Evaluation of current statistical approaches for predictive geomorphological mapping

Predictive models are increasingly used in geomorphology, but systematic evaluations of novel statistical techniques are still limited. The aim of this study was to compare the accuracy of generalized linear models (GLM), generalized additive models (GAM), classification tree analysis (CTA), neural networks (ANN) and multiple adaptive regression splines (MARS) in predictive geomorphological modelling. Five different distribution models both for non-sorted and sorted patterned ground were constructed on the basis of four terrain parameters and four soil variables. To evaluate the models, the original data set of 9997 squares of 1 ha in size was randomly divided into model training (70%, n=6998) and model evaluation sets (30%, n=2999).In general, active sorted patterned ground is clearly defined in upper fell areas with high slope angle and till soils. Active non-sorted patterned ground is more common in valleys with higher soil moisture and fine-scale concave topography. The predictive performance of each model was evaluated using the area under the receiver operating characteristic curve (AUC) and the Kappa value. The relatively high discrimination capacity of all models, AUC=0.85–0.88 and Kappa=0.49–0.56, implies that the model's predictions provide an acceptable index of sorted and non-sorted patterned ground occurrence. The best performance for model calibration data for both data sets was achieved by the CTA. However, when the predictive mapping ability was explored through the evaluation data set, the model accuracies of CTA decreased clearly compared to the other modelling techniques. For model evaluation data MARS performed marginally best.Our results show that the digital elevation model and soil data can be used to predict relatively robustly the activity of patterned ground in fine scale in a subarctic landscape. This indicates that predictive geomorphological modelling has the advantage of providing relevant and useful information on earth surface processes over extensive areas, such data being unavailable through more conventional survey methods.     

Modelling patterned ground distribution in Finnish Lapland: an integration of topographical, ground and remote sensing information

New data technologies and modelling methods have gained more attention in the field of periglacial geomorphology during the last decade. In this paper we present a new modelling approach that integrates topographical, ground and remote sensing information in predictive geomorphological mapping using generalized additive modelling (GAM) . First, we explored the roles of different environmental variable groups in determining the occurrence of non‐sorted and sorted patterned ground in a fell region of 100 km2 at the resolution of 1 ha in northern Finland. Second, we compared the predictive accuracy of ground‐topography‐ and remote‐sensing‐based models. The results indicate that non‐sorted patterned ground is more common at lower altitudes where the ground moisture and vegetation abundance is relatively high, whereas sorted patterned ground is dominant at higher altitudes with relatively high slope angle and sparse vegetation cover. All modelling results were from good to excellent in model evaluation data using the area under the curve (AUC) values, derived from receiver operating characteristic (ROC) plots. Generally, models built with remotely sensed data were better than ground‐topography‐based models and combination of all environmental variables improved the predictive ability of the models. This paper confirms the potential utility of remote sensing information for modelling patterned ground distribution in subarctic landscapes.     

喀斯特山地土壤侵蚀和产水量的最优解释力

喀斯特生态系统服务主导影响因子识别是科学治理石漠化的前提,然而环境因子对生态系统服务空间分布的解释力受尺度变换的影响,其尺度效应的定量研究仍需进一步加强。为定量厘定环境因子解释力的尺度效应,本文从多尺度视角出发,利用地理探测器定量分析不同空间分辨率下环境因子对土壤侵蚀和产水量的解释力,并探求其在不同地貌形态类型区的差异性规律。结果表明,坡度和植被覆盖度是影响土壤侵蚀空间分布的主导因子,两者的交互作用对土壤侵蚀空间分布的解释力更强。受研究区地形起伏普遍性和景观破碎化的影响,坡度和土地利用类型在低分辨率下解释力最优。降水、海拔和土地利用类型是产水量空间分异的主导因子。降水和土地利用类型的交互作用对产水量的解释力达95%以上,海拔在不同地貌形态类型区的空间变异性影响其最优解释力水平。具体表现为：在台地、丘陵类型区海拔空间变异性较小,在高分辨率下其解释力最优;在山地类型区,海拔空间变异性较大,在低分辨率下其解释力更强。本文通过多尺度分析定量甄别生态系统服务变量的最优解释力,以期为喀斯特山地生态系统服务的主导因子精准辩识和分区优化提供途径和依据。     

