Hierarchical Prediction Machines and Big Data Analytics

An emerging consensus in cognitive science views the biological brain as a hierarchically-organized predictive processing system. This is a system in which higher-order regions are continuously attempting to predict the activity of lower-order regions at a variety of (increasingly abstract) spatial and temporal scales. The brain is thus revealed as a hierarchical prediction machine that is constantly engaged in the effort to predict the flow of information originating from the sensory surfaces. Such a view seems to afford a great deal of explanatory leverage when it comes to a broad swathe of seemingly disparate psychological phenomena (e.g., learning, memory, perception, action, emotion, planning, reason, imagination, and conscious experience). In the most positive case, the predictive processing story seems to provide our first glimpse at what a unified (computationally-tractable and neurobiological plausible) account of human psychology might look like. This obviously marks out one reason why such models should be the focus of current empirical and theoretical attention. Another reason, however, is rooted in the potential of such models to advance the current state-of-the-art in machine intelligence and machine learning. Interestingly, the vision of the brain as a hierarchical prediction machine is one that establishes contact with work that goes under the heading of 'deep learning'. Deep learning systems thus often attempt to make use of predictive processing schemes and (increasingly abstract) generative models as a means of supporting the analysis of large data sets. But are such computational systems sufficient (by themselves) to provide a route to general human-level analytic capabilities? I will argue that they are not and that closer attention to a broader range of forces and factors (many of which are not confined to the neural realm) may be required to understand what it is that gives human cognition its distinctive (and largely unique) flavour. The vision that emerges is one of 'homomimetic deep learning systems', systems that situate a hierarchically-organized predictive processing core within a larger nexus of developmental, behavioural, symbolic, technological and social influences. Relative to that vision, I suggest that we should see the Web as a form of 'cognitive ecology', one that is as much involved with the transformation of machine intelligence as it is with the progressive reshaping of our own cognitive capabilities.

Racial and spatial interaction for neighborhood dynamics in Chicago

We look at at the empirical validity of Schelling’s models for racial residential segregation applied to the case of Chicago. Most of the empirical literature has focused exclusively the single neighborhood model, also known as the tipping point model and neglected a multineighborhood approach or a unified approach. The multi-neighborhood approach introduced spatial interaction across the neighborhoods, in particular we look at spatial interaction across neighborhoods sharing a border. An initial exploration of the data indicates that spatial contiguity might be relevant to properly analyse the so call tipping phenomena of predominately non-Hispanic white neighborhoods to predominantly minority neighborhoods within a decade. We introduce an econometric model that combines an approach to estimate tipping point using threshold effects and a spatial autoregressive model. The estimation results from the model disputes the existence of a tipping point, that is a discontinuous change in the rate of growth of the non-Hispanic white population due to a small increase in the minority share of the neighborhood. In addition we find that racial distance between the neighborhood of interest and it surrounding neighborhoods has an important effect on the dynamics of racial segregation in Chicago.

How well do regional or national Breeding Bird Survey data predict songbird population trends at an intact boreal site?

A study to monitor boreal songbird trends was initiated in 1998 in a relatively undisturbed and remote part of the boreal forest in the Northwest Territories, Canada. Eight years of point count data were collected over the 14 years of the study, 1998-2011. Trends were estimated for 50 bird species using generalized linear mixed-effects models, with random effects to account for temporal (repeat sampling within years) and spatial (stations within stands) autocorrelation and variability associated with multiple observers. We tested whether regional and national Breeding Bird Survey (BBS) trends could, on average, predict trends in our study area. Significant increases in our study area outnumbered decreases by 12 species to 6, an opposite pattern compared to Alberta (6 versus 15, respectively) and Canada (9 versus 20). Twenty-two species with relatively precise trend estimates (precision to detect > 30% decline in 10 years; observed SE ≤ 3.7%/year) showed nonsignificant trends, similar to Alberta (24) and Canada (20). Precision-weighted trends for a sample of 19 species with both reliable trends at our site and small portions of their range covered by BBS in Canada were, on average, more negative for Alberta (1.34% per year lower) and for Canada (1.15% per year lower) relative to Fort Liard, though 95% credible intervals still contained zero. We suggest that part of the differences could be attributable to local resource pulses (insect outbreak). However, we also suggest that the tendency for BBS route coverage to disproportionately sample more southerly, developed areas in the boreal forest could result in BBS trends that are not representative of range-wide trends for species whose range is centred farther north.

