Overview Objectives

Sloping lands comprise an estimated 45 percent of the total land area in the Philippines and directly support approximately 30 percent of the population. Agricultural development on sloping lands is intensifying in response to population increase and the concomitant need to diversify food and income streams. However, sloping lands are vulnerable to erosion and degradation of watershed function, which reduces the capacity to provide vital economic benefits and ecological services. Poverty can be addressed by increasing total factor agricultural productivity from sloping lands and protecting the watersheds where these changes are occurring, but success depends on carefully matching land use with land suitability.
ACIAR recently sponsored a national workshop at which participants concluded that the nature of upland Philippines watersheds is not well enough known to allocate land use across watersheds or to guide the development of improved land management systems. This new ACIAR project will lead a focused process of watershed characterisation - analysing the suitability of land for agricultural development and planning land use to meet multiple social and economic objectives. These will be the basis for a new phase of improved management for both productivity and sustainability.
This project has the following major objectives: 1) develop and apply efficient methods to characterise the Cabulig (Misamis Oriental), Inabanga (Bohol) and Billabong watersheds (NSW) to include biophysical and socio-economic information, with particular emphasis on mapping land and soil attributes using digital technology; 2) develop improved approaches to analysing the suitability of sloping land for agricultural intensification within a watershed context; 3) inform and enhance local land-use planning processes at both the watershed and community scales in Cabulig and Inabanga watersheds; 4) design and implement ongoing monitoring programs in the Philippine watersheds that allow critical assessment of the adoption and impacts of land-use planning.

Progress Reports (Year 1, 2, 3 etc)

During the first year of the project there has been considerable progress on developing and applying efficient methods to characterise the Cabulig watershed (Misamis Oriental) to include biophysical and socio-economic information, with particular emphasis on mapping land and soil attributes using digital technology (Objective 1). The project plan involves a staggered approach to meeting the project objectives in the two southern Philippine watersheds, starting with field survey in the Cabulig watershed. Survey work will take place in the Inabanga watershed during the second year, while work on objectives 2 and 3 begins in Cabulig.
The project inception meeting was held in Cagayan de Oro and Tagbilaran in May 2011 at which the all participants were exposed to different aspects of the Cabulig and Inabanga watersheds. A reconnaissance survey of Cabulig in November-December 2011 provided an overview of the soils and landscapes of the watershed. A sampling design was developed that selected seed sites to cover a wide range of environments within the watershed. Transect sites were also selected along transects away from each seed site at 3, 30, 150 and 500 m. A sampling protocol was developed that emphasized the collection of soil samples for rapid soil analysis, rather than the conventional collection of soil descriptions. The sampling protocols were tested during the reconnaissance survey as was the feasibility of using various geophysical methods (gamma radiometrics, electro-magnetic induction and magnetic susceptibility) to add extra data. However, only magnetic susceptibility proved useful in this watershed. Protocols were also developed for the socio-economic survey, which aimed to interview farmers at two of the soil survey sites on each transect.
After the reconnaissance survey the protocols were finalised for use in the main field survey in February-March 2012. The main survey was preceded by a one week training course “Introduction to field survey” for staff from seven organisations. Many of the trainees went on to participate in the main survey, which helped to reinforce the training and improve ‘soil literacy’ in the various organisations.
During the main field survey, sites were sampled in most parts of the Cabulig watershed, except those that were inaccessible due to very steep or heavily vegetated terrain or poor security. A total of 221 sites were sampled along 44 transects during the reconnaissance and main surveys,. Soil samples from the seed sites were sent for routine soil analysis at Bureau of Soil and Water Management laboratories in Quezon City. To date analysis is complete for samples from the reconnaissance survey.
A field laboratory is being designed for rapid soil analysis during a soil survey. The laboratory will be portable so it can easily be set up in a survey base. The laboratory will analyse samples from all soil layers at all sites using rapid soil analysis. Mid-infrared (MIR) spectrometry will be used to rapidly estimate many soil parameters using locally derived calibrations. The calibrations will be developed using analytical results from the seed sites. Other methods are required to estimate the availability of plant nutrients since they are not estimated by MIR. Work is currently underway to choose and test suitable methods, concentrating on N, P and K.