Effects of sample size on the accuracy of geomorphological models

Commonly, the most costly part of geomorphological distribution modelling studies is gathering the data. Thus, guidance for researchers concerning the quantity of field data needed would be extremely practical. This paper scrutinises the relationship between the sample size (the number of observations varied from 20 to 600) and the predictive ability of the generalized linear model (GLM), generalized additive model (GAM), generalized boosting method (GBM) and artificial neural network (ANN) in two data settings, i.e., independent and split-sample approaches. The study was performed using empirical data of periglacial processes from an area of 600 km² in northernmost Finland at grid resolutions of 1 ha (100 × 100 m) and 25 ha (500 × 500 m). A rather sharp increase in the predictive ability of the models was observed when the number of observations increased from 20 to 100, and the level of robust predictions was reached with 200 observations. The result indicates that no more than a few hundred observations are needed in geomorphological distribution modelling at a medium scale resolution (ca. 0.01–1 km²).     

GEOMORPHOLOGICAL CHANGES WITHIN A HILLSLOPE CAUSED BY A WINDTHROW EVENT IN THE TATRA MOUNTAINS,SOUTHERN POLAND

Tree uprooting plays an important role in hillslope evolution. The geomorphological impact of tree uprooting after a foehn wind occurrence, in December 2013 in the Tatra Mountains, was investigated. Geomorphological mapping was conducted in three watersheds. Additionally, in one of the watersheds, 459 windthrow pits were measured, in an area of 6.4 ha. The mean volume of a pit was 2.41 m³, and the mean surface area was 5.47 m². 3.9% of the area was affected by windthrow pits, however locally the magnitude of changes was significantly higher, reaching up to 14.5% of the surface area. Slope inclination weakly influenced the effects of uprooting, and a decrease in the average depth of pits on steep slopes was observed. Individual windthrow pits (five cases) initiated the activity of geomorphological processes, and two cases of periodic springs were noted. Changes in the relief of small landforms caused by tree uprooting were documented. Windthrow creation facilitated the delivery of the soil material from the slopes into the channels.     

模型驱动数据空间分辨率对模拟生态水文过程的影响

设定两种独立的调参方案,检验将高/低分辨率下的调参值应用到低/高分辨率中驱动模型模拟的可行性;同时分析驱动数据分辨率的差异对模型模拟生态水文过程的影响;此外,设置9种集水面积阈值,分析驱动数据分辨率与集水面积阈值对模拟结果的影响。结果表明两种调参方案下模拟的水文与生态数据相关性强、差异小,说明调参方案受驱动数据分辨率的影响不明显;分辨率对土地覆被和土壤类型面积影响小,而对高程和坡度影响大;驱动数据分辨率降低,模拟径流的均值减小、峰值增大、生态变量值增大;驱动数据分辨与集水面积阈值之间不存在交互作用;集水面积阈值增大,模拟径流的均值、峰值均减小,生态变量值也减小,但阈值变化对峰现时间无影响。     

Exploring spatially varying and scale-dependent relationships between soil contamination and landscape patterns using geographically weighted regression

Landscape pattern is an important determinant of soil contamination at multiple scales, and a proper understanding of their relationship is essential for alleviating soil contamination and making decisions for land planners. Both soil contamination and landscape patterns are heterogeneous across spaces and scale-dependent, but most studies were carried out on a single scale and used the conventional multivariate analyses (e.g. correlation analysis, ordinary least squared regression-OLS) that ignored the issue of spatial autocorrelation. To move forward, this paper examined spatially varying relationships between agricultural soil trace metal contamination and landscape patterns at three block scales (i.e. 5 km × 5  km, 10 km × 10 km, 15 km × 15 km) in the Pearl River Delta (PRD), south China, using geographically weighted regression (GWR). This paper found that GWR performed better than OLS in terms of increasing R square of the model, lowering Akaike Information Criterion values and reducing spatial autocorrelation. GWR results revealed great spatial variations in the relationships across scales, with an increasing explanatory power of the model from small to large block scales. Despite a few negative correlations, more positive correlations were found between soil contamination and different aspects of landscape patterns of water, urban land and the whole landscape (i.e. the proportion, mean patch area, the degree of landscape fragmentation, landscape-level structural complexity, aggregation/connectivity, road density and river density). Similarly, more negative correlations were found between soil contamination and landscape patterns of forest and the distance to the river and industry land (p < 0.05). Furthermore, most significant correlations between soil contamination and landscape variables occurred in the western PRD across scales, which could be explained by the prevailing wind, the distribution of pollutant sources and the pathway of trace metal inputs.     