Use of fused airborne scanning laser altimetry and digital map data for urban flood modelling

Flood modelling of urban areas is still at an early stage, partly because until recently topographic data of sufficiently high resolution and accuracy have been lacking in urban areas. However, Digital Surface Models (DSMs) generated from airborne scanning laser altimetry (LiDAR) having sub-metre spatial resolution have now become available, and these are able to represent the complexities of urban topography. The paper describes the development of a LiDAR post-processor for urban flood modelling based on the fusion of LiDAR and digital map data. The map data are used in conjunction with LiDAR data to identify different object types in urban areas, though pattern recognition techniques are also employed. Post-processing produces a Digital Terrain Model (DTM) for use as model bathymetry, and also a friction parameter map for use in estimating spatially-distributed friction coefficients. In vegetated areas, friction is estimated from LiDAR-derived vegetation height, and (unlike most vegetation removal software) the method copes with short vegetation less than ~1m high, which may occupy a substantial fraction of even an urban floodplain. The DTM and friction parameter map may also be used to help to generate an unstructured mesh of a vegetated urban floodplain for use by a 2D finite element model. The mesh is decomposed to reflect floodplain features having different frictional properties to their surroundings, including urban features such as buildings and roads as well as taller vegetation features such as trees and hedges. This allows a more accurate estimation of local friction. The method produces a substantial node density due to the small dimensions of many urban features.

Stochastic modelling of rainfall from satellite data

Satellite-based rainfall monitoring is widely used for climatological studies because of its full global coverage but it is also of great importance for operational purposes especially in areas such as Africa where there is a lack of ground-based rainfall data. Satellite rainfall estimates have enormous potential benefits as input to hydrological and agricultural models because of their real time availability, low cost and full spatial coverage. One issue that needs to be addressed is the uncertainty on these estimates. This is particularly important in assessing the likely errors on the output from non-linear models (rainfall-runoff or crop yield) which make use of the rainfall estimates, aggregated over an area, as input. Correct assessment of the uncertainty on the rainfall is non-trivial as it must take account of • the difference in spatial support of the satellite information and independent data used for calibration • uncertainties on the independent calibration data • the non-Gaussian distribution of rainfall amount • the spatial intermittency of rainfall • the spatial correlation of the rainfall field This paper describes a method for estimating the uncertainty on satellite-based rainfall values taking account of these factors. The method involves firstly a stochastic calibration which completely describes the probability of rainfall occurrence and the pdf of rainfall amount for a given satellite value, and secondly the generation of ensemble of rainfall fields based on the stochastic calibration but with the correct spatial correlation structure within each ensemble member. This is achieved by the use of geostatistical sequential simulation. The ensemble generated in this way may be used to estimate uncertainty at larger spatial scales. A case study of daily rainfall monitoring in the Gambia, west Africa for the purpose of crop yield forecasting is presented to illustrate the method.

An observing system simulation experiment for climate monitoring with GNSS radio occulation data: setup and testbed study