The second year of the project has seen the on-going work for a number of tasks concerned with data collection for Objective 1, including survey, rapid soil analysis, laboratory analysis, digital soil mapping and training. A summary of achievements is listed here, and expanded on in the following section.
Development of data atlas for Cabulig and Inabanga-Wahig catchments continued.
- New digital elevation models (DEM) at 20 m resolution were developed for both catchments to provide much greater detail than previously available.
Statistical sampling design for Inabanga-Wahig catchment in Bohol was developed.
- Digital land resource assessment sampling methods were refined.
- A statistically based sampling design was used with seed site selected using a stratified random approach incorporating physiographic units and landform classes at two scales that were tailored to the relief of each physiographic unit.
- Sampling sites were located at 0 (seed site), and transect sites at 3, 30, 150, 500m from the seed site.
- The sampling design used for the main Inabanga-Wahig catchment was the most satisfactory procedure to date because it is not only statistically valid but also mimics the way a conventional soil surveyor samples along toposequences.
The land resource survey of Inabanga-Wahig catchment in Bohol was conducted.
- The reconnaissance survey was conducted in November 2012.
- The main survey was conducted in February - March 2013.
- 373 sites along 85 transects were investigated.
- Information was recorded and soil samples collected according to project established protocols.
- Site protocols emphasised the collection of soil specimens for rapid soil analysis, with less emphasis on descriptions of soil morphology.
- Soil samples from the 85 seed sites were sent to BSWM for conventional laboratory analysis.
Data collection continued in the Cabulig catchment.
- Data collection is largely finished.
- Laboratory analysis on soil samples from 44 seed sites was completed by BSWM.
- Interviews for collection of socio-economic data were completed.
Development and testing of rapid soil analysis methods commenced.
- MIR spectroscopy started using a small portable MIR spectrometer.
- Problems were encountered with operating MIR at as a field laboratory at BISU to support the field team. The equipment did not work well without developed laboratory infrastructure, so it was transferred to BSWM in Manila.
- MIR samples from both Cabulig and Inabanga are currently being processed.
- Preliminary MIR calibrations for Cabulig were developed using conventional laboratory analyses of samples from the seed sites.
- Calibrations for clay content, pH, organic carbon, cation exchange capacity, and exchangeable Ca, Mg and acid showed that these properties can be reliably estimated by MIR.
- MIR estimates of iron and exchangeable K are of indicator quality.
Three training courses were run:
- “Trees for Multi-Use Landscapes in the Philippines (TULPhil): A negotiation support toolbox for Integrated Watershed Resources Management”.
- Soil and Land Resources Survey Training Course on “Rapid Soil Analysis”. The course concentrated on the use of MIR spectroscopy for the rapid estimation of soil properties.
- Soil and Land Resources Survey Training Course on “Introduction to Field Survey”. This provided a team to assist with conducting the Inabanga survey.
Six proposals for implementation of TULPhil rapid appraisal tools in the Inabanga catchment were developed led by BISU and another six for Cabulig watershed led by MOSCAT. Implementation will start in 2012-13.