地形信息对确定DEM适宜分辨率的影响

分辨率会直接影响基于栅格数字高程模型（DEM）的数字地形分析结果,因此在实际应用中,需要选择适宜的DEM分辨率。目前采取的基本方法,基于某种地形信息定量刻画尺度效应曲线,从而确定DEM适宜分辨率,但对于采用不同地形信息时所产生的影响尚缺乏研究。本文针对该方法中通常采用的坡度、剖面曲率、水平曲率等3 种地形信息,每种地形信息提取时,分别使用两种不同的常用算法,在3 个不同地形特征的研究区中,逐一计算其在不同分辨率下的局部方差均值,以刻画尺度效应曲线,确定相应的DEM适宜分辨率,并进行对比分析。结果表明：① 采用剖面曲率或水平曲率所得适宜分辨率结果基本相同,但采用坡度所得出的适宜分辨率结果则有明显差别,后者所得的适宜分辨率更粗;② 采用不同地形信息时,越是在平缓地形为主的研究区,所得的适宜分辨率结果越相近,在复合地形特征的研究区所得到的适宜分辨率区间均明显较宽;③ 地形属性计算时所用的算法对适宜分辨率结果的影响不明显。     

SPATIAL ASSOCIATIONS BETWEEN LONGEST-LASTING WINTER SNOW COVER AND COLD REGION LANDFORMS IN THE HIGH DRAKENSBERG, SOUTHERN AFRICA

Although snow is known to influence landform genesis and distribution, the spatial associations between snow and landforms within particular cold regions has received limited research attention. We present a case study from the high Drakensberg of southern Africa, comparing the contemporary spatial pattern of longest-lasting cold-season snow patches with the distribution patterns of active and relic cold region landforms. Two 30 m resolution sets of TM images dated 3 and 19 August 1990 and a DEM were used to demonstrate the geographic trends of snow patch depletion during late winter. Geomorphological phenomena with known coordinates were then incorporated into the GIS. The spatial distribution of several periglacial land-forms (earth hummocks, stone-/turf-banked lobes, block deposits, large sorted patterned ground) coincides with topographic positions that limit snow accumulation. However, the strong spatial association between longest-lasting snow patches and palaeo-moraines implies substantial snow accumulation at some high altitude south-facing sites during the last glacial cycle.     

RATES OF HEAVE AND SURFACE ROTATION OF PERIGLACIAL FROST BOILS IN THE WHITE MOUNTAINS,CALIFORNIA

The White Mountains of California have a cold semi-arid climate that has led to the development of large areas of periglacial patterned ground. Rates of activity vary widely, and no present activity has been noted below 3600 m. Three distinct scales of patterned ground—having average diameters of 3–5 m, 0.5–1.0 m, and 0.1–0.5 m—exist together in a nested fashion at a favorable location at 3914 m. Vertical heave, horizontal displacement, and surface rotation were investigated on landforms of each scale, to determine rates of activity and the spatial and temporal variation of heave. No movement was detected in the largest landforms, but frost boils located in the centers of the larger sorted polygons were highly active. Rates of heave increased dramatically at the beginning of the thaw season, approaching 3.6 mm per day. Dowels were heaved an average of 16.8 cm over ten months. Disks used as surface markers demonstrated exceptional rotational and radial displacement over a nine-month period; 29% of the markers were overturned. The processes responsible for the observed movement in the White Mountains are discussed in relation to processes affecting other alpine environments.     

