#Maxent

[dropsofsciencenews]

OpenStreetMap to Stop Cheetah Trafficking

ITA version ESP version

Behind the glamorous social media photos of cheetah cubs lies a dark reality. With fewer than 7,000 individuals in the wild, the cheetah (Acinonyx jubatus) is one of the world’s most endangered species, and the illegal trade of live cubs, especially in the Horn of Africa, is devastating its populations. Poachers take cubs from their mothers while they hunt, selling them into the luxury pet market at prices reaching $50,000 per cub. Between 2020 and 2023, this trade increased by 50%, posing a serious threat to the species' survival.

However, a new methodology developed by a research team could finally help monitor and disrupt illegal cheetah trafficking through a three-tiered approach. Scientists are mapping not only the trafficking routes but also previously undocumented cheetah populations, combining prey species distribution models, habitat suitability, and a trafficking network based on OpenStreetMap data.

The first step involves locating cheetah prey, such as gazelles and antelopes, which are essential for their survival. The team used Maxent, a machine learning method, to cross-reference prey presence data with environmental variables like climate and land cover, thus mapping the most suitable areas for cheetah survival. This step identifies key conservation areas and, unfortunately, those most vulnerable to poaching.

Based on these data, researchers developed a Habitat Suitability Index (HSI) specific to cheetahs, highlighting areas where their survival is most probable. This model identifies regions with optimal conditions for stable populations and locates areas at risk of trafficking, enabling targeted interventions.

Finally, a model was created to understand how traffickers move cheetahs from remote origin areas to Arabian Peninsula markets. By integrating OpenStreetMap data on roads, ports, and border crossings, researchers identified the most likely land and sea routes and main transit points. Route optimization tools mapped out trafficking “hotspots,” allowing law enforcement to plan strategic interventions at critical crossing points, such as less monitored ports and borders.

This model is a breakthrough in conservation, as it not only enables mapping and monitoring of trafficking routes but also reveals previously undocumented cheetahs, giving scientists and conservationists a more complete picture of the situation. With this discovery, we finally have a system to protect one of Africa’s most iconic species and provide local communities with a tool to defend their natural resources.

See You Soon and Good Science!

source

[i-am-l-ananas]

born to find joy in catching and looking at bugs; forced to do math about it

[captain-price-unofficially]

Wrecked Light Tank Mk VIC on a road between Huppy and Saint-Maxent, Picardy, France. 29 May 1940

[max1461]

So Yoruba vowel harmony, huh. You can model it with maxent. But should you? We shall see.

[botanyone]

AI Spots Plant Changes from Space Using Citizen Science

AI Spots Plant Changes from Space Using Citizen Science https://ift.tt/ajD9Vvy Gillespie and colleagues have created an AI tool called Deepbiosphere to track plant biodiversity. Using satellite images and data from citizen scientists, this deep learning model has mapped over 2,000 plant species across California. It outperforms traditional methods, spotting both towering redwoods and wildflowers with high accuracy. Deepbiosphere could revolutionise global efforts to monitor plant life, helping us understand how climate change and human activity are reshaping ecosystems. The authors claim that Deepbiosphere outperforms traditional species distribution models, achieving higher accuracy. It can map species at up to a few metres resolution and accurately identify plant communities. The model detected both mature and young regrowth in redwood forests, showing the lasting effects of deforestation. It can also identify rapid changes in plant communities after events like wildfires, demonstrating its potential for monitoring biodiversity changes over time. Gillespie and colleagues developed a deep learning model, Deepbiosphere, using a  modified convolutional neural network architecture to fed with combined aerial imagery from the National Agriculture Imagery Program with over 650,000 plant observations from citizen scientists across California. The model was trained to predict the presence of 2,221 plant species. Its performance was compared to traditional modeling approaches like MaxEnt and Random Forest. Plant biodiversity is changing rapidly due to habitat destruction and climate change. Traditional methods lack the spatial and temporal resolution to detect these rapid changes for individual species. Deepbiosphere’s approach, combining deep learning with remote sensing, offers new possibilities for high-resolution biodiversity monitoring. Ultimately, we envision a paradigm shift toward open-source foundation models that are continuously trained and improved with new remote sensing data, citizen science observations, and data modalities as they become available. Achieving this from public airborne or satellite imagery and growing citizen science observations will make biodiversity monitoring more accessible, thus advancing local and global nature conservation goals. Gillespie, L. E., Ruffley, M., & Exposito-Alonso, M. (2024). Deep learning models map rapid plant species changes from citizen science and remote sensing data. Proceedings of the National Academy of Sciences, 121(37), e2318296121. https://doi.org/10.1073/pnas.2318296121 (OA) Cross-posted to Bluesky, Mastodon & Threads. The post AI Spots Plant Changes from Space Using Citizen Science appeared first on Botany One. via Botany One https://botany.one/ September 17, 2024 at 12:30PM

