In the first year the project team made substantial progress towards each of the three project goals for the major research sites in Sumberjaya (Lampung, Indonesia) and Mae Chaem (northern Thailand).
Analysis of historical records of rainfall and river flow revealed a) the differences and similarities between the two study catchments, and b) the relative importance of changes in tree cover for total water yield (especially at 1.5 m of rain per annum in northern Thailand) and of changes in soil properties for the quick flow component (especially at 2.5 m of rain per annum in Lampung). To bridge between plot-level data and landscape level effects, a spatially explicit rainfall simulator proved to be an essential tool. Analysis of available data showed the low spatial correlation in rainfall events in the Sumberjaya watershed if distances between recording stations are larger than 2-3 km.
Uncertainty on total input of rainfall is a major constraint to the validation of models with the scarce data sets available. Project members have to rely on a comparison of statistical distributions, rather than event-level correspondence between observed and predicted flows. Quantitative tests of the models for both catchments show satisfactory correspondence on indicators of total water yield, buffering of flow and dry-season flows. The land use change in the period 1975 (still 60% forest) to 2000 (15% forest) in Sumberjaya has increased the total water yield per unit rainfall (from 45 to 70%) as well as the number of days that the target water flow for the hydroelectric facility can be met. A new 'buffer indicator' that relates peak flows to peak rain events showed a parallel shift, but there was no consistent change in either the river volume in the driest month or the maximum daily flow.
Equipment was obtained and installed to monitor sediment loads of the river in two subcatchments of Sumberjaya. Initial results of monitoring water quality identified limited problems with sediment loads but valid concerns about pesticide levels linked to pockets of intensive horticulture, in both Mae Chaem and Sumberjaya. A lot of the riparian vegetation just next to the river-albeit often a very narrow strip-is still intact, stimulating biological life. Model analysis suggested that the riparian zone forest has a more-than-proportionately positive impact on reducing sediment loads of the river, provided that there are no 'break-through' points where footpaths, motorbike trails or roads intercept streams and rivers.
The first agreements were signed between local government and farmer groups in Sumberjaya under the 'community forest management' umbrella to allow multistrata coffee gardens to be maintained within the 'protection forest' zone, provided watershed functions are maintained. The project research contributes to the development of criteria and indicators for such agreements and the way they can be monitored.
With the low current prices of coffee the profitability of coffee-based systems has dropped to a level that the opportunity cost of labour is too high for intensive management. Mixed tree-based systems are superior to intensive coffee monocultures when long-term price estimates are used. With improved road access to the area the option of intensive horticulture is gaining popularity, although it only covers a small fraction of the land in Sumberjaya so far. By contrast intensive horticulture is a major source of income in the Mae Chaem watershed, especially by the Hmong farmers using the ridge tops. Its impact on the soil and on infiltration, as well as on river flow due to dry-season irrigation, is a major reason for concern.

This project aims to:
Characterise current land use and ongoing trends in land use change in selected watersheds, and their effects on water quantity, seasonality and sediment load, through participatory resource mapping and dynamic modelling at plot scale of water and sediment flows

Test the effectiveness of existing filter elements and incorporation of these effects into spatially explicit models of soil and water movement in landscape mosaics at catchment scale, and

Quantify the trade-off between watershed functions and profitability of land use for current and possible future land use mosaics, leading to participatory exploration of alternative practices.

In the second year of the project we made further progress towards each of these three goals for the major research sites in Sumberjaya (Lampung, Indonesia) and Mae Chaem (northern Thailand):

Analysis of the river flow records of the past 30 years for Mae Chaem and Way Besai showed some evidence for impacts of the land use change that took place. The overall 'transmission' fraction of rainfall increased for the Way Besai due to the substantial reduction (from about 60 to about 15%) in tree cover, while the newly developed 'buffering indicator' and evidence of river flow in dry periods showed little change. The Mae Chaem with its much more seasonal climate shows a large difference in transmission fraction and buffering between the first and second halves of the rainy season, linked to the degree of saturation of the catchment as a whole; the second half of the season approaches the Way Besai relationship between rainfall and discharge. The overall water balance and dynamics of river flow can be broadly understood and reconstructed by applications of the GenRiver and SpatRain models. There is, however, considerable scope for refinement and improvement. Uncertainty on the actual mean catchment rainfall remains, with indications of the topographic effect shifting with dominant rainfall direction between first and second half of the rainy season in Thailand and short-range rainfall variability in the Way Besai. For the Mae Chaem situation details in phenology of the vegetation influence model results, while the variation in soil properties and their sensitivity to lack of soil cover are important in the Way Besai.

