into what is now a completely separate sub-catchment within the broader megafan. This has had the effect of separating this area from the upper source areas of the broader Gilbert megafan catchment starving it of “fresh” alluvium from the major volcanic and igneous source zones of the upper catchment. Hence, the Dismal Creek catchment has largely been eroding, weathering, and redistributing the old alluvial fan sediments that were deposited several millions of years ago. It is for this reason that this landscape unit is characterised by extremely weathered lateritic soils which have been leached and weathered in the monsoonal savannah environment for millions of years, leaving extremely low nutrient, and potentially unstable sodic soils (discussed further below).

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 34

Figure 33 A DEM of a section of the upper Gilbert Fan showing the highly dissected topography encompassed within the IFED irrigation area. Also shown is the highly dissected nature of the drainage network with the irrigation area. The Irrigation area falls within the Dismal Creek catchments, which is a relatively steep, well developed catchment in its own right, set within the broader context of the Gilbert Fan. As can be ascertained from this image, this area is not a flat alluvial plain that one might expect in the typical irrigation areas found on the lower Burdekin or the Murray Darling.

Also evident from this blow up image of the megafan topography, is that far from being a relatively flat floodplain system, as most irrigated agricultural areas tend to be whether they are in the tropics (eg. the Burdekin) or the temperate zone (Murray-Darling), this area is highly dissected, forming a complex dendritic stream network in its own right. As can be seen from Figure 34, which shows a series of topographic cross sections through the proposed IFED agricultural precinct with the Dismal Creek catchment, there is considerable relief, both across the fan (dissecting the channels) and longitudinally down the major tributary systems within the zone (Figure 35). Cross valley slopes (i.e. between the interfluves of the adjacent creeks and the creek line) are typically in the order of 0.3 - 0.6%, while longitudinal (down valley) slopes are in the order of 0.14-0.18 %. This is around three times the down-valley slope on the Burdekin irrigation area and around 10 times the typical cross floodplain slopes in the Burdekin.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 35

Figure 34 Cross sections of the Dismal Ck catchment showing the considerable relief (> 20m) between the ridges separating consecutive creek lines. Cross section locations are shown on the 30m DEM

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 36

Figure 35 Longitudinal profiles of the major creek lines running through the Dismal Ck catchment showing the relief encompassed within the irrigation area (indicated in red ), and the relatively steep gradient of the streams within the irrigation area (~ 0.14 – 0.18%)

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 37

8 Field reconnaissance of the IFED agricultural precinct

While it was not within the scope of this project to undertake detailed soils mapping within the agricultural precinct, and hence to comprehensively assess the suitability of the landscape for irrigated agriculture, we have been able to undertake a preliminary field reconnaissance which enables us to highlight some of the inherent land degradation risks associated with the intensive development in this area. As part of this analysis we undertook some initial soil chemistry analyses in areas of know erosion hazard (i.e. existing gully erosion sites) (Figure 36). While it is acknowledged that a much more comprehensive analysis of soil chemistry would be required to fully assess the erosion risks of this landscape, even this preliminary data enables us to highlight some of the inherent risks, particularly when we place this in the context of the regional landscape evolutionary history, the topographic constraints, and the distribution of existing gully erosion. It should be stressed that the current landuse is relatively low intensity grazing, and so the extant gullies would represent the starting point from which we could expect considerable expansion under a land-use intensification scenario, unless careful measures are put in place to minimise the threat.

Figure 36 Map showing the distribution of existing gullies (red polygons) within the Dismal Creek catchment. Most of these existing gullies are alluvial gullies that are located within or near to drainage lines in the area