[codeshive]

LIN570: HW10 – maxent solved

1. Q1 (55 points): Create a MaxEnt POS tagger, maxent tagger.sh. • The command line is: maxent tagger.sh train file test file rare thres feat thres output dir • The train file and test file have the format (e.g., test.word pos): w1/t1 w2/t2 … wn/tn • rare thres is an integer: any words (in the train file and test file) that appear LESS THAN raw thres times in the train file are treated as rare…

[lettucefather]

Pray for me in my MAXENT adventures (it's been reading the environmental file for 6 hours and it's still 0%)

[geckophysics]

Born to MAXENT, Forced to Equal a Priori Probabilities.

[444names]

Names generated from the full list of Moon selenographical features and French names

Aford Afrey Ageasbeas Agnetinz Albeauzeau Aldis Alement Alley Alpoune Alvenois Amauberis Amblamist Amene Amons Aneau Apent Arbin Ardan Aseidger Asheille Assen Astine Auxure Aviconne Avien Ayermonna...

Bacharyn Bacquer Baire Bancitat Bannyson Bardimer Barte Batillesse Bayeaudeck Bazine Beaur Beauval Beaux Belley Belmir Berno Biert Binglacour Bisoner Bizée Bjesse Blanque Blault Boischaël Bolzeau Bommon Bonius Boragore Borber Borene Borélodil Bosselle Bouay Boulta Bound Bounsuels Bouron Boynisell Bozon Brach Bradin Brantion Brodour Brosimirs Broux Brussau Bujobile Bunoister Burche Burcouray Burier Bustelin Cadebon Callard Capances Caphéopey Caplourie Caravalet Carna Casque Casse Chard Chazo Cheect Cheland Cheliandre Chene Choamy Choeurne Clois Clorectoue Cloux Coderthwev Coison Colaford Cotte Coudeau Couis Coupur Coute Couth Crard Crocher Croterre Custivalps Cécierise Célin Dabeau Dabing Daner Daramorva Darmy Darvilemie Davoyeux Deate Deaulis Decon Dejousts Delenter Dellanger Dellegrass Derafors Desmarrise Desquer Dessamas Desse Desset Dettegis Doine Donton Dorcad Dornanne Dorte Doufore Doviergeog Drablante Drailau Drivet Dubertain Dublet Dubrice Dubue Dubuilance Duchefand Ducierthin Ducrèchay Dudecurin Duernien Dullin Dunne Dupina Dupornaud Dupra Durea Délinoël Emessaveux Erdefoux Ertalps Exhinck Fabiernis Fabontorne Fallard Fauchomarl Faville Fayet Febrier Fecoutais Feneve Flecov Flemi Flordin Fordine Forochau Fraist Freau Fretteline Frier Frobin Galen Galvegerd Ganhord Garks Genion Gidosteen Gimon Glard Goigni Gotaudes Gouille Grache Grade Graninne Graste Grefed Grich Grier Groquiront Guilia Gärtionoy Hasset Hathe Heaudi Hiboix Hiche Hotterun Houchard Hugan Humeng Hunet Héniatrun Jacieux Jacom Jacquer Jaloquel Jambon Janarobir Janbron Janneks Jearmán Jeaux Jesjasher Joldukev Josylvale Julamo Krachard Laclandris Lacolzeaux Lafain Lafolze Lagaisegir Lagriault Laharoury Lallooks Lamilet Lanchee Lanissago Lapen Laper Larbeley Lardoise Larolia Lathing Lavergay Lavide Laïsen Leauzier Lecre Lemette Leminette Lerroux Leton Leuregate Lexan Liedolson Lierlev Livigne Lormedis Lorétatte Loutartion Loyothie Luaillan Luclanill Lunetiss Mahin Maile Maimo Malme Mangen Manne Manstrain Mantio Marchalp Mardille Mardinet Marics Marivas Marle Marmon Marnes Martanifte Mashilie Mathélous Matiaquert Maudin Maxelle Maxente Mayohanne Maëll Medefau Medier Mellierond Menacke Menan Mierdonc Misimax Mobin Mocho Monne Monovskill Moroit Motte Moure Mukeve Nange Narder Nelleguis Nelogette Nesarrim Niline Nistin Nolia Nonser Ochamp Océcil Offeaulin Ohneautrev Olfrable Oreyoe Oucyrenne Palen Pandive Paque Pardé Paste Patte Payere Peliphone Pettette Phabon Phienne Phome Phorree Plageollin Plope Poldrau Polet Popper Poqueriver Poutridon Prett Pricouis Proch Rabriquese Rainien Rallet Rance Rante Raument Rence Retion Reurn Ricle Rigok Risson Ristry Robeauc Robeyriff Roier Romonse Rompoisle Romsdebon Rondoux Roodev Rotiedge Rourity Ruglia Runierisem Rélémy Sabarge Sardes Schoux Selegous Seline Sellagard Shaine Sichenato Siolau Slitz Smane Soncy Sophaléo Sopponne Soutre Spard Spatrum Spenesieu Spuyeau Stion Stivilloë Streher Stéon Stéont Suille Suprin Suriste Sylvy Talins Tatier Telaux Tervier Tesmytoux Theeperse Thenarraim Theushard Thiree Tinke Tonneau Treauzie Trulx Trumnerey Tudemined Ungeau Ussette Ustus Vadejeau Venyu Verry Verte Vetterry Villanie Villemes Voites Vopoix Wanusionis Whartouche Whoside Wiche Wikisevre Wikoe Witte Worasser Wriau Yalmodnard Yamescasim Zemadne Éliandry Élinette Élène