Data gathering in both the Cabulig and Inabanga watersheds has largely been completed. This includes digital land resource survey, socio-economic survey and rapid appraisal using the TULPhil (Trees for multi-use landscapes in the Philippines) tools.
Follow-up soil sampling took place in Inabanga in October 2013 and in Cabulig in March 2014 to increase the range of landscapes sampled.
In Cabulig, socio-economic data was collected by ICRAF at a sub-set of the soil observation sites. In Inabanga, ICRAF collected data from a wider range of households within the catchment in conjunction with another project. The surveyed households have been geo-referenced so that they can be compared to biophysical factors from the soil survey.
Six TULPhil rapid appraisal tools were used to gather data in both watersheds including Participatory Landscape Appraisal, Rapid Carbon Stock Assessment, Rapid Hydrological Appraisal, Rapid Marketing Appraisal, Rapid Agrobiodiversity Appraisal and Rapid Land Tenure Appraisal. The tools were implemented in the Cabulig and Inabanga watersheds by teams led by Misamis Oriental State College for Agriculture and Technology (MOSCAT) and Bohol Island State University (BISU), respectively, and which included other local partners from each region. Apart from gathering data, these projects also served as a capacity building exercise.
The Mid-Project Review meeting was successfully held in Bohol with all counterpart agencies attending and presenting. Information from the meeting is documented in a separate project report ‘Mid Project Review Report: Review Meeting 9 to 11 October 2013’.
Conventional analysis of 792 soil samples from the 137 seed sites of the soil surveys (approximately 21% of all sites) has been completed by Bureau of Soils and Water Management (BSWM). Rapid analysis by mid-infrared spectroscopy (MIR) of 3338 soil samples from all 636 sites has also been completed by BSWM. Partial least squares calibration of the MIR spectra was successfully carried out for the Cabulig watershed to allow estimation of several soil attributes. The statistics of the MIR calibration showed that estimates of organic carbon are equivalent to analytical quality; those of pH, exchangeable Ca, exchangeable Mg, sum of exchangeable cations, exchangeable acid, effective cation exchange capacity, cation exchange capacity and clay content were good quality; and those of exchangeable K, extractable Al, and sand and silt contents were of indicator quality. The use of MIR to estimate this range of parameters means that qualitative analyses are available for all sample sites in the survey and greatly increases the ‘data density’ compared to conventional survey using conventional laboratory methods. MIR calibrations are currently being developed for the Inabanga watershed.
MIR spectroscopy cannot be used to estimate N, P and K, so equipment for rapid soil fertility analysis using ion-selective electrodes (ISE) was selected, purchased and commissioned at BSWM, BISU and MOSCAT. The equipment allows rapid estimation of pH, ammonium, nitrate and potassium. At this stage an ISE method for measuring phosphorus is not available. A one week training course on rapid soil fertility analysis was held at BSWM in Quezon City in February 2014 and was followed by a week of on-the-job training at MOSCAT. The equipment is being used to measure potassium and pH buffering capacity on surface soil samples collected during the surveys.
The final stage of digital land resource mapping (DLRM) involves spatial prediction of individual soil attributes over the survey area. This is currently being undertaken using software developed for DLRM by CSIRO which uses the Cubist program for rule-based regression modelling. The technique develops models between soil attributes at the sample sites and ‘spatial covariates’ available for the whole watershed. In the case of Cabulig these covariates are mainly terrain attributes derived from the digital elevation model. Once a successful model has been developed it is applied over the whole survey area to predict the soil attribute in question. DLRM produces individual maps of each soil attribute at each depth - rather than the single soil type map produced by conventional land resource survey.
The land resource mapping will be used to produce land suitability maps for a variety of crops and cropping systems, based on rules about the requirements for each. Information to define both the range of crops and cropping systems as well as the suitability rules was collated from the literature as well as BSWM. Further information to refine these rules for the individual watersheds is being gathered by visiting a selection of the sample sites and interviewing farmers, extension officers and other local agronomists about crop production problems for existing crops and possible alternative crops. These interviews took place in Cabulig in February/March 2014, but interviewing in Inabanga was interrupted by the October 2013 earthquake in Bohol and is being continued in July-August 2014. The earthquake also prevented the final field soil sampling mission and review of landscapes in the Inabanga Watershed, reducing the number of sample sites. While not ideal, this reduction in field verification is not likely to be a significant limitation for the DLRM.