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 38

8.1 Current gully extent

A total of 106 ha of gullies were mapped within IFED agricultural precinct, which represents around 0.16% of the total land area. Whilst this is below the average for the broader Gilbert study area, as can be seen in Figure 36, the gullies are distributed throughout the site, and are prevalent in areas in and around drainage lines and water courses. There are also many examples of gullies initiated by road drains, which is one of the common initiation mechanisms for gullies in this landscape, along with cattle pads through riparian areas (see Shellberg & Brooks., 2012, 2013). While the extent of gullies in this area is currently at moderate levels, the fact that they are distributed fairly evenly across this landscape, highlights the fact that there are plenty of susceptible soils in the area. Hence there is substantial risk of the extent and severity of gullies increasing if the proposed scale of intensive development proceeds. In particular, this development is based on the intensive cultivation of around 1600 small plots and therefore necessitates a very extensive network of roads and tracks through this landscape. Such a high density network of roads and tracks represents a significant threat to the initiation of gully erosion, given that it would be impossible to design a road network in addition to that 65,000 ha of irrigation, that doesn’t intersect numerous times with the relatively dense stream network in this area.

8.2 Soil chemistry and gully hazard

Soil chemistry data was collected in a pilot study focusing on gully erosion sites within the IFED Irrigation area. Initial soil chemistry analysis was undertaken at nine sites, with the results indicating considerable complexity in the soil chemistry. These preliminary results indicate that a much greater sampling density is required to understand the suitability of the soils within the area for agriculture, and their relative susceptibility to erosion. The variability of the soil chemistry between sites that are relatively close together, and all of which are eroding, indicates there is a complex interaction of factors driving the erodibility of these soils. The base saturation data associated with sub-surface soils at each of the sampled gully sites, shown in Figure 39 and Figure 40, indicate that no one soil chemistry indicator can be used as a predictor of soil erodibility. High exchangeable sodium percentage (ESP) is generally thought to represent erodible soils, whereas a high Ca/Mg ratio is thought to indicate a low susceptibility to erosion. These results raise questions about the validity of these assumptions in this landscape given that all of these samples come from sites that are experiencing active gully erosion. This is further demonstrated in the dispersion data also presented below (Figure 41) which indicates that dispersibility alone is not the sole arbiter of whether the soil is susceptible to gully erosion, and that there is not always a clear relationship between high ESP and erodibility. Less dispersive soils are also subject to gully erosion where there is sufficient flow concentration to drive gully incision and expansion. Some of these samples, however, are so dispersive that the Emmerson Aggregate Tests, which are usually run for 22 hours, were complete within 30 seconds. The most erodible samples (GIL10216) came from a gully that has no outlet connected to a stream network, and yet is evidently highly active, indicating that the material eroded within the gully is transported out of the gully in suspension via overland flow that passes through and over the gully. Such behaviour represents an extreme level of erosion hazard, which if more widespread would be a considerable cause for concern. Interestingly this sample had Ca/Mg ratio greater than 1.0 and a relatively low ESP, which ordinarily would place this soil in a fairly stable soil category. Hence, there is clearly much still to be learnt about the behaviour of soils in this landscape.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 39

Figure 37 Location of the sampled gully sites within the study area Figure 36 Location of the sampled gully sites within the study area

Figure 38 37Gully Gully soil sample soil samplesites within sites the IFED within irrigation the area IFED irrigation area

51 Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 40

Figure 38 Base saturation associated with sub-surface soils at each of the sampled gully sites. Figure 39 Base saturation associated with sub-surface soils at each of the sampled gully sites. These data indicate that no Theseone soil data chemistry indicate indicator that can beno used one as soila predictor chemistry of soil erodibility. indicator High can ESP be is generallyused as thought a predicto to representr of soilerodible erodibility.soils, whereas aHigh high Ca/Mg ESP isratio generally is thought thoughtto indicate ato low represent susceptibility erodible to erosion. soils, These resultswhereas raise questionsa high Ca/Mg about the ratiovalidity is of thought these assumptions to indicate in this alandscape low susceptibility to erosion. These results raise questions about the validity of these assumptions in this landscape. 90.0 Gilbert soil samples % sat Ca 80.0 % sat Mg 70.0 % sat K 60.0 % sat Na 50.0 (ESP)

40.0

30.0

20.0

10.0

0.0 GIL10211 GIL10212 GIL10213 GIL10214 GIL10215 GIL10216 GIL10217 GIL10218 GIL10219

Figure 40 39Base Base saturation saturation % values for % sample values sites for sample sites