The long-term stability, high accuracy, all-weather capability, high vertical resolution, and global coverage of Global Navigation Satellite System (GNSS) radio occultation (RO) suggests it as a promising tool for global monitoring of atmospheric temperature change. With the aim to investigate and quantify how well a GNSS RO observing system is able to detect climate trends, we are currently performing an (climate) observing system simulation experiment over the 25-year period 2001 to 2025, which involves quasi-realistic modeling of the neutral atmosphere and the ionosphere. We carried out two climate simulations with the general circulation model MAECHAM5 (Middle Atmosphere European Centre/Hamburg Model Version 5) of the MPI-M Hamburg, covering the period 2001–2025: One control run with natural variability only and one run also including anthropogenic forcings due to greenhouse gases, sulfate aerosols, and tropospheric ozone. On the basis of this, we perform quasi-realistic simulations of RO observables for a small GNSS receiver constellation (six satellites), state-of-the-art data processing for atmospheric profiles retrieval, and a statistical analysis of temperature trends in both the “observed” climatology and the “true” climatology. Here we describe the setup of the experiment and results from a test bed study conducted to obtain a basic set of realistic estimates of observational errors (instrument- and retrieval processing-related errors) and sampling errors (due to spatial-temporal undersampling). The test bed results, obtained for a typical summer season and compared to the climatic 2001–2025 trends from the MAECHAM5 simulation including anthropogenic forcing, were found encouraging for performing the full 25-year experiment. They indicated that observational and sampling errors (both contributing about 0.2 K) are consistent with recent estimates of these errors from real RO data and that they should be sufficiently small for monitoring expected temperature trends in the global atmosphere over the next 10 to 20 years in most regions of the upper troposphere and lower stratosphere (UTLS). Inspection of the MAECHAM5 trends in different RO-accessible atmospheric parameters (microwave refractivity and pressure/geopotential height in addition to temperature) indicates complementary climate change sensitivity in different regions of the UTLS so that optimized climate monitoring shall combine information from all climatic key variables retrievable from GNSS RO data.

Estimating the spatial scales of regionalized variables by nested sampling, hierarchical analysis of variance and residual maximum likelihood

The variogram is essential for local estimation and mapping of any variable by kriging. The variogram itself must usually be estimated from sample data. The sampling density is a compromise between precision and cost, but it must be sufficiently dense to encompass the principal spatial sources of variance. A nested, multi-stage, sampling with separating distances increasing in geometric progression from stage to stage will do that. The data may then be analyzed by a hierarchical analysis of variance to estimate the components of variance for every stage, and hence lag. By accumulating the components starting from the shortest lag one obtains a rough variogram for modest effort. For balanced designs the analysis of variance is optimal; for unbalanced ones, however, these estimators are not necessarily the best, and the analysis by residual maximum likelihood (REML) will usually be preferable. The paper summarizes the underlying theory and illustrates its application with data from three surveys, one in which the design had four stages and was balanced and two implemented with unbalanced designs to economize when there were more stages. A Fortran program is available for the analysis of variance, and code for the REML analysis is listed in the paper. (c) 2005 Elsevier Ltd. All rights reserved.

A geomatics-based approach for the derivation of the spatial distribution of sediment transport processes in periglacial mountain environments

Within this paper modern techniques such as satellite image analysis and tools provided by geographic information systems (GIS.) are exploited in order to extend and improve existing techniques for mapping the spatial distribution of sediment transport processes. The processes of interest comprise mass movements such as solifluction, slope wash, dirty avalanches and rock- and boulder falls. They differ considerably in nature and therefore different approaches for the derivation of their spatial extent are required. A major challenge is addressing the differences between the comparably coarse resolution of the available satellite data (Landsat TM/ETM+, 30 in x 30 m) and the actual scale of sediment transport in this environment. A three-stepped approach has been developed which is based on the concept of Geomorphic Process Units (GPUs): parameterization, process area delineation and combination. Parameters include land cover from satellite data and digital elevation model derivatives. Process areas are identified using a hierarchical classification scheme utilizing thresholds and definition of topology. The approach has been developed for the Karkevagge in Sweden and could be successfully transferred to the Rabotsbekken catchment at Okstindan, Norway using similar input data. Copyright (C) 2008 John Wiley & Sons, Ltd.

The spatial prediction of soil mineral N and potentially available N using elevation

The precision farmer wants to manage the variation in soil nutrient status continuously, which requires reliable predictions at places between sampling sites. Ordinary kriging can be used for prediction if the data are spatially dependent and there is a suitable variogram model. However, even if data are spatially correlated, there are often few soil sampling sites in relation to the area to be managed. If intensive ancillary data are available and these are coregionalized with the sparse soil data, they could be used to increase the accuracy of predictions of the soil properties by methods such as cokriging, kriging with external drift and regression kriging. This paper compares the accuracy of predictions of the plant available N properties (mineral N and potentially available N) for two arable fields in Bedfordshire, United Kingdom, from ordinary kriging, cokriging, kriging with external drift and regression kriging. For the last three, intensive elevation data were used with the soil data. The mean squared errors of prediction from these methods of kriging were determined at validation sites where the values were known. Kriging with external drift resulted in the smallest mean squared error for two of the three properties examined, and cokriging for the other. The results suggest that the use of intensive ancillary data can increase the accuracy of predictions of soil properties in arable fields provided that the variables are related spatially. (c) 2005 Elsevier B.V. All rights reserved.