[malcolmgg]

GGBet - legalne kasyno i bukmacher z licencją MGA

Czy zastanawiałeś się, czy GGBet to legalne kasyno i bukmacher? Jeśli tak, mamy dla Ciebie dobrą wiadomość - tak jest! Platforma GGBet działa w oparciu o międzynarodową licencję MGA/B2C/210/2011 wydaną przez Malta Gaming Authority. W tym artykule omówimy kluczowe informacje na temat licencji GGBet oraz sposobu działania kasyna i bukmachera.

GGBet jest własnością spółki Maxent Limited, zarejestrowanej zgodnie z prawem Unii Europejskiej i Malty. Numer rejestracyjny spółki to C47261, a adres siedziby to JPR Buildings, Level 2, Triq Taz-Zwejt, San Gwann, SGN3000, MALTA. Wszystkie te informacje świadczą o legalności działalności GGBet.

GGBet działa w oparciu o międzynarodową licencję wydaną przez Malta Gaming Authority. Licencja ta, o numerze MGA/B2C/210/2011, została wydana 1 sierpnia 2018 roku i jest zgodna z Gaming Authorisations Regulations (L.N. 243 of 2018) oraz prawem maltańskim. Dzięki tej licencji GGBet może oferować swoje usługi w wielu krajach europejskich.

Działalność ggbets-pl.com to nie tylko kasyno, ale również bukmacherka. Aby rozpocząć grę na platformie, należy założyć konto przez stronę internetową lub aplikację mobilną. Po założeniu konta i wpłaceniu depozytu, gracz może wybierać spośród wielu zakładów bukmacherskich i gier kasynowych online. Pełna oferta platformy jest dostępna dla pełnoletnich użytkowników z kontem na stronie.

Ważne jest, aby grać odpowiedzialnie i korzystać z najlepszych metod ochrony danych i finansów, co zapewnia GGBet. Platforma oferuje także akcje promocyjne, które umożliwiają graczom zdobycie dodatkowych środków do gry.