SCHMIDT‐HAMMER EXPOSURE AGES FROM PERIGLACIAL PATTERNED GROUND (SORTED CIRCLES) IN JOTUNHEIMEN,NORWAY, AND THEIR INTERPRETATIVE PROBLEMS

Periglacial patterned ground (sorted circles and polygons) along an altitudinal profile at Juvflya in central Jotunheimen, southern Norway, is investigated using Schmidt‐hammer exposure‐age dating (SHD). The patterned ground surfaces exhibit R‐value distributions with platycurtic modes, broad plateaus, narrow tails, and a negative skew. Sample sites located between 1500 and 1925 m a.s.l. indicate a distinct altitudinal gradient of increasing mean R‐values towards higher altitudes interpreted as a chronological function. An established regional SHD calibration curve for Jotunheimen yielded mean boulder exposure ages in the range 6910 ± 510 to 8240 ± 495 years ago. These SHD ages are indicative of the timing of patterned ground formation, representing minimum ages for active boulder upfreezing and maximum ages for the stabilization of boulders in the encircling gutters. Despite uncertainties associated with the calibration curve and the age distribution of the boulders, the early‐Holocene age of the patterned ground surfaces, the apparent cessation of major activity during the Holocene Thermal Maximum (HTM) and continuing lack of late‐Holocene activity clarify existing understanding of the process dynamics and palaeoclimatic significance of large‐scale sorted patterned ground as an indicator of a permafrost environment. The interpretation of SHD ages from patterned ground surfaces remains challenging, however, owing to their diachronous nature, the potential for a complex history of formation, and the influence of local, non‐climatic factors.     

The scale effects of the spatial autocorrelation measurement: aggregation level and spatial resolution

The scale effects of the spatial autocorrelation (SA) measurement has been explored for decades. However, the effects of the data aggregation levels and spatial resolution on the SA measurement are often confused. Whether the two types of scale effects are the same is still unclear and requires further investigation. We retrieved the land surface temperature (LST) from Landsat 8 images in 30 capital cities of China. By aggregating the LST images, we observed a decrease in the SA of the LST as the data aggregation level increased; this relationship can be fitted well with a negative logarithmic function. We derived an indicator to measure the scale effects intensity of SA, which was negatively correlated with the spatial complexity of LST. Both aggregating images and the increasing spatial resolution induce weaker SA, but the effect of the former was stronger. The aggregating images negatively affected the SA degree regardless of the spatial resolutions of the original images. The SA degrees of the aggregated images were far below those of the real-life images. This study suggests that the scale effects caused by aggregation levels and spatial resolutions are different, and cautions should be taken when applying relevant conclusions derived from aggregating images.     

MUDBOIL AND ICE‐WEDGE DYNAMICS INVESTIGATED BY ELECTRICAL RESISTIVITY TOMOGRAPHY,GROUND TEMPERATURES AND SURFACE MOVEMENTS IN SVALBARD

Arctic tundra surfaces are dominated by a variety of patterned ground forms. Whereas a large number of studies have described morphology, structure and processes of patterned ground, few have monitored detailed patterned ground dynamics and subsurface environments continuously. We applied electrical resistivity tomography (ERT) to understand near‐surface conditions of two types of patterned ground, ice‐wedge polygons and mudboils in Svalbard, where periglacial processes associated with permafrost are intensively monitored. Automated monitoring shows surface movement characterized by annual cycles of frost heave and thaw settlement, the amounts and rates of which are influenced by the intensity of ice segregation. A time series of ERT shows (1) a distinct resistivity boundary delimiting the active‐layer depth, (2) seasonal variation in resistivity controlled by thermo‐hydrological dynamics and (3) spatial variation in resistivity reflecting desiccation in summer and intensive ice segregation in winter. These results demonstrate ERT as a useful complementary technique for monitoring active‐layer depths and near‐surface hydrological conditions at periglacial patterned ground sites, where automated soil thermal and moisture measurements are limited.     

Earth hummocks in north-east Okstindan,northern Norway: Morphology,distribution and environmental constraints

Gurney, S.D. & Hayward, S. 2015. Earth hummocks in north-east Okstindan, northern Norway: Morphology, distribution and environmental constraints. Norsk Geografisk Tidsskrift–Norwegian Journal of Geography. ISSN 0029-1951.