Digital land resource mapping
Previously, sampling strategies and rapid soil analysis protocols for digital land resource mapping (DLRM) were developed for upland catchments in the Philippines. The final component of DLRM has now been completed with the generation of maps of individual soil properties for the Cabulig watershed. That for the Inabanga catchment is on-going. Generation of these maps requires spatial prediction using:
a) Sample site data - collected according the sampling strategy designed for the project and estimated using rapid soil analysis protocols;
b) Terrain attributes for the whole catchment - derived from a digital elevation model (DEM) created by the project with a resolution of 20 m.
Rule-based regression modelling with ‘Cubist’ was used to find relationships between the site data and the terrain attributes at those sites. Essentially Cubist breaks the data into groups using rules, then applies multiple linear regression to each group in order to predict the soil property. The resulting models are then applied over the whole catchment using the terrain attributes to map the desired soil properties. Cubist was applied iteratively using TERN Soils (developed by CSIRO for the Terrestrial Ecosystem Research Network) with additional modifications to allow for the transect sampling scheme. TERN Soils first removed a random 20% of the transects as an external validation set. Cubist was run 50 times for each soil property. During each iteration a further random 20% of transects was removed for internal validation. A model was then generated and assessed using the internal validation set. Before starting the next iteration the internal validation set was returned to the dataset. The final spatial prediction was the mean of the 50 models. The mean prediction was assessed using the external validation set. In addition, the variability between the 50 predictions at each pixel was used as a measure of uncertainty.
Instead of the single, polygonal, soil type maps produced by conventional soil survey, DLRM maps are generated for each soil property at each depth. The maps show how the property varies continuously over the landscape and is accompanied by an uncertainty map. The table below shows how well the predictions of each soil parameter match the validation set for DLRM in the Cabulig catchment.

Parameter

External concordance*
Depth of layer, cm:

0-10
10-20
20-30
30-50
Soil depth
0.59

Rock content

0.56
0.67
0.43
0.49
Clay

0.57
0.34
0.29
0.44
Silt

0.35
0.63
0.30
0.22
Sand

0.65
0.57
0.61
0.51
Organic C

0.46
0.17
0.46
0.75
pH(CaCl2)

0.86
0.57
0.73
0.64
ECEC

0.74
0.63
0.51
0.71
Exchangeable acid

0.80
0.77
0.65
0.72
Exchangeable K

0.60
0.61
0.62
0.41
Exchangeable acid percent

0.70
0.68
0.63
0.50
ECEC/Clay

0.30
0.66
0.51
0.47
*Concordance is a statistical measure of the agreement between the predicted values and those in the external validation set - i.e. how closely the data lie to the 1:1 line in a plot of observed versus predicted.
The soil maps produced for Cabulig were ‘sensibility tested’ by pedologists to ensure the basic patterns of variation of each soil property were in accordance with soil-forming processes operating in the Cabulig landscape. The concordance values were within the range expected for DLRM.
Land suitability analysis
Semi-structured interviews of farmers at a variety of the seed sites were conducted in the Inabanga catchment. This activity was originally scheduled for October 2013 but was interrupted by the Bohol earthquake.
The rules for land suitability analysis are currently being constructed for both catchments using generic rules supplied by BSWM modified with the results of the interviews. Programming to implement the rules in a GIS framework is on-going.
TULPhil (Trees in multi-use Landscapes in the Philippines) activities
Previously, six TULPhil rapid appraisal activities were undertaken in each catchment by teams from Bohol Island State University, Misamis Oriental State College of Agriculture and Technology and the Local Government Units of Claveria and Pilar:
Participatory Landscape Appraisal Rapid Hydrological Appraisal
Rapid Biodiversity Appraisal Rapid Market Appraisal
Rapid Carbon Stocks Appraisal Rapid Land Tenure Appraisal
The TULPhil teams took varied approaches, with some gathering new data for each catchment and others collating existing data. The teams benefitted from conducting their own research and from building partnerships between different agencies within each catchment.
Following the TULPhil review in May 2014, when each team presented their completed project and received feedback from the review panel, the different teams proceeded with writing final reports. These are now being reviewed by ICRAF for potential working papers and journal publications.