GGBet to legalne kasyno i bukmacher, działające na podstawie międzynarodowej licencji MGA/B2C/210/2011 wydanej przez Malta Gaming Authority. Dzięki tej licencji platforma może oferować swoje usługi w wielu krajach europejskich. Gra na GGBet jest bezpieczna, a platforma korzysta z najlepszych metod ochrony danych i finansów. Odpowiedzialna gra i korzystanie z oferowanych akcji promocyjnych mogą przynieść graczom wysokie wygrane.

[dropsofsciencenews]

OpenStreetMap per fermare il traffico di ghepardi

ENG version ESP version

Dietro le foto patinate dei cuccioli di ghepardo sui social si nasconde un mondo oscuro. Con meno di 7.000 esemplari in natura, il ghepardo (Acinonyx jubatus) è una delle specie più minacciate al mondo, e il commercio illegale di cuccioli vivi, specialmente nel Corno d'Africa, ne sta decimando le popolazioni. I bracconieri sottraggono i piccoli alle madri mentre queste sono a caccia e li destinano al mercato del lusso come animali da compagnia esotici, a un costo che può arrivare fino a 50.000 dollari per esemplare. Tra il 2020 e il 2023, questo commercio è aumentato del 50%, rappresentando una minaccia concreta alla sopravvivenza della specie.

Ora, però, una nuova metodologia sviluppata da un team di ricerca potrebbe finalmente aiutare a monitorare e interrompere il traffico illegale di ghepardi, con un approccio innovativo a tre livelli. Gli scienziati stanno mappando non solo le rotte di traffico, ma anche le popolazioni di ghepardi non ancora documentate, combinando i modelli di distribuzione delle specie preda, l’idoneità dell’habitat e una rete di traffico basata sui dati di OpenStreetMap.

Il primo passo consiste nel localizzare le prede dei ghepardi, come gazzelle e antilopi, essenziali per la loro sopravvivenza. Per farlo, il team ha utilizzato Maxent, un metodo di apprendimento automatico, per incrociare i dati di presenza delle specie preda con variabili ambientali come clima e copertura del suolo, mappando così le aree più adatte a sostenere i ghepardi. Questo passaggio consente di identificare le zone chiave per la conservazione e, purtroppo, anche quelle più vulnerabili al bracconaggio.

Sulla base di questi dati, i ricercatori hanno sviluppato un Indice di Idoneità dell'Habitat (HSI) specifico per i ghepardi, che evidenzia le aree dove la loro sopravvivenza è maggiormente probabile. Questo modello permette di individuare le aree con le condizioni migliori per ospitare popolazioni stabili e di identificare i luoghi a rischio di traffico, favorendo interventi mirati.

Infine, è stato costruito un modello per comprendere come i trafficanti trasportano i ghepardi dalle aree di origine, spesso remote, fino ai mercati della Penisola Arabica. Integrando i dati di OpenStreetMap su strade, porti e punti di confine, i ricercatori hanno individuato le rotte terrestri e marittime più probabili e i principali punti di transito. Grazie agli strumenti di ottimizzazione delle rotte, sono stati mappati i percorsi più caldi del traffico, permettendo alle forze dell’ordine di pianificare interventi strategici nei punti di passaggio critici, come porti e confini meno sorvegliati.

Questo modello rappresenta una svolta per la conservazione, non solo perché consente di mappare e monitorare le rotte di traffico, ma anche perché ha permesso di identificare ghepardi che finora sfuggivano alle ricerche, dando a scienziati e conservazionisti un quadro più completo della situazione. Grazie a questa scoperta, abbiamo finalmente in mano un sistema che permetterà di proteggere una delle specie più iconiche dell’Africa e di offrire alle comunità locali uno strumento per difendere le proprie risorse naturali.

A Presto e Buona Scienza! fonte

[i-am-l-ananas]

GIS work is such a slippery slope. You start out going yay :) I get to make a pretty map :)!! to what the fuck do you mean I need to make a space time cube??? the tardis?? the fucking tardis?? you want me to make an insect tardis????

[gaare-geog560]

Annotated Bibliography

Adhikari, Pradeep & Lee, Yong Ho & Adhikari, Prabhat & Hong, Sun & Park, Yong-Soon. (2022). Climate change-induced invasion risk of ecosystem disturbing alien plant species: An evaluation using species distribution modeling. Frontiers in Ecology and Evolution. 10:880987.. 13. 10.3389/fevo.2022.880987. 