Earth hummocks (also termed pounus or thúfur) are a common form of periglacial non-sorted patterned ground. The study objectives were to determine the morphology, distribution and development on slopes of earth hummocks in north-east Okstindan, Norway, an area with many hummocks but few documented accounts. The methodology involved detailed geomorphological mapping and precise measurement with a profileometer. The internal structure of the hummocks was investigated through excavations and sediment sample analyses. Fourteen sites with well-developed earth hummocks (accounting for over 650 individual hummock forms) were investigated. The sites have an average altitude of 750?m and occur on slopes with an average gradient of 7°. The hummock heights are in the range 0.11–0.52?m and their diameters 0.7–1.5?m, although coalescent forms are up to 5?m in length. The hummock morphology is characterised by a variable plan form, asymmetry with respect to upslope and downslope forms, downslope elongation, coalescence, and superimposed microtopography. The hummocks’ distribution appeared to have been controlled by the existence of a frost-susceptible ‘host’ sediment, but moisture availability and topographic position played a role. The authors conclude that differential frost heave and vegetation cover stability are critical for the hummocks’ longevity in the studied landscape.     

土地利用变化的多尺度空间自相关分析——以内蒙古翁牛特旗为例

用传统统计学方法模拟和解释土地利用变化的前提条件是研究分析的数据在统计上必须独立且均匀分布。但是空间数据相互之间通常具有依赖性 (即空间自相关),某一变量的值随着测定距离的缩小而变得更相似或更为不同。由于经典线性回归方法未能抓住数据的空间自相关特征,而空间自相关包含一些有用的信息,为了克服这一缺点,利用Moran的I系数自相关图来描述研究区土地利用变化的空间自相关,并且建立了不仅考虑回归而且又考虑空间自相关的混合回归-空间自相关回归模型 (即空间滞后模型)。研究得到：① 研究区土地利用变化模型中不但自变量之间而且因变量之间存在空间正自相关,这表明土地利用变化数据的空间自相关很强;② Moran的I系数随着尺度的变粗而减小,这是由于数据平均时的滤波特性和Moran的I系数对距离的非线性特征造成的;③ 经典线性回归模型的残差也表现出正相关,这表明标准的多元线性回归模型未能考虑土地利用数据所存在的空间依赖性;④ 混合回归-空间自相关回归模型 (即空间滞后模型) 的残差未存在空间自相关,并且有更好的拟合度;⑤ 相对于经典线性回归模型,混合回归-空间自相关回归模型 (即空间滞后模型) 对于存在空间自相关性的数据来说有着统计上的合理性,而经典线性回归模型未能考虑这些因素。     

The use of Landsat imagery to assess large-scale forest cover changes in space and time,minimizing false-positive changes

A new approach, based on the application of multi-spectral remote sensing data of Landsat imagery, is introduced to determine large-scale spatiotemporal variations of forest cover changes quantitatively and with a high degree of precision. The test area covers about 837,330.5 ha of a mountainous region in Central Italy. The approach employs several multi-temporal Landsat acquisitions to account for forest cover changes larger than 0.5 ha for the period from March 2002 to July 2011. In contrast to automated approaches that strongly curtail mapping time, the approach introduced here allowed us to map only the real forest cover change, based on a robust validation and rectification of the detected forest change. Derived high spatial resolution data of forest change estimates indicate that about 5.7% (47,670.5 ha) of the observed forest area was subject to human-induced change between 2002 and 2011. Moreover, the detected forest cover changes, most of which are identifiable as timber harvesting, are considerably larger than those reported in the official statistics and often fall within the perimeter of restricted areas (i.e., national parks and natural reserves).     