Equipment installed to monitor sediment loads of the Way Ringkih and Way Petai subcatchments of the Way Besai functioned after some start-up problems. It confirmed a substantial difference between the two streams that relates primarily to the difference in soils and in stability of the riparian zone. The primary 'filter' effect in the coffee landscape of Sumberjaya is provided by the surface litter layer. The time pattern of rebuilding a litter layer accounts for much of the variation between land use types in the effective 'entrainment' factor, that relates erosion to overland flow. Second level filter effects are provided by zones of accumulated weed biomass along contour lines in the open phases of coffee gardens and strips of grass and weeds at field boundaries. Most of the stream sediment loads may derive from the motorbike trails and paths that lack filters before transmission to the streams. Analysis of the riparian vegetation also progresses in the Mae Chaem catchment in Thailand and shows that this is generally of higher integrity there than in Indonesia, linked to lower land use intensities. Specific weak spots in the riparian filter relate to access points for domestic livestock to streams.

A ten-step approach was formulated that leads to the compilation of the biophysical and socio-economic information necessary for assessment of the tradeoffs between land use and impacts on quality and quantity of stream flows as part of a negotiation support system. In Sumberjaya the discussion progressed on monitoring and impact assessment of the first agreements between local government and farmer groups under the 'community forest management' (HKM) umbrella. A detailed household characterization was initiated to assess tradeoffs and decision making. With the low coffee prices use of chemical fertilizer on coffee decreased rapidly, but herbicides as labour-saving method of weed control in the more open coffee gardens remained. Some of the farmers are focusing more on the vegetable component of the coffee gardens, intensifying production. Others allow an extensification of the garden and evolution towards multistrata coffee gardens. In the Mae Chaem valley intensive vegetable production during the dry season and its use of any available surface or shallow groundwater source is still on the increase, with conflicts between villages and ethnic groups over access to streams increasing. As a spin-off of the project a simplified 'rapid hydrological appraisal' (RHA) methodology was developed that allows for a comparison of stakeholder perspectives and simple water balance models as a start of a local negotiation process.

This project aims to:
1. Characterise current land use and ongoing trends in land use change in selected watersheds, and their effects on water quantity, seasonality and sediment load, through participatory resource mapping and dynamic modelling at plot scale of water and sediment flows
2. Test the effectiveness of existing filter elements and incorporation of these effects into spatially explicit models of soil and water movement in landscape mosaics at catchment scale, and
3. Quantify the trade-off between watershed functions and profitability of land use for current and possible future land use mosaics, leading to participatory exploration of alternative practices.

In the third year of the project we made further progress towards each of these three goals for the major research sites in Sumberjaya (Lampung, Indonesia) and Mae Chaem (northern Thailand):
1. Analysis of the rainfall-discharge records of the Way Besai showed that the perceived increased frequency in low flows during dry seasons is to a large extent due to the increased frequency of El Nio years in the '90ies compared to the '70ies and '80ies. The 'buffer analysis' that relates river flow to rainfall indicates no substantive change in the water release characteristics of the catchment. The available rainfall-discharge data did not allow a good parametrisation of the IHACRES model, which could separate effects of rainfall vs. land use change on discharge. This was probably due to the high spatial variability of rainfall, not fully represented in the available rainfall station data and perhaps also in poor quality of the available discharge data. Two extensive field campaigns, in collaboration with local farmers and the hydro-power company were organized in February-March in order to assess sediment transport at the sub-catchment level. Variability in sediment yield between sub-catchments was smaller than the variability observed at plot level. Two small rivers were found to already have a significant sediment load during rainfall events before leaving the forest, adding extra evidence that land use might not be the dominant factor to explain erosion and sediment yield. The variation in soils within the catchment remains a key feature, with some soils 'erosive' even with full forest cover, some having a micro-aggregate structure that survives two years without soil cover without generating much erosion and others that do respond in the 'normal' way to lack of cover.
2. Data analysis confirmed that at plot level surface litter indeed functions as the main filter, substantially modifying the sediment concentration in overland flows. In the build up phase of a litter layer (in recently established coffee gardens), movement of litter itself is an important process that slows down the 'filter functions' - any ground vegetation that stops litter movement is important in that respect. At the landscape level footpaths and motorbike trails form channels for sediment transport that bypasses the existing filters and delivers sediment directly to the streams. A sub-project lead by Brawijaya University will collect further data on that process, complementing estimates of bank instability as source of river sediment. Field observations indicated that land slides and bank instability are a likely important sediment contributor in the area. It remains a challenge how these land slide events can be built into the existing models. Even under forest cover, (small) land slides do happen, and affect the sediment load. In the Mae Chaem catchment a typology of riparian vegetation has been developed that will be used to delineate the 'hot spot' areas of sediment delivery.
3. A detailed household survey in Way Besai area provided baseline data that help us understand the plot-level strategies in the context of the overall resources at household level. A new way of defining 'land use intensity' was tested. With (robusta) coffee prices only gradually increasing from their low level, diversification is still an important aspect of strategies. For Mae Chaem a summary was made of the way participatory landscape mapping can be combined with formal GIS techniques. The integrative landscape dynamics model, FALLOW, was further developed to include household-level decision making on the basis of locally acquired knowledge ('adopt and learn').
4. A mid-term review of the project concluded that the project is well placed to meet all the intended deliverables and that cooperation with the national partners is smooth.

