Browsing by Person "Liu, Hongxi"

Now showing 1 - 1 of 1

Modelling weed management effects on soil erosion in rubber plantations in Southwest China
(2018) Liu, Hongxi; Cadisch, Georg
Land use in Xishuangbanna, Southwest China, a typical subtropical rain forest region, has been dramatically changed over the past 30 years. Driven by favorable market opportunities, a rapid expansion of rubber plantations has taken place. This disturbs forests and land occupied by traditional swidden agriculture thus strongly affecting hydrological/erosion processes, and threatening soil fertility and water quality. The presented PhD thesis aimed at assessing farmer acceptable soil conservation strategies in rubber plantations that efficiently control on-site soil loss over an entire rotation time (25 – 40 years) and off-site sediment yield in the watershed. The study started with field investigations on erosion processes and soil conservation management options in rubber plantations (Chapter 2 and 3). Based on the field data, the physically based model “Land Use Change Impact Assessment” (LUCIA) was employed to assess long-term conservation effects in rubber plantations (Chapter 4) and scale effects on sediment yield in the watershed (Chapter 5). Specifically, the first study aimed at assessing soil loss in rubber plantations of different ages (4, 12, 18, 25 and 36 year old) and relating erosion potential to surface cover and fine root density by applying the Universal Soil Loss Equation (USLE) model. This study adopted the space-for-time substitution for field experimental design instead of establishing a long-term observation. Spatial heterogeneity of soil properties (e.g. texture, organic carbon content) and topography (slope steepness and length) interfered erosion at different plantation ages. To meet this challenge, namely account for possible impacts of soil properties and slope on erosion, the empirical USLE model was applied in data analysis to calculate the combined annual cover, management and support practice factor CP, which represents ecosystem erosivity. Calculated CP values varied with the growth phase of rubber in the range of 0.006 - 0.03. Surface cover was recognized as the major driver responsible for the erosive potential changes in rubber plantations. The mid-age rubber plantation exhibited the largest erosion (3 Mg ha-1) due to relatively low surface cover (40%-60%) during the rainy season, which was attributed to low weed cover (below 20%) and the low surface-litter cover favored by a high decomposition rate. Based on the results of the first study, the second study focused on reducing soil loss in rubber plantations by maintaining a high surface cover through improved weed management. Among the different weeding strategies tested, no-weeding most efficiently reduced on-site soil loss to 0.5 Mg ha-1. However, due to the low farmer acceptance of the no-weeding option, we recommend reducing herbicide application to a single dose at the beginning of the rainy season (once-weeding) to better conserve soil as well as inhibiting overgrowth of the understory vegetation. As the second experiment lasted only one-year, while rubber plantation is a perennial crop with a commercial lifespan of 25 – 40 years, the third study applied the LUCIA model to simulate the temporal dynamics of soil erosion in rubber plantations under different weeding strategies. The erosion module in LUCIA was extended to simulate both runoff and rainfall based soil detachment to better reflect the impact of the multi-layer structure of the plantation canopy. The improved LUCIA model successfully represented weed management effects on soil loss and runoff at the test site with a modelling efficiency (EF) of 0.5-0.96 and R2 of 0.64-0.92. Long-term simulation results confirmed that “once-weeding” controlled annual soil loss below 1 Mg ha-1 and kept weed cover below 50%. Therefore, this weeding strategy was suggested as an eco- and farmer friendly management in rubber plantations. Furthermore, LUCIA was applied at watershed level to evaluate plot conservation impact on sediment yield. Two neighboring sub-watersheds with different land cover were chosen: one a forest dominated (S1, control), the other with a mosaic land use (S2), which served to assess mono-conservation (conservation only in rubber plantations) and multi-conservation (conservation in maize, rubber and tea plantations) effects on total sediment yields. The model was well calibrated and validated based on peak flow (EF of 0.70 for calibration and 0.83 for validation) and sediment yield (EF of 0.71 for calibration and 0.95 for validation) measured from the two watersheds outlet points. Model results showed that improved weed management in rubber plantations can efficiently reduce the total sediment yields by 20%; while multi-conservation was largely able to offset increased sediment yields by land use change. In summary, while exploring the dynamics of erosion processes in rubber plantations, a physically based model (LUCIA) was extended and applied to simulate weed management effects over an entire crop cycle (40 years) and implications at higher scale level (watershed sediment yield). Once-weeding per year was identified as an improved management to reduce on-site erosion and off-site sediment yield. But to fully offset increased sediment yield by land use change, a multi-conservation strategy should be employed, which not only focuses on new land uses, like rubber plantations, but also takes care of traditional agricultural types. A conceptual framework is proposed to further assess the specific sub-watershed erosion (e.g. sediment or water yield) effects in large watersheds by spatially combining process-oriented and data-driven (e.g. statistic based, machine learning based) models. This study also serves as a case study to investigate ecological issues (e.g. erosion processes, land use change impact) based on short-term data and modelling in the absence of long-term observations.