Estimating the spatial pattern of human-caused forest fires using a generalized linear mixed model with spatial autocorrelation in South Korea

Most forest fires in Korea are spatially concentrated in certain areas and are highly related to human activities. These site-specific characteristics of forest fires are analyzed by spatial regression analysis using the R-module generalized linear mixed model (GLMM), which can consider spatial autocorrelation. We examined the quantitative effect of topology, human accessibility, and forest cover without and with spatial autocorrelation. Under the assumption that slope, elevation, aspect, population density, distance from road, and forest cover are related to forest fire occurrence, the explanatory variables of each of these factors were prepared using a Geographic Information System-based process. First, we tried to test the influence of fixed effects on the occurrence of forest fires using a generalized linear model (GLM) with Poisson distribution. In addition, the overdispersion of the response data was also detected, and variogram analysis was performed using the standardized residuals of GLM. Second, GLMM was applied to consider the obvious residual autocorrelation structure. The fitted models were validated and compared using the multiple correlation and root mean square error (RMSE). Results showed that slope, elevation, aspect index, population density, and distance from road were significant factors capable of explaining the forest fire occurrence. Positive spatial autocorrelation was estimated up to a distance of 32 km. The kriging predictions based on GLMM were smoother than those of the GLM. Finally, a forest fire occurrence map was prepared using the results from both models. The fire risk decreases with increasing distance to areas with high population densities, and increasing elevation showed a suppressing effect on fire occurrence. Both variables are in accordance with the significance tests.     

陕西省人口分布影响因素的空间计量分析

人口分布影响因素研究有利于揭示人口分布规律，预判人口分布趋势。基于陕西省区县级人口、经济社会、自然地理等数据，通过因子分析方法和空间计量建模解析人口分布的影响因素。研究发现，人口地域别比率不仅取决于一个特定区县内可观测的经济社会、历史基础、自然地理等外在特征，还取决于该区县不可观测的、模型遗漏的其他共有特征，其中经济与公共服务因子、人口基底因子对人口分布具有最显著的正向解释力，其他因素影响较弱或统计不显著；城市等级可显著强化产业结构、人均收入和地形因素对人口分布的影响。研究认为，经济与公共服务因素是优化人口分布的关键，同时需考虑自然地理因素的限制作用。研究对人口分布优化政策的制定具有参考价值。     

中国新型城镇化的空间格局演变及影响因素分析——基于285个地级市的面板数据

对中国2003~2012年285个地级市的新型城镇化水平进行测度,对其空间格局的演变进行深刻剖析。基于距离平方倒数权重矩阵,检验空间相关性,建立空间计量模型分析其影响因素,得出以下结论：① 新型城镇化的空间格局总体呈东高西低态势,且不断向内陆地区扩展,大城市水平明显高于中小城市;② 中国城市的新型城镇化水平具有明显的空间相关性,劳动力、投资水平、经济发展、政府能力、基础设施以及产业结构均起到正向的促进作用,且经济发展、劳动力与基础设施具有显著的正向空间溢出效应,其他几个变量均为负溢出。③ 估计效果最好的SDM模型显示不仅是本地区解释变量,还包括周边地区的相关变量都会影响各城市新型城镇化进程。     

Comparison of input data with different spatial resolution in landscape pattern analysis – A case study from northern Latvia

A suitable spatial scale needs to be selected in geographical and landscape ecological research, and this requires great consideration as different scales have profound effect on derived landscape spatial patterns. Numerous studies have investigated the effects of different scales on landscape metrics using simulated patterns, but few have been conducted to compare different data sources with variable scale for regional- and landscape-scale assessments. Possibly this has occurred because researchers have been prone to use the best available source, a well-known standard, and easiest to use. This study was conducted to assess the impact of input data resolution on values of landscape pattern metrics in four landscapes at scales 1:10 000, 1:50 000 and 1:100 000. The aim was to determine the applicability of three data sources for thematic models in landscape pattern analyses in the Eastern Baltic region. We found that the utility of CORINE Land Cover data for comprehensive structural assessment in mosaic-type landscapes was very limited, as the level of cartographic generalization excluded many small and linear landscape structure elements with potentially high importance for landscape functioning, such as habitat continuity. We also found that actual area harvested using clearcuts was considerably higher than shown in CORINE data, due to clearcuts size being much smaller than the minimum mapping unit. In the light of this, we suggest using data with spatial resolution corresponding to a cartographic scale of at least 1:50 000, in cases when spatial patches have size up to 25 ha.