52 Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion

slaked and and slaked

omment C Little evidence if slaking or dispersion 2 after mins after Fully slaked 60 secs Full y 60 after dispersed secs after Fully slaked 30 secs slaked Partially mins 2 after

Dispersion @ 120 secs Dispersion@ 120

Dispersion @ 90 secs Dispersion@ 90

54 Dispersion @ 60 secs Dispersion@ 60

Dispersion @30 secs Dispersion @30

Gully

- - sfce sfce

floor ~1m below orig Sample # GIL10211 Abingdon Gully 'A' Horizon soil GIL10212 Abingdon Gully 'B' Horizon soil; ~0.5m below orig sfce soil GIL10213 Abingdon Gully 'C' Horizon soil; soil GIL10214 Kutchera Rd Gully material GIL10215 Kutchera Rd Gully sub -

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 42

minor slakingminor

after 2 mins 2 after Fully slaked and 30 after dispersed secs Very Minimal slaking mins 2 after Significant slaking mins 2 after these samples are are samples these

usually run for 22 hours, were complete within 30 secs. The most erodible samples (GIL10216) come from a come (GIL10216) samples most erodible The 30 secs. within complete were hours, run 22 for usually

ow that passes through and over the gully. the over and passes through that ow Pictorial representation of the soil dispersion tests carried out at selected sites within the IFED irrigation area. Some of Some area. irrigation IFED within the sites selected carried at testsout dispersion soil the representationPictorial of

Gully with no outlet! GIL10216 Kutchera Rd Gully GIL10217 Kutchera Rd Gully GIL10218 Kutchera Rd Gully GIL10219 Kutchera Gully near Nthrn Entrance Rd. Figure 40 so dispersive that the tests, which are gully that has no outlet connected to a stream network, and yet is evidently highly active. Suspended sediment must be being removed via via removed being must be sediment Suspended active. highly evidently is yet network,stream and a connected to no outlet gully has that overland fl

Figure 41 Pictorial representation of the soil dispersion tests carried out at selected sites within the IFED irrigation area. Some of these samples are so dispersive that the tests, which are usually run for 22 hours, were complete within 30 secs. The most erodible samples (GIL10216) come from a gully that has no outlet connected to a stream network, and yet is evidently highly active. Suspended sediment must be being removed via overland flow that passes through and over the gully.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 43

8.2 Identify the alluvial stream and river channel network at highest available 8.3 Identifyresolution the &alluvial identifying str appropriateeam and buffers river channel network at highest available resolution and identifying appropriate buffersEphemeral stream channels like the one shown in Figure 41 are a common feature of the study area and collectively may well be a more dominant source of sediment to the channel network Ephemeralthan stream gully channels erosion. likeHence the anyone development shown in Figure activity 42 are that a mightcommon accelerate feature theof the erosion study of area and collectivelychannels may like well this be needs a more to bedominant carefully source controlled. of sediment It is also to clearthe channel that gully network erosion than is generally gully erosion. veryHence closely any developmentlinked to the streamactivity network,that might with accelerate many gullies the erosion initiating of channelsfrom disturbances like this needs in the to be carefullyriparian controlled. areas adjacent It is also to stream clear that lines. gully To erosion this end is wegenerally have proposed very closely a network linked to of the buffers stream be network,left with retained many gullies under initiatingany development from disturbances scenario which in the has riparian the effect areas ofadjacent minimising to stream new lines. To this enddisturbances we have proposed anywhere a networkclose to drainageof buffers lines, be left and retained provides under for maximum any development dissipation scenario of energy which hasfor the flows effect generated of minimising off cleared new disturbancespaddocks or anywherethe extensive close road to drainagenetwork thatlines, will and be provides required in for maximumthe proposed dissipation development of energy for. flows generated off cleared paddocks or the extensive road network that will be required in the proposed development.

Figure 42 EphemeralFigure 41 stream Ephemeral channels like stream this one channels in the Gilbert like catchment this one can contribute in the Gilbert significant catchment volumes of sediment can to the channelcontribute network, and significantfurther disturbance volumes to such of channels sediment from clearing, to the road channel crossings network, etc., could significantly and further increase this sedimentdisturbance input. to such channels from clearing, road crossings etc., could significantly increase this sediment input.