Species distribution models have become an effective tool for monitoring invasive species. In South Korea, invasive species cause a loss of 22.6 billion Korean Won per year. Because of the extreme effect that invasives pose on the South Korean economy, scientists and land managers are motivated to monitor and manage this problem. Two statistical regression models were used along with three machine learning models. The two regression models used in this study were generalized linear model (GLM) and multivariate adaptive regression splines (MARS). The three machine learning models used in this study were artificial neural network (ANN), maximum entropy (MaxEnt), and random forest (RF). One of the two main objectives for this study was to determine the best algorithm and estimate the potential distribution area of 12 invasive species. Several variables were considered for this study including land cover change, distance from the road. distance from water and 19 bioclimatic variables. Through analysis of the various algorithms, it was determined that the random forest model had the highest predictive performance. The random forest model was determined to be the best model based on the average area under the curve (AUC), true skill statistics (TSS), and Kappa scores. The random forest model also determined that coastal areas of the northwestern and southern regions are at high risk for invasion of alien species. Some large cities of South Korea including Incheon, Seoul, and Busan are also at high risk for invasion. Climate change still poses a great threat to South Korea in regard to invasive species spread. 

Bonneau, Laurent & Shields, Kathleen & Civco, Daniel. (1999). Using Satellite Images to Classify and Analyze the Health of Hemlock Forests Infested by the Hemlock Woolly Adelgid. Biological Invasions. 1. 10.1023/A:1010021629127. 

The eastern hemlock (Tsuga canadensis) is a coniferous tree that can be found across the eastern United States. In southern New England, it can be found often in rocky ravines. This particular tree retains its lower branches, providing thermal cover for birds and small mammals during cold, winter months. T. canadensis has a unique ecological niche in that it provides structural habitat diversity for populations of small mammals and fish. This species is threatened by the hemlock woolly adelgid, which is an insect that was introduced from Asia. The hemlock woolly adelgid feeds on eastern hemlock by feeding on storage cells and can introduce toxins into the tree. Satellite imaging has made it possible to study insect infestations on forest lands. The main objectives of this study are to 1) evaluate a variety of image enhancement techniques to classify hemlock health and 2) to correlate hemlock stand health with landscape features. In the first phase of this study, the authors created a map to identify the location of hemlock in the study area, evaluate the health of hemlock stands through field surveys and compare image enhancing techniques to classify hemlock pixels based on health. In the second phase of this study, the authors examined the distribution of hemlock health and associated the following attributes: slope, aspect, basal area, canopy height and digital soils data. Through this study, the authors found that the MSAVI2 vegetation index produced the most accurate map regarding hemlock forest health. It was also found that hemlock forest health was positively correlated with deep, excessively drained, medium textured soils. The use of the techniques from this study could help further understand migration patterns of the Hemlock woolly adelgid. 

Boyte, Stephen & Wylie, Bruce. (2016). Near-Real-Time Cheatgrass Percent Cover in the Northern Great Basin, USA, 2015. Rangelands. 38. 278-284. 10.1016/j.rala.2016.08.002. 

Cheatgrass has become a great concern threatening sagebrush steppe ecosystems in the Great Basin. With the invasion of the annual grass, comes an increased fire frequency. With increased fire frequency comes altered ecosystems susceptible to higher rates of spread of invasive species. The invasive annual grass known as cheatgrass contributes to soil erosion on the landscape and destroys wildlife habitat, specifically for the now endangered species, the greater sage grouse. Greater sage grouse populations have declined an average 2% every year from 1965 to 2003. They have now been listed as an endangered species. Understanding the spread of invasive annual grasses, such as cheatgrass, can assist land managers, fire modelers and policy makers with addressing relevant issues such as decreased wildlife habitat and increased wildfire frequency. The authors used a two-step process to produce an estimation of percent cover and location of cheatgrass in the Northern Great Basin. A regression-tree software was used to develop parameters for an ecological model. Cheatgrass data from 2001 to 2006 was used to train the ecological model. Four datasets were added to the model, specifically expedited Moderate Resolution Imaging Spectroradiometer17 (eMODIS) normalized difference vegetation index (NDVI) data. The data collected from eMODIS and NDVI are available as a continuous weekly time series, which added a time-step feature to the cheatgrass model. The model allowed for the production of a cheatgrass cover dataset and maps from 2000 to 2013. The ecological model created was then applied to site specific variables from 2015. This was able to create the 2015 near-real-time cheatgrass cover estimate, which showed a slightly higher cover percentage of cheatgrass than the previous 14 year model. The authors determined that the increase in cover percentage was likely due to higher average annual precipitation. For the development and creation of future models, historical data is important and needed for accuracy. 

