COUNCIL MEETING,CHICAGO, ILLINOIS November 3-4, 1972

This paper questions the validity and relevance of the application of procedures of classical statistical inference to population data in geography. Arguments for such procedures include the notions of temporal and spatial samples, measurement error, modifiable areal units, data vetting and stochastic processes. It is concluded that such arguments cannot be justified in terms of statistical theory.     

THE SENSITIVITY OF TWO DISTRIBUTED NON-POINT SOURCE POLLUTION MODELS TO THE SPATIAL ARRANGEMENT OF THE LANDSCAPE

Distributed hydrological models are becoming increasingly complex with respect to spatial phenomena, and with the widespread availability of spatial data from GIS, this trend is likely to increase. In all such models the spatial arrangement of phenomena, such as soil properties and land-use categories is fundamental, and so the arrangement should have an influence on the model output. Testing for this influence we term spatial sensitivity analysis. Here, we report on the spatial sensitivity of two widely used models, AgNPS (agricultural non-point source pollution model) and ANSWERS (areal nonpoint source watershed environment response simulation). The input spatial data were subjected to spatially random mixing to varying degrees, such that the organized landscape became disorganized. The chemical discharge from AgNPS, and the sediment and water discharge from ANSWERS, are examined. In both cases most outputs exhibited little or no sensitivity to the spatial distribution of most input data. Only infiltration-related inputs produced large variations, but these changes were not in the sense that might have been predicted. Although the analytical methods used require further refinement, there must now be some doubt as to the validity of the models, and whether they repay their computational complexity. Furthermore, it is felt that spatial sensitivity analysis should become a fundamental part of the verification of all such models. © 1997 by John Wiley & Sons, Ltd.     

POPULATIONS,SAMPLES AND STATISTICAL INFERENCE IN GEOGRAPHY

This paper questions the validity and relevance of the application of procedures of classical statistical inference to population data in geography. Arguments for such procedures include the notions of temporal and spatial samples, measurement error, modifiable areal units, data vetting and stochastic processes. It is concluded that such arguments cannot be justified in terms of statistical theory.     

测量的广义可靠性研究

首次给出了广义可靠性的定义,对于测量误差理论进行了进一步论述,分析了偶然误差的系统性影响和系统误差的偶然化两大问题,讨论了目前流行的测量不确定度概念,指出了广义可靠性的研究内容和若干应用,对于测量方案设计、平差数据处理、仪器检定以及测量规范的理解等具有一定参考意义。     

Identifying insect infestation hot spots: an approach using conditional spatial randomization

Epidemic populations of mountain pine beetle highlight the need to understand landscape scale spatial patterns of infestation. The observed infestation patterns were explored using a randomization procedure conditioned on the probability of forest risk to beetle attack. Four randomization algorithms reflecting different representations of the data and beetle processes were investigated. Local test statistics computed from raster representations of surfaces of kernel density estimates of infestation intensity were used to identify locations where infestation values were significantly higher than expected by chance (hot spots). The investigation of landscape characteristics associated with hot spots suggests factors that may contribute to high observed infestations.     

Randomization test and its application

In this paper, one of the distribution-free tests — randomization test, is briefly described. It doesn’t need any distribution assumption and its related parameter estimation and is applicable to random and nonrandom sample. Then it is used to the test of migration of strong earthquakes on the Xianshuihe Fault Belt and “immunity” of large earthquakes in the large northern reigon of China. The test results show that there is 98.7% confidence degree for the migration of strong earthqueks on the Xianshuihe Fault Belt and “immunity” of earthqueks withM _S⩾8 toM _S⩾7 is significant in the large northern region of China. The obtained test results and the test method itself have certain application in the practice. The Chinese version of this paper appeared in the Chinese edition ofActa Seismologica Sinica,15, 484–489, 1993.     

Influence of Velocity and Types of Beam Trawl Towing on Deep-Sea Demersal Fish and Decapod Crustacean Samples

The sampling efficiency of two types of beam trawl towed at different velocities was compared when sampling deep-sea demersal fish and decapod crustaceans. The trawls were a Sigsbee-Agassiz type of 2 m span, and an Oregon type of 3 m span. The different hauls caught by the Oregon type beam trawl of 3 m span towed at 1.5 knot (3m 1.5 kt), the Oregon type towed at 0.8 knot (3m 0.8 kt) and the Sigsbee-Agassiz type towed at 1.5 knot (2m 1.5 kt) were compared by analysis of variance (ANOVA), the Tukey test, cluster analysis based on Morisita-Horn index, and correspondence analysis. The minimum viable sample size was determined by a randomization test. For demersal fish, the type of gear used had more influenced on the samples obtained than the towing velocity. For decapod crustaceans, no significant difference was detected between different trawls towed at different velocities. The common idea that if these types of trawl are towed at 1 knot or more, they cannot efficiently sample epifauna, but demersal fish can be well sampled, should be abandoned. The samples of demersal fish caught by 3m 1.5 kt revealed the greatest consistency and lowest variation than the others. We recommend using an Oregon type beam trawl of 3 m span towed at 1.5 knot as an ordinary deep-sea sampling device. Only a single successful haul or a sample size of 52 individuals was sufficient for estimating demersal fish diversity, but at least two successful hauls or a sample size of 54 individuals were needed for decapod crustaceans. If possible, seven successful hauls or a sample size of 189 individuals of decapod crustaceans are recommended for community study. This revised version was published online in July 2006 with corrections to the Cover Date.