一种改进的多学习教与学优化算法

针对传统教与学算法在解决复杂多峰函数优化问题时,具有局部最优且搜索开发能力较差的缺点,提出了一种改进的多学习教与学优化算法,新算法为学员的每一维加入不同的教学因子,设计了基于学员均值比较的教师选择策略和向教师及学员学习的多学习策略。基于多个单峰和多峰函数的仿真结果表明,新算法跟传统的、改进的教与学算法相比,在稳定性、寻优精度和收敛速度方面更具优势。     

Adoption Potential of Conservation Agriculture Practices in Sub-Saharan Africa: Results from Five Case Studies

Despite the reported benefits of conservation agriculture (CA), its wider up-scaling in Sub-Saharan Africa (SSA) has remained fairly limited. This paper shows how a newly developed qualitative expert assessment approach for CA adoption (QAToCA) was applied to determine its adoption potential in SSA. CA adoption potential is not a predictor of observed adoption rates. Instead, our aim was to systematically check relevant factors that may be influencing its adoption. QAToCA delivers an assessment of how suitable conditions “and thus the likelihood for CA adoption” are. Results show that the high CA adoption potentials exhibited by the Malawi and Zambia case relate mostly to positive institutional factors. On the other hand, the low adoption potential of the Zimbabwe case, in spite of observed higher estimates, is attributed mainly to unstable and less secured market conditions for CA. In the case of Southern Burkina Faso, the potential for CA adoption is determined to be high, and this assessment deviates from lower observed figures. This is attributed mainly to strong competition of CA and livestock for residues in this region. Lastly, the high adoption potential found in Northern Burkina Faso is explained mainly by the fact that farmers here have no alternative other than to adopt the locally adapted CA system—Zaï farming. Results of this assessment should help promoters of CA in the given regions to reflect on their activities and to eventually adjust or redesign them based on a more explicit understanding of where problems and opportunities are found.     

Structure of pathways in ecological networks: relationships between length and number

In ecosystems network, structure determines adjacent (direct) and non-adjacent (indirect) pathways over which energy, matter, and information can flow. The more pathways, the more possible ways the conservative substance can move in zero-sum transactions between network nodes that the pathways interconnect, and the more possible non-conservative, nonzero-sum relations can be secondarily derived from these. Structural analysis is a tool we employ, from a family of input–output methods for exploring zero- and nonzero-sum attributes of environmental networks, to count pathways of varying length between network nodes. In this paper, we examine the relationship between pathway length (k) and number (P_k) as determined by system size (n, number of nodes) and extent and pattern of connectance (C). We develop a measure (m_a) of pathway growth in numbers with increasing length, and then normalize this to the maximum rate possible (m_a/m_c) for a given system size. These measures apply to two pathway types—paths, m_a(0) and m_a(0)/m_c(0), which forbid adjacent node repetitions, and walks, m_a(1) and m_a(1)/m_c(1), which allow them. We find that network size has a curvilinear effect on the pathway number versus length relationship, and extent and pattern of connectance are convolved. Values computed for the paths and walks of three ecosystem models (oyster reef, freshwater marsh, and reservoir cove) are used to compare their pathway structure.     

Joint application of an empirical and mechanistic model for regional lake acidification

The empirical direct distribution model for lake acidification is calibrated for use in an integrated assessment model which predicts the regional impact of an acid deposition control strategy. The calibration is based on the mechanistic Model of Acidification of Groundwater in Catchments (MAGIC). The models are applied jointly to a set of 33 statistically-selected lakes in the Adirondack region of New York. Calibration of the direct distribution model is based on a step-function application of acid deposition to MAGIC. Comparative evaluations of the resulting model predictions are made using historic deposition estimates and two alternative future deposition scenarios. The predictions of the direct distribution model match well the shapes and patterns of change of the regional distributions of ANC and pH predicted by MAGIC, the short- and medium-term dynamics of these changes, and the effect of including organic acids. However, small, long-term decreases in the fraction of incoming acid deposition neutralized by lakes and their watersheds predicted by MAGIC are not reproduced.