In order to minimise the risk of initiating sub-surface erosion around the stream network, buffers around all stream lines and existing gully areas need to be maximised. To test the viability of different buffers we modelled a range of buffer widths that are a function of stream order, according to the equation:

Where S0 > 1: Bw (m) = n x S0,

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 44

Where S0 > 1: Bw (m) = n x S0,

Where S0 = 1: Bw (m) = 2n x S0

Bw = buffer width (each side of stream), n = buffer multiplier (10 – 45m); and S0 = stream order (Strahler)

All mapped gullies also include a buffer of the same extent as the 1st and 2nd order streams.

We initially modelled the scenario where all stream orders where treated equally (i.e. no additional factor added to first order stream lines), and these results are presented as well (Table 12 and Figure 43), but given the large number of first order streams and the propensity for gullies to initiate in these first order stream we believe that the additional buffer around first order streams is necessary, and achievable. Examples of some of the buffer scenarios are shown in Figure 44 - Figure 46, which also provide an example of the 40 ha development plots for scale.

Based on the modelling, a buffer multiplier of 45m would appear to be achievable and would still leave around 75,000 ha of land that could potentially be developed, while minimising the threat of gully erosion initiation in and around the drainage network (Figure 47). Anything less than 30m is insufficient to provide an adequate buffer between the developed farm land and the associated infrastructure and the stream lines. Further work is required to refine the stream network given that this analysis was based on the available 1:100,000 scale stream network, and there are some drainage lines not captured in this analysis. However, the total area would not differ significantly from that represented here.

If it did transpire that there was a significant length of stream not represented in the 1:100K stream network it may be necessary to reduce the buffer multiplier, but we would not recommend that it be reduced to less the 30m.

The aim with these buffers would be to completely eliminate all land use pressure from these zones, including clearing, roads and tracks and disturbances associated with water infrastructure.

These buffers should be the minimum area left uncleared within the agricultural precinct. Additional wildlife corridors should be superimposed on this network to provide wildlife movement across the drainage network (i.e. in a NW/SE direction).

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 45

35% Stream Buffer Area 35% Stream Buffer Area 30% 30% Stream Buffer Area 25% 25% Stream Buffer Area Condition 1 Condition 1 20% 20% Linear (Stream 15% Linear (Stream 15% Buffer Area) Buffer Area) 10% 10% Linear (Stream 5% Linear (Stream 5% Buffer Area Buffer Area 0% Condition 1) 0% Condition 1) Percentage of Irrigation Area 0 Percentage of Irrigation Area 01010 20 20 30 30 40 40 50 50 Buffer DistanceBuffer Distance(m) (m)

Figure 42Figure Plot 42showing Plot showing the proportion the proportion of the IFED of the area IFED that area would that needwould to need be left to outbe left out of production to provide adequate buffers Figure 43 Plotof productionshowing the proportion to provide of the adequateIFED area that buffers would need to be left out of production to provide adequate buffers Table 12Table Effect 12 of Effect different of different buffer widths buffer onwidths the proportionon the proportion of land oftaken land out taken of out of Tableproduction. 12 Effectproduction. of different buffer widths on the proportion of land taken out of production.

ApproximateApproximate IFED IFED IrrigationIrrigation Area Area 2 (ha) (km ) 2 (ha) (km ) Total Total 103,539103,539 1,035 1,035 exclusionexclusion

