Nonparametric machine learning for mapping forest cover and exploring influential factors

Context

The contribution of forest ecosystem services to human well-being varies over space following the dynamics in forest cover. Use of machine learning models is increasing in projecting forest cover changes and investigating the drivers, yet references are still lacking for selecting machine learning models for spatial projection of forest cover patterns.

Objectives

We assessed the ability of nonparametric machine learning techniques to project the spatial distribution of forest cover and identify its drivers using a case study of Tasmania, Australia.

Methods

We developed, evaluated, and compared the performance of four nonparametric machine learning models: support vector regression (SVR), artificial neural networks (ANN), random forest (RF), and gradient boosted regression trees (GBRT).

Results

The results demonstrated that RF far outperformed the other three models in both fitting and projection accuracy, and required less computional costs. GBRT outperformed SVR and ANN in projection accuracy. However, RF exhibited serious overfitting due to the full growth of its decision trees. The influence rankings of explanatory variables on spatial patterns of forest cover were different under the four models. Land tenure type and rainfall were identified among the top four most influential variables by all four models. The ranking produced by the RF model was significantly different with topographic factors associated with land clearing and production costs (elevation and distance to timber facilities) being the two most influential variables.

Conclusions

We encourage practitioners to consider nonparametric machine learning methods, especially RF, when facing problems of complex environmental data modelling.

     

Using occupancy and species distribution models to assess the conservation status and habitat use of the goldline darter (Percina aurolineata) in Georgia,USA

The goldline darter (Percina aurolineata) is threatened by recent increases in urbanisation in the Coosawattee River watershed, but no studies have addressed their current status. Similarly, limited data on habitat use make it difficult to understand factors that may impact this species. We compared occurrence data before (1996–2000) and after (2009–2011) a period of rapid development and population growth within the watershed. Single‐season occupancy models were built to account for imperfect species detection and to identify habitat covariates. MaxEnt was used to identify important stream reaches for conservation and to understand landscape scale factors correlated with the distribution of goldline darters. Our results indicate a high proportion of sites occupied upstream of Carters Lake during the historic and recent time periods, with no evidence of decline. However, occurrences of goldline darters for sites in Talking Rock Creek and the lower Coosawattee River were sparse in the historic period and absent during the recent period. The probability of detecting goldline darters is low and was positively associated with the occurrence of small substrate. Species distribution models were strongly influenced by watershed area and elevation and indicated a high probability of suitable habitat within the Coosawattee River and large tributaries upstream of Carters Lake. While goldline darter occupancy is currently stable upstream of Carters Lake, continued urbanisation is a threat to long‐term persistence. We recommend additional monitoring and describe a protocol that allows for precise estimates of species occupancy while minimising the risk of sampling‐related mortality.     

The Kuiper belt and the solar system's comet disk

Our planetary system is embedded in a small-body disk of asteroids and comets, vestigial remnants of the original planetesimal population that formed the planets. Once formed, those planets dispersed most of the remaining small bodies. Outside of Neptune, this process has left our Kuiper belt and built the Oort cloud, as well as emplacing comets into several other identifiable structures. The orbits in these structures indicate that our outer solar system's comet disk was shaped by a variety of different physical processes, which teach us about how the giant planets formed. Recent work has shown that the scattered disk is the most likely source of short-period comets. Moreover, a growing body of evidence indicates that the sculpting of the Kuiper belt region may have involved large-scale planetary migration, the presence of other rogue planetary objects in the disk, and/or the close passage of other stars in the Sun's birth cluster.     

Marked decline in atmospheric carbon dioxide concentrations during the Paleogene

The relation between the partial pressure of atmospheric carbon dioxide (pCO2) and Paleogene climate is poorly resolved. We used stable carbon isotopic values of di-unsaturated alkenones extracted from deep sea cores to reconstruct pCO2 from the middle Eocene to the late Oligocene (approximately 45 to 25 million years ago). Our results demonstrate that pCO2 ranged between 1000 to 1500 parts per million by volume in the middle to late Eocene, then decreased in several steps during the Oligocene, and reached modern levels by the latest Oligocene. The fall in pCO2 likely allowed for a critical expansion of ice sheets on Antarctica and promoted conditions that forced the onset of terrestrial C4 photosynthesis.     