Carrión-Klier C, Moity N, Sevilla C, Rueda D, Jäger H. The Importance of Very-High-Resolution Imagery to Map Invasive Plant Species: Evidence from Galapagos. Land. 2022; 11(11):2026. https://doi.org/10.3390/land11112026

Biodiversity on the Galapagos Islands is threatened by invasive plant species. Invasive species alter ecosystem functions and displace native and endemic species. Invasive species also have a huge impact on the economy, contributing to an estimated cost of 120 billion USD in the United States. In this study, the authors evaluated the effects spatial resolution has on accurately mapping invasive species on the Galapagos Islands. The guava tree (Psidium guajava), Cuban cedar (Cedrela odorata) and blackberry bush (Rubus niveus) are the invasive species of interest in the study. The study area is located on Santa Cruz Island, in the highlands of the Galapagos National Park. The authors in the study used very-high-resolution (VHR) WorldView-2 images to identify the invasive species of interest. Images from the hot and wet season in the Galapagos were used with <10% cloud cover. The vegetation is green and lush during the hot and wet seasons, making it easier to distinguish between species in the images. The authors used both VHR imagery and medium-resolution (MR) imagery and found that VHR imagery produced more accurate species distribution models. VHR imagery is a great resource for identifying and mapping invasive species. 

Cho, Ki Hwan & Park, Jeong-Soo & Kim, Ji & Kwon, Yong & Lee, Do-Hun. (2022). Modeling the distribution of invasive species (Ambrosia spp.) using regression kriging and Maxent. Frontiers in Ecology and Evolution. 10. 10.3389/fevo.2022.1036816. 

Biodiversity is threatened by the invasion of non-native species. Human activities can promote the spread of invasive species by unintentional seed dispersal. In this study, the authors use spatial and non-spatial models for distribution of Ambrosia artemisiifolia and Ambrosia trifida. A. artemisiifolia and A. trifida are two invasive plant species that are widely distributed throughout South Korea. These two species have difficulty dispersing long distances by wind due to their morphological characteristics Two modeling techniques used in this study were regression kriging (RK) and Maxent. Regression kriging is an interpolation method that takes into account spatial autocorrelation. The regression model allows for predictions to be made. Maxent is a machine learning-based species distribution model that was used the model the distributions of Ambrosia spp. Maxent uses presence-only data to model species distributions by evaluating habitat suitability. Habitat suitability is evaluated by comparing the probability density of background samples with the probability density of presence data. The habitat suitability index (HSI) estimated by Maxent indicated a high HSI for A. artemisiifolia in the northwestern part of the study area and a low HSI for A. trifida in the northwestern and eastern parts of the study area. The authors found that the RK model provides better predictive performance than Maxent. The RK model predicted lower occurrence probability of A. artemisiifolia because fewer individuals were identified in the study area. Maxent did not take into account the spatial context of the data used in the model and predicted a higher HSI. It was determined that invasive species distribution models are likely to produce more accurate model results with the use of spatial autocorrelation. 

Henry, Annie & González, Eduardo & Robinson, W. & Sher, Anna. (2018). Spatial modeling improves understanding patterns of invasive species defoliation by a biocontrol herbivore. Biological Invasions. 20. 10.1007/s10530-018-1794-0. 