Spatial prediction of soil properties for the Inabanga catchment has been completed using similar methods to that for the Cabulig catchment reported previously. The main inputs were:
a) Sample site data - collected according to the sampling strategy designed for the project and either measured in the laboratory or estimated using mid-infrared spectroscopy;
b) Terrain attributes for the whole catchment - derived from a digital elevation model (DEM) with a resolution of 30 m.
As for Cabulig, the spatial prediction used rule-based regression modelling with Cubist with boot-strapping and with 20% of sampling transects removed for external validation.
However, the start of digital soil mapping was considerably delayed by the discovery of serious errors in the DEM for Inabanga (20 m resolution) which had been generated by digitising topographic maps. Either the original maps or the scans provided to the project were incorrectly registered and had been warped. This meant that landscape features were incorrectly located in the DEM. The effect was that the apparent terrain attributes at the sample sites were incorrect making it impossible to derive predictive relationships between those attributes and the soil attributes.
Considerable time was spent generating ground control points over the area in order to calculate the spatial errors and warp the original contours into the correct locations, before regenerating the DEM. Unfortunately this method was unsuccessful as it appeared to introduce as many errors as it removed.
By this time the 1 arc-second (~30 m) SRTM DEM of the globe had been made available on-line. This DEM was derived from synthetic aperture radar by NASA’s Shuttle Radar Topography Mission in 2000. (Previously this data was only available at a 3 arc-second, about 90 m, resolution, which was found to be too coarse for digital soil mapping.) Although this DEM is noisier that that generated from the topographic maps and has a coarser resolution (30 m versus 20 m), it is correctly located and sufficiently detailed for digital soil mapping after some processing. It also has the advantage that the flatter areas in the centre of the catchment are better represented than in the topographic maps.
The results of using this DEM for digital soil mapping of the Inabanga catchment are shown in the following table:

Parameter

External concordance*
Depth of layer, cm:

0-10
10-20
20-30
30-50
50-70
70-100
Soil depth
0.38

Rock content

0.22
0.31
0.27
0.40
0.50
0.55
Clay

0.49
0.35
0.33
0.35
0.32
0.44
Silt

0.27
0.30
0.20
0.22
0.26
0.25
Sand

0.48
0.48
0.36
0.48
0.61
0.64
Organic C

0.58
0.67
0.50
0.35
0.42
0.31
pH(CaCl2)

0.45
0.56
0.51
0.73
0.57
0.54
ECEC

0.48
0.47
0.47
0.52
0.36
0.36
Exchangeable acid

0.40
0.54
0.47
0.51
0.51
0.40
Exchangeable K

0.47
0.21

0.22

Exchangeable acid percent

0.47
0.38
0.41
0.37
0.40
0.23
ECEC/Clay

0.52
0.49
0.46

*Concordance is a statistical measure of the agreement between the predicted values and those in the external validation set - i.e. how closely the data lie to the 1:1 line in a plot of observed versus predicted.
In general the concordances are lower than those for the Cabulig catchment. The reasons include the coarser scale of the DEM and the greater variety of landscapes in the Inabanga catchment.
Land suitability analysis
Land suitability rules have been compiled for a variety of crops and land uses. The starting point was the land evaluation system currently used in the Philippines by Sys et al. (1993). The rules are being modified for use in sloping areas of the southern Philippines based on findings from the semi-structured interviews conducted in 2014-15. A system to apply these rules to digital land resource information is being developed. The system will allow easy modification of the rules in a spreadsheet format as well quick generation of land suitability maps so that the effects of rule changes can be seen within the context of a group dialogue.
Land use planning
Preparations for catchment and village scale land use planning were made during a trip by Anthony Ringrose-Voase in August 2015. The following process was decided:
Half day workshop with all stakeholders in a catchment, including choosing a smaller group of stakeholder representatives.
Planning workshop for remainder of week with stakeholder representatives
Preparation of planning document for review by all stakeholders, followed modifications
Final presentation to all stakeholders.
The delay in completing the digital soil mapping has meant some of the momentum for these activities may have been lost.

Project ID
SMCN/2009/031
Project Country
Inactive project countries
Commissioned Organisation
CSIRO Land and Water, Australia
Project Leader
Dr Anthony Ringrose-Voase
Email
Anthony.Ringrose-Voase@csiro.au
Phone
+61 2 6246 5956
Fax
+61 2 6246 5965
Collaborating Institutions
World Agroforestry Centre, Philippines
Bureau of Soil and Water Management, Philippines
Department of Environment and Natural Resources, Philippines
Landcare Foundation of the Philippines Inc, Philippines
Misamis Oriental State College of Agriculture and Technology, Philippines
Project Budget
$2,180,574.00
Start Date
01/05/2011
Finish Date
31/10/2015
Extension Finish Date
30/06/2017
ACIAR Research Program Manager
Dr Robert Edis