Flood volcanism in the northern high latitudes of Mercury revealed by MESSENGER

MESSENGER observations from Mercury orbit reveal that a large contiguous expanse of smooth plains covers much of Mercury's high northern latitudes and occupies more than 6% of the planet's surface area. These plains are smooth, embay other landforms, are distinct in color, show several flow features, and partially or completely bury impact craters, the sizes of which indicate plains thicknesses of more than 1 kilometer and multiple phases of emplacement. These characteristics, as well as associated features, interpreted to have formed by thermal erosion, indicate emplacement in a flood-basalt style, consistent with x-ray spectrometric data indicating surface compositions intermediate between those of basalts and komatiites. The plains formed after the Caloris impact basin, confirming that volcanism was a globally extensive process in Mercury's post-heavy bombardment era.     

Molecular Characterization of Atlas 66-Derived Wheat Near-lsogenic Lines Contrasting in Aluminum （Al） Tolerance

Aluminum （Al） toxicity is the major limiting factor for wheat growth in acidic soils. Genetic improvement of Al tolerance is one of the most cost-effective solutions to improve wheat productivity. The objective of this study was to characterize near isogenic lines （NILs） contrasting in Al tolerance derived from Atlas 66 in the backgrounds of Al-sensitive cultivars Chisholm and Century using amplified fragment length polymorphism （AFLP） and simple sequence repeat （SSR）. A total of 200 AFLP and 88 SSR primer pairs were screened and 12 markers （11 AFLPs and one SSR） were associated with Al-tolerance in NILs of at least one recurrent parental background. Among them, nine were linked to A1 tolerance in the Chisholm-derived NILs, seven were associated with Al-tolerance in the Century-derived NILs, and three AFLPs derived from the primer combinations of pAG/mGCAG, pCAG/mAGC and pGTG/mGCG, and one SSR, Xwmc331 on chromosome 4D, associated with A1 tolerance in NILs of both recurrent parental backgrounds. Those common markers across two backgrounds may be the major marker loci associated with Al-tolerance in Atlas 66 and could be useful for marker-assisted breeding to improve Al tolerance in wheat. In addition, evaluation of Al tolerance among different genotypes using hematoxylin stain and relative root growth revealed that Atlas 66 was more tolerant to Al stress than the NILs, therefore suggested that the Al-tolerant NILs might not carry all Al-tolerance loci from Arias 66 and inheritance of Al tolerance in Arias 66 is more likely multigenic.     

Monovalent salt-induced gelation of enzymatically deesterified pectin

Pectin gels were induced by monovalent salts (0.2 M) concurrently with deesterification of high methoxy pectin using a salt-independent orange pectin methylesterase (PME). Constant pH was maintained during deesterification and gelation. If salt or PME was absent, the pectin did not form a gel. The gel strength was influenced by both pH and species of monovalent cation. At pH 5.0, the pectin gel induced by KCl was significantly stronger than the NaCl-induced gel. In contrast, a much stronger gel was produced in the presence of NaCl as compared to KCl at pH 7.0. LiCl did not induce pectin gelation at either pH. Molecular weights of pectins increased from 1.38 x 10(5) to 2.26 x 10(5) during NaCl-induced gelation at pH 7. One proposal to explain these pectin molecular weight changes is a hypothetical PME transacylation mechanism. However, these pectin molecular weight changes can also be explained by metastable aggregation of the enzymatically deesterified low methoxy pectin. We postulate that gelation was induced by a slow deesterification of pectin under conditions that would normally salt out (precipitate) low methoxy pectin in the absence of PME.     

A QTL for early heading in wheat cultivar Suwon 92

Heading date is an important trait that determines wheat adaptation to environments. A recombinant inbred line (RIL) population derived from CI 13227 × Suwon 92 was employed to tag the quantitative trait locus (QTL) for early heading in Suwon 92. This population was phenotyped for heading date in 1994, 1995, and 1997, and analyzed with AFLP and SSR markers. Two AFLP markers (XGCTG.CGCT118 and XGCTG.CGCT60) closely associated with heading date were identified. Across years, XGCTG.CGCT118 and XGCTG.CGCT60 explained 40.4% and 32.2% of the total phenotypic variances, respectively. Interval analysis revealed a major QTL for heading date, designated QHd.pser-2DS, between AFLP marker XGCTG.CGCT118 and SSR marker Xgwm261. Based on the linkage map, QHd.pser-2DS was about 41.2 cM proximal to the distal end of chromosome 2DS, and explained 40.5% of the phenotypic variance across three years. The identified markers associated with the early heading QTL have the potential to be used in wheat breeding programs.