In the American southwest, land managers and ecologists struggle with controlling the invasive tree Tamarix spp. Spatial modeling has become useful for understanding plant populations and their growth habits but has yet to understand the impacts of a biological control agent on invasive species. In this study, a biocontrol herbivore was introduced to aid in the containment of the invasive species Tamarix spp. Tamarix spp. was introduced to the U.S. in the early 19th century. This tree was primarily used as a bank stabilizer, an ornamental and a windbreak. River regulations and the changing of flood regimes assisted in the establishment of Tamarix spp. as a dominant species. This species severely affects ecosystems in that it changes the soil salinity of an ecosystem, increases fire frequency and alters river geomorphology. In this study, the authors quantified various environmental variables against spatial structure in determining defoliation patterns of Tamarix. The biocontrol agent used in this study was Diorhabda spp., which is a low cost and effective method for reducing Tamarix dominance. Diorhabda  carinulata was released at 12 locations between 2004 and 2006. 79 field sites were established throughout Grand County, Utah to monitor defoliation in Tamarix after Northern tamarisk beetles were released. Monitoring occurred once per growing season between 2013 and 2016. Point intercept method was used to measure canopy cover. Permanent transects were established along the edge of Tamarix stands to record whether an individual was alive or dead. Geographic features, stand characteristics and soil characteristics were included in the dataset. Average live canopy of Tamarix varied from 42% to 59%. 2013 showed the highest average live canopy for Tamarix with a slight decrease the following years. It was determined that live canopy cover increases with the number of years since first defoliated and decreases with stand age. It was also found that most of the variability within the canopy was not related to environmental variables. This study will open new pathways for understanding biological control agents effects on invasive plant species. 

Holcombe, Tracy & Stohlgren, Thomas & Jarnevich, Catherine. (2007). INVASIVE SPECIES MANAGEMENT AND RESEARCH USING GIS.

Geographic Information Systems (GIS) and Global Positioning Systems (GPS) are key tools for spatial analysis. Maps can be quickly and easily produced and spatial analysis becomes extremely easier. Many industries have been positively impacted through the development of GIS including health care, agriculture and the environmental industry. A large, ongoing problem within the environmental industry is the development of invasive species. Invasive species typically thrive in all types of environments, with little to no resources (ex. sunlight, water availability, etc). GIS is a very important, applicable tool in the environmental industry because it can create potential distribution maps for early detection and rapid assessment of invasive species. Using GIS applications, the authors analyzed the distribution patterns of the invasive species, Cane Toad, in the United States at the 6-digit Hydrologic Unit Code (HUC) level. The authors used data obtained from USGS Florida Integrated Science Center’s Non-Indigenous Aquatic Species database on the invasive cane toad. Using a logistic regression model, the authors determined potential habitat for the invasive cane toad within the US. Variables used in the logistic regression model included minimum temperature, minimum radiation, mean temperature, maximum temperature, maximum humidity, and maximum growing degree day. The results of this model showed that the cane toad population had already invaded the majority of suitable habitat (mostly within the state of Florida) from the model's prediction. The authors used another GIS application to estimate potential suitable habitat for the cane toad, which resulted in the cane toad not having much more suitable habitat than the areas it was already occupying within the state of Florida. 

Joshi, Chudamani & De Leeuw, Jan & Van Duren, Iris. (2004). Remote sensing and GIS applications for mapping and spatial modeling of invasive species. ISPRS. 35. 

Remote sensing (RS) and Geographic Information Systems (GIS) have become important tools for land management agencies. RS and GIS can produce habitat suitability models for early detection of invasive species. There are various systems used in remote sensing including aerial imagery, multispectral scanners, broad-band scanners and hyperspectral scanners. The authors in this article review remote sensing techniques and GIS used for mapping potential distribution patterns of invasive species. Since remote sensing is based on the use of satellite imagery, remote sensing data is strongly correlated to canopy cover. In this article, invasive species are categorized based on their canopy cover within an ecosystem. Canopy cover is broken down into four classes including class I: Canopy dominating species, class II: Mixed canopy dominant species, class III: Invaders influencing canopy dominant species and class IV: Understory species. Various studies have noted the significance of differences in canopy architecture, leaf pubescence and vegetative density for mapping invasive species using aerial photography. Using reflectance imagery for aquatic invasive species has been described as limiting, because little light is reflected off of submerged vegetation, resulting in inaccurate measurements. Other studies have found remote sensing reflectance imagery to be an effective and accurate tool for mapping invasives. It has also been found extremely challenging to map invasive plants and animals occupying the understory using any of the remote sensing techniques mentioned. There are still improvements to be made for accurately documenting invasive species distributions using remote