Stream StreamBuffers BuffersArea Area Gully Buffer Gully Area Buffer Area

2 2 (ha) (ha) (km ) (km(ha)2) (ha) (km ) (km2) Stream Buffer (m) = X * S0 Stream Buffer (m) = X * S0 Gully BufferGully (m) Buffer = X (m) = X where Xwhere is the bufferX is the multiplier buffer multiplier 30 30 15,325 15,325153.25 153.25 324 3243.24 3.2415% 15% 35 35 17,671 17,671176.71 176.71 366 3663.66 3.6617% 17% 40 40 19,963 19,963199.63 199.63 408 4084.08 4.0820% 20% 45 45 22,210 22,210222.10 222.10 453 4534.53 4.5322% 22% Where S0 = 1; Buffer (m) = X * 2 * Where S0 = 1; Buffer (m) = X * 2 * S0, S0, Where S0 > 2; Buffer (m) = X * S0 Where S0 > 2; Buffer (m) = X * S0 Gully BufferGully (m) Buffer = X *(m) 2 = X * 2 where Xwhere is the bufferX is the multiplier buffer multiplier

30 30 19,628 19,628196.28 196.28 594 5945.94 5.9420% 20% 35 35 22,647 22,647226.55 226.55 695 6956.95 6.9523% 23% 40 40 25,618 25,618256.18 256.18 802 8028.02 8.0226% 26% 45 45 28,522 28,522285.22 285.22 914 9149.14 9.1428% 28%

59 Developing guidelines for agriculture development in northern Australia to avoid59 and reduce accelerated erosion

FigureFigure 44 Example 43 Example of two buffering of two options buffering within the options IFED area within(20x stream the order IFED (darker area colour) (20x & stream 45 x stream order order). Yellow(darker boxFigure is acolour) 40 ha43 plot Example& for 45 scale x stream of two order). buffering Yellow options box iswithin a 40 hathe plot IFED for area scale. (20x stream order (darker colour) & 45 x stream order). Yellow box is a 40 ha plot for scale.

Figure 45 Blow up showing the 20m buffer and the 45m buffer. The yellow box represents a 40 ha plot 60 60

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 47

Figure 46 Blow up of a section of the Kutchera ag area showing the proposed 45m buffer (i.e. 90m for 1st and 2nd order stream lines). The box here represents the size of the 30m pixels used in the FGARA soil suitability analysis.

Figure 47 Map of the whole IFED showing the proposed minimum buffer (45m x stream order; with 1st order streams = 2nd streams). For scale a 40 ha plot is shown in the centre of the image

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 48

8.4 Potential offset areas

One of the aims of this research was to identify potential areas that could be rehabilitated as a means of offsetting some of the inevitable impacts that will flow from a major development such as that proposed in the IFED. Alluvial gully erosion represents the largest source of accelerated erosion in the Gilbert and many other catchments in northern Australia (Shellberg & Brooks, 2012), and it has been demonstrated that many gullies will continue to supply sediment at equivalent rates to those observed over recent years and decades for many hundreds if not thousands of years. (e.g. Shellberg, 2011). By comparison, sediment supply rates from hillslope gullies, such as those observed in the Georgetown granites, are likely to be starting to decline, in the same way that the hillslope gullies in the Burdekin have been shown to be well into the non-linear phase of negative exponential decay (Wilkinson et al., 2013). For this reason, major reductions in sediment yield are most likely to be achieved through active management of active alluvial gullies, which if left unchecked will continue to contribute excessive sediment to the stream network for the foreseeable future (on management timelines). Investment in managing the majority of hillslope gullies, will not be cost effective given that most are declining anyway. Based on the reasoning outlined above, the largest aggregation of alluvial gullies that could be managed with appropriate investment, are those along the Einasleigh River in the vicinity of the Confluence with the Etheridge River (Figure 48). A variety of approaches are appropriate here, although experience from elsewhere (Shellberg and Brooks, 2013), indicates that active remediation will be required if significant reductions in erosion rates are to be achieved.

Figure 48 Map showing the aggregation of alluvial gullies in the area around the confluence of the Einasleigh and Etheridge RiversFigure and on 47 the GilbertMap showingRiver in the Greenhillsthe aggregation areas. Note theof gullyalluvial mapping gullies shown in here the is areathe earlier around remote the sensing-based mappingconfluence which has ofbeen the shown Einasleigh to over represent and Etheridge the extent of Rivers gullies in and the areason the of high Gilbert gully concentration. River in the Hence the actual extent of gullies is less than represented here, but we do know that there are some major gully complexes here that couldGreenhills be managed areas with appropriate. Note the investment. gully mapping shown here is the earlier remote sensing- based mapping which has been shown to over represent the extent of gullies in the Developingareas guidelines of high for gully agriculture concentration. development in northern Hence Australia the toactual avoid and extent reduce accelerated of gullies erosion is less than represented here, but we do know that there are some major gully complexes here that could be managed with appropriate investment.