The Representative Soil Sampling Scheme of England and Wales: the spatial variation of topsoil nutrient status and pH between 1971 and 2001

The Representative Soil Sampling Scheme (RSSS) has monitored the soil of agricultural land in England and Wales since 1969. Here we describe the first spatial analysis of the data from these surveys using geostatistics. Four years of data (1971, 1981, 1991 and 2001) were chosen to examine the nutrient (available K, Mg and P) and pH status of the soil. At each farm, four fields were sampled; however, for the earlier years, coordinates were available for the farm only and not for each field. The averaged data for each farm were used for spatial analysis and the variograms showed spatial structure even with the smaller sample size. These variograms provide a reasonable summary of the larger scale of variation identified from the data of the more intensively sampled National Soil Inventory. Maps of kriged predictions of K generally show larger values in the central and southeastern areas (above 200 mg L-1) and an increase in values in the west over time, whereas Mg is fairly stable over time. The kriged predictions of P show a decline over time, particularly in the east, and those of pH show an increase in the east over time. Disjunctive kriging was used to examine temporal changes in available P using probabilities less than given thresholds of this element. The RSSS was not designed for spatial analysis, but the results show that the data from these surveys are suitable for this purpose. The results of the spatial analysis, together with those of the statistical analyses, provide a comprehensive view of the RSSS database as a basis for monitoring the soil. These data should be taken into account when future national soil monitoring schemes are designed.

Spatial and temporal scale issues in determining biomass burning regimes in Bolivia and Peru

ATSR-2 active fire data from 1996 to 2000, TRMM VIRS fire counts from 1998 to 2000 and burn scars derived from SPOT VEGETATION ( the Global Burnt Area 2000 product) were mapped for Peru and Bolivia to analyse the spatial distribution of burning and its intra- and inter-annual variability. The fire season in the region mainly occurs between May and October; though some variation was found between the six broad habitat types analysed: desert, grassland, savanna, dry forest, moist forest and yungas (the forested valleys on the eastern slope of the Andes). Increased levels of burning were generally recorded in ATSR-2 and TRMM VIRS fire data in response to the 1997/1998 El Nino, but in some areas the El Nino effect was masked by the more marked influences of socio-economic change on land use and land cover. There were differences between the three global datasets: ATSR-2 under-recorded fires in ecosystems with low net primary productivities. This was because fires are set during the day in this region and, when fuel loads are low, burn out before the ATSR-2 overpass in the region which is between 02.45 h and 03.30 h. TRMM VIRS was able to detect these fires because its overpasses cover the entire diurnal range on a monthly basis. The GBA2000 product has significant errors of commission (particularly areas of shadow in the well-dissected eastern Andes) and omission (in the agricultural zone around Santa Cruz, Bolivia and in north-west Peru). Particular attention was paid to biomass burning in high-altitude grasslands, where fire is an important pastoral management technique. Fires and burn scars from Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper (ETM) data for a range of years between 1987 and 2000 were mapped for areas around Parque Nacional Rio Abiseo (Peru) and Parque Nacional Carrasco (Bolivia). Burn scars mapped in the grasslands of these two areas indicate far more burning had taken place than either the fires or the burn scars derived from global datasets. Mean scar sizes are smaller and have a smaller range in size between years the in the study area in Peru (6.6-7.1 ha) than Bolivia (16.9-162.5 ha). Trends in biomass burning in the two highland areas can be explained in terms of the changing socio-economic environments and impacts of conservation. The mismatch between the spatial scale of biomass burning in the high-altitude grasslands and the sensors used to derive global fire products means that an entire component of the fire regime in the region studied is omitted, despite its importance in the farming systems on the Andes.

