Exploring impacts of rainforest disturbance history and conversion to oil palm on water quality dynamics in eastern Sabah using a novel concentration-discharge pattern approach

Assessment of water quality impacts of rainforest disturbance and land-use change on water quality in the wet tropics is hampered by the technical difficulty and prohibitive costs of collection of multi-catchment high-frequency streamflow and water quality datasets for a period of record that covers a sufficiently representative range of storm events and preceding weather and baseflow conditions. This problem is magnified in agricultural plantation areas, where water quality responses in storm events vary also with the nature of management practices, including time since fertilizer application.  There has also been little focus on impacts of land-use history such as multiple phases of logging and mature agricultural plantation.  This paper addresses these issues using a novel classification of  daily C-Q (solute concentration – specific discharge) patterns in a multi-catchment study in eastern Sabah, Malaysian Borneo. The five small (1.7 – 4.6 km2) study catchments  lie in the upper reaches of the Brantian, Kalabakan and Segama river systems within the 10,000 km2 Yayasan Sabah Forest Concession area, where rainforest since the 1970s has been either (1) selectively logged (and left to recover) up to three times, (2) subsequently converted to oil plantations, or (3) protected as primary forest in three large Conservation Areas (Danum Valley, Maliau Basin and Imbak Canyon) or as near-primary forest in  smaller Virgin Jungle Reserves.  Two of the study catchments are under primary  and near-primary rainforest; two are under forest recovering from two and three episodes of selective logging respectively; and the final catchment is covered by mature (>20 year-old) oil palm.  Annual rainfalls for the catchments are 2500-2880 mm.  Water depth, conductivity and turbidity sensors linked to Campbell data loggers have recorded readings at 5-minute intervals in each catchment from 2011. Catchment-specific solute concentration/specific conductance and stage-discharge relationships were used to derive the 5-minute solute concentration (C, mg L-1) and specific discharge (Q, m3 km-2 s-1) data series.  To compare their water quality dynamics, C-Q  relationships for each day over the 22-months period November 2011 to August 2013 were analysed for each catchment.  For each day, the correlation coefficient (r) and slope (b) of the best-fit logC-LogQ regression were calculated and graphs of Log C/Log Q scatter and  C and Q against time were produced.  Days were divided into Storm Days and Recession/Baseflow Days. A typology of C-Q patterns (eight Storm Day and four Recession/Baseflow Day types) was devised using: the r and b values; (for Storm Days) the order and relative dominance of any dilution and flushing response features; and (for Recession/Baseflow Days) the ranges in Q and C values.  Each day of each data series was classified and percentage frequency distributions of C-Q types for each catchment were derived and compared.  The frequency distribution of the oil palm catchment is markedly different (fewer “dilution” and more “flushing” storm days) than for the forested catchments  - which  can be linked to fertilizer sources and lower baseflows of the oil palm catchment. Lesser, but important differences between the forested catchments are also identified and discussed.