9 References

Bartley R, Thomas MF, Clifford D, Phillip S, Brough D, Harms D, Willis R, Gregory L, Glover M, Moodie K, Sugars M, Eyre L, Smith DJ, Hicks W and Petheram C (2013) Land suitability: technical methods. A technical report to the Australian Government for the Flinders and Gilbert Agricultural Resource Assessment (FGARA) project, CSIRO. Brooks, A. Borombovits, D., Spencer, J., Pietsch, T., Olley, J., (2014a). Measured hillslope erosion rates in the wet-dry tropics of Cape York, northern Australia Part 1: A low cost sediment trap for measuring hillslope erosion in remote areas - trap design and evaluation Catena. Brooks, A. P., Shellberg, J. G., Knight, J., & Spencer, J. (2009). Alluvial gully erosion: an example from the Mitchell fluvial megafan, Queensland, Australia. Earth surface processes and landforms, 34(14), 1951. Brooks, A. Spencer, J., Borombovits, D., Pietsch, T., Olley, J., (2014b). Measured hillslope erosion rates in the wet-dry tropics of Cape York, northern Australia: Part 2, RUSLE-based modeling significantly over-predicts hillslope sediment production Catena.

9 References

Bartley, R., Thomas, M.F., Clifford, D., Phillip, S., Brough, D., Harms, D., Willis, R., Gregory, L., Glover, M., Moodie K., Sugars, M., Eyre L., Smith, D.J., Hicks, W. and Petheram, C. (2013) Land suitability: technical methods. A technical report to the Australian Government for the Flinders and Gilbert Agricultural Resource Assessment (FGARA) project, CSIRO.

Brooks, A. Borombovits, D., Spencer, J., Pietsch, T., Olley, J., (2014a). Measured hillslope erosion rates in the wet-dry tropics of Cape York, northern Australia Part 1: A low cost sediment trap for measuring hillslope erosion in remote areas - trap design and evaluation Catena.

Brooks, A. P., Shellberg, J. G., Knight, J., & Spencer, J. (2009). Alluvial gully erosion: an example from the Mitchell fluvial megafan, Queensland, Australia. Earth surface processes and landforms, 34(14), 1951.

Brooks, A. Spencer, J., Borombovits, D., Pietsch, T., Olley, J., (2014b). Measured hillslope erosion rates in the wet-dry tropics of Cape York, northern Australia: Part 2, RUSLE-based modeling significantly over-predicts hillslope sediment production Catena.

Brooks, A., Spencer, J., & Knight, J. (2007, May). Alluvial gully erosion in Australia’s tropical rivers: a conceptual model as a basis for a remote sensing mapping procedure. In Proceedings of the 5th Australian Stream Management Conference. Australian rivers: making a difference (pp. 43-48).

Brooks, A.P., Spencer, J., Olley, J., Pietsch, T., Borombovits, D., Curwen, G., Shellberg, J., Howley, C., Gleeson, A., Simon, A., Bankhead, N., Klimetz, D., Eslami-Endargoli, L., Bourgeault, A., (2013) An Empirically-based Sediment Budget for the Normanby Basin: Sediment Sources, Sinks, and Drivers on the Cape York Savannah. Griffith University, 506pp. http://www.capeyorkwaterquality.info/references/ cywq-229

Caitcheon, G. G., Olley, J. M., Pantus, F., Hancock, G., & Leslie, C. (2012). The dominant erosion processes supplying fine sediment to three major rivers in tropical Australia, the Daly (NT), Mitchell (Qld) and Flinders (Qld) Rivers. Geomorphology, 151, 188-195.