This project aims to:
1. Characterise current land use and ongoing trends in land use change in selected watersheds, and their effects on water quantity, seasonality and sediment load, through participatory resource mapping and dynamic modelling at plot scale of water and sediment flows
2. Test the effectiveness of existing filter elements and incorporation of these effects into spatially explicit models of soil and water movement in landscape mosaics at catchment scale, and
3. Quantify the trade-off between watershed functions and profitability of land use for current and possible future land use mosaics, leading to participatory exploration of alternative practices.
In the fourth year of the project we made further progress towards each of these three goals for the major research sites in Sumberjaya (Lampung, Indonesia) and Mae Chaem (northern Thailand):
1. Synthesis of the sediment concentration data for Sumberjaya revealed the interplay of multiple factors, but the complex geology of the area dominates over land cover. Surprisingly, concentrations as high as 1606 mg/l were measured where streams leave the forested top of the Bukit Rigis mountain. Frequent small landslides in unstable tuff (derived from the historic Ranau volcanic eruption) material cannot be prevented by the existing forest cover. The western half of the Way Besai catchment is also characterized by outcrops of Ranau tuff, generally more susceptible to erosion, likely contributing to the large sediment loads in rivers like Way Lirikan, Way Ringki, Way Kabul and Air Napalan. The Tangkit Tebak mountain in the east underlying a number of catchments with generally low sediment loads has a more stable lithology and soils with 23 to 32 % (v/v) drainage pores in the top soil layer, in contrast with sils with only 12 to 13% (v/v) drainage pores and high erodibility elsewhere. The area upstream of one of the cleanest measuring points, covers 36% of the catchment area and would rank low on the list of priority catchments for 'watershed rehabilitation'. Ironically, the past eviction and reforestation efforts were targeted to those headwaters. In hindsight, the evictions and conflict had no rationale in measurable watershed functions. While earlier analyses indicated that available rainfall-discharge data did not allow a good parametrisation of the IHACRES model, significant progress was made in the fourth year. Considerable effort to strengthen the rating curves for three positions in the main river and better understanding of the backflow problems that reduce the validity of the measurement site on which most previous analyses were based, has lead to a tighter fit between data and model. The average water yield per unit area is approximately constant in this new adjusted model, leaving small-range spatial variability of rainfall as key explanation of the lack of fit between model and measurements. Significant progress was made on the participatory assessment of the landscape resources, the sources of sediment loads to the rivers and the opportunities for 'river care'. Schools in the area became involved in the monitoring of water quality though macro-invertebrate indicators. These surveys proved to be important entry points for further community involvement in the watershed.
2. Earlier analyses focussed on the dynamics of the litter layer as the primary filter for overland flow, consistent with the GUEST model. Simulation models thus have to include the dynamics of litterfall, the rates of decomposition and overland transport of litter, in order to adequately describe sediment influx to streams. Coffee gardens generally reach a stable litter layer at 3-5 ages of age and beyond that point most sediment derives from the trails used for motorbike access to the gardens. After initial problems with calibration, a full years dataset of sediment on such trails is currently compiled. Current estimates suggest that about half of the sediment in the streams derives from bank instability, with significant differences between trees depending on their rooting characteristics. As a combination of 'soil anchoring' and 'topsoil binding' is desirable, a mixed vegetation is to be preferred. Detailed analysis of riparian vegetation in Mae Chaem now allows a direct comparison between the two catchments in this regard.
The integrative landscape dynamics model, FALLOW, was adapted to the Mae Chaem landscape and will be used for overall comparisons between the Indonesian and Thai watershed. With co-funding from other sources a more direct approach to the opportunity costs of soil conservation in the context of coffee gardens was made, using an auction approach to establish the level of financial rewards necessary before farmers would be able to make the necessary investments. In parallel, farmer groups undertaking a 'river care' commitment, build on the joint analysis of 'weak links' in the landscape filters for an outcome based reward scheme for reducing sediment loads. After a very gradual 'warm-up', the hydroelectricity company showed active interest in these results and discussions on a continuation of their involvement after the end of the ACIAR-supported project are promising. The number of farmer groups with signed agreements for conservation and use of the forest margin increased from 5 to 23, with significantly reduced transaction costs for the new agreements. A transition from 'proof of principle' to 'proof of application' can now be made. An impact study documents a doubling of farmer assets, a potential income increase of about 30% and enhanced equity for farm households engaged in these contracts, which now cover 70% of the relevant area.