Identifying management zones in agricultural fields using spatially constrained classification of soil and ancillary data

Site-specific management requires accurate knowledge of the spatial variation in a range of soil properties within fields. This involves considerable sampling effort, which is costly. Ancillary data, such as crop yield, elevation and apparent electrical conductivity (ECa) of the soil, can provide insight into the spatial variation of some soil properties. A multivariate classification with spatial constraint imposed by the variogram was used to classify data from two arable crop fields. The yield data comprised 5 years of crop yield, and the ancillary data 3 years of yield data, elevation and ECa. Information on soil chemical and physical properties was provided by intensive surveys of the soil. Multivariate variograms computed from these data were used to constrain sites spatially within classes to increase their contiguity. The constrained classifications resulted in coherent classes, and those based on the ancillary data were similar to those from the soil properties. The ancillary data seemed to identify areas in the field where the soil is reasonably homogeneous. The results of targeted sampling showed that these classes could be used as a basis for management and to guide future sampling of the soil.

Variograms of ancillary data to aid sampling for soil surveys

To provide reliable estimates for mapping soil properties for precision agriculture requires intensive sampling and costly laboratory analyses. If the spatial structure of ancillary data, such as yield, digital information from aerial photographs, and soil electrical conductivity (EC) measurements, relates to that of soil properties they could be used to guide the sampling intensity for soil surveys. Variograins of permanent soil properties at two study sites on different parent materials were compared with each other and with those for ancillary data. The ranges of spatial dependence identified by the variograms of both sets of properties are of similar orders of magnitude for each study site, Maps of the ancillary data appear to show similar patterns of variation and these seem to relate to those of the permanent properties of the soil. Correlation analysis has confirmed these relations. Maps of kriged estimates from sub-sampled data and the original variograrns showed that the main patterns of variation were preserved when a sampling interval of less than half the average variogram range of ancillary data was used. Digital data from aerial photographs for different years and EC appear to show a more consistent relation with the soil properties than does yield. Aerial photographs, in particular those of bare soil, seem to be the most useful ancillary data and they are often cheaper to obtain than yield and EC data.

Sampling the soil in long-term forest plots: The implications of spatial variation

Long-term monitoring of forest soils as part of a pan-European network to detect environmental change depends on an accurate determination of the mean of the soil properties at each monitoring event. Forest soil is known to be very variable spatially, however. A study was undertaken to explore and quantify this variability at three forest monitoring plots in Britain. Detailed soil sampling was carried out, and the data from the chemical analyses were analysed by classical statistics and geostatistics. An analysis of variance showed that there were no consistent effects from the sample sites in relation to the position of the trees. The variogram analysis showed that there was spatial dependence at each site for several variables and some varied in an apparently periodic way. An optimal sampling analysis based on the multivariate variogram for each site suggested that a bulked sample from 36 cores would reduce error to an acceptable level. Future sampling should be designed so that it neither targets nor avoids trees and disturbed ground. This can be achieved best by using a stratified random sampling design.

Using the variogram to explore imagery of two different spatial resolutions

The resolution of remotely sensed data is becoming increasingly fine, and there are now many sources of data with a pixel size of 1 m x 1 m. This produces huge amounts of data that have to be stored, processed and transmitted. For environmental applications this resolution possibly provides far more data than are needed: data overload. This poses the question: how much is too much? We have explored two resolutions of data-20 in pixel SPOT data and I in pixel Computerized Airborne Multispectral Imaging System (CAMIS) data from Fort A. P. Hill (Virginia, USA), using the variogram of geostatistics. For both we used the normalized difference vegetation index (NDVI). Three scales of spatial variation were identified in both the SPOT and 1 in data: there was some overlap at the intermediate spatial scales of about 150 in and of 500 m-600 in. We subsampled the I in data and scales of variation of about 30 in and of 300 in were identified consistently until the separation between pixel centroids was 15 in (or 1 in 225pixels). At this stage, spatial scales of about 100m and 600m were described, which suggested that only now was there a real difference in the amount of spatial information available from an environmental perspective. These latter were similar spatial scales to those identified from the SPOT image. We have also analysed I in CAMIS data from Fort Story (Virginia, USA) for comparison and the outcome is similar.:From these analyses it seems that a pixel size of 20m is adequate for many environmental applications, and that if more detail is required the higher resolution data could be sub-sampled to a 10m separation between pixel centroids without any serious loss of information. This reduces significantly the amount of data that needs to be stored, transmitted and analysed and has important implications for data compression.