CSIRO (2012) Proposed project methods. A report to the Australian Government from the CSIRO Flinders and Gilbert Agricultural Resource Assessment, part of the North Queensland Irrigated Agriculture Strategy. CSIRO Water for a Healthy Country and Sustainable Agriculture flagships, Australia.

Gulf Savannah Development (2009) http://www.cairnsinvestment.com.au/resources/research- hub/11/16/doc-164 (accessed 7-7-2015)

IFED (2012) http://i-fed.com.au/ (accessed 7-7-2015)

Knight, J., Spencer, J., Brooks, A., & Phinn, S. (2007, May). Large-area, high-resolution remote sensing based mapping of alluvial gully erosion in Australia’s tropical rivers. In Proceedings of the 5th Australian Stream Management Conference (pp. 199-204).

Leier, A.L., DeCelles, P.G., Pelletier, J.D., 2005. Mountains, monsoons and megafans. 930 Geology 33, 289–292.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion 50

Olley, J. Brooks, A. Spencer, J. Pietsch, T. Borombovits, D. (2013) Subsoil erosion dominates the supply of fine sediment to rivers draining into Princess Charlotte Bay, Australia. Journal of Environmental Radioactivity v 124 pp 121-129.

Petheram C, Watson I and Stone P (eds) (2013) Agricultural resource assessment for the Gilbert catchment. A report to the Australian Government from the CSIRO Flinders and Gilbert Agricultural Resource Assessment, part of the North Queensland Irrigated Agriculture Strategy. CSIRO Water for a Healthy Country and Sustainable Agriculture flagships, Australia.

Shellberg JG. 2011. Alluvial Gully Erosion Rates and Processes across the Mitchell River Fluvial Megafan in Northern Queensland, Australia. PhD Thesis, Griffith University, Brisbane, Qld, AUS.

Shellberg, J. G., Brooks, A. P., Spencer, J., & Ward, D. (2013). The hydrogeomorphic influences on alluvial gully erosion along the Mitchell River fluvial megafan.Hydrological Processes,27(7), 1086-1104.

Shellberg, J., Brooks, A., & Spencer, J. (2010, August). Land-use change from indigenous management to cattle grazing initiates the gullying of alluvial soils in northern Australia. In 19th World Congress of Soil Science, Soil Solutions for a Changing World (pp. 1-6).

Shellberg, J.G and Brooks, A.P. 2012 Alluvial Gully Erosion: A Dominant Erosion Process Across Tropical Northern Australia. An Educational Report for the Tropical Rivers and Coastal Knowledge program. http://www.track.org.au/sites/default/files/managed/file-attach/biblio/Full_Report_Alluvial_Gully_ Erosion_final.pdf

Shellberg, J.G., Brooks, A.P. (2013) Alluvial Gully Prevention and Rehabilitation Options for Reducing Sediment Loads in the Normanby Catchment and Northern Australia. Prepared by Griffith University, Australian Rivers Institute for the Australian Government Caring for Our Country Reef Rescue Program, Cooktown, Qld. https://www.researchgate.net/publication/258333846,

Shellberg, J.G., Brooks, A.P. and Rose, C.W. (2013). Sediment production and (suspended) sediment yield from an alluvial gully: empirical approaches: Earth Surface Processes and Landforms. DOI: 10.1002/esp.3414

Shellberg, J.G., Brooks, A.P., Spencer, J. and Ward, D., (2013). The hydrogeomorphic influences on alluvial gully erosion along the Mitchell River fluvial megafan, northern Australia. Hydrological Processes. DOI: 10.1002/hyp.9240.

Wilkinson SN, Kinsey-Henderson AE, Hawdon AA, Ellis TW, Nicholas DM (2013) Gully erosion and its response to grazing practices in the Upper Burdekin catchment. A report to NQ Dry Tropics for the Paddock to Reef program. CSIRO Water for a Healthy Country Flagship, Australia.

Developing guidelines for agriculture development in northern Australia to avoid and reduce accelerated erosion

NERPNorthern

@NERPNorthern

Northern Biodiversity This research was supported by funding from the Australian Email: [email protected] Government’s National Environmental Northern Biodiversity Research Program Phone: 08 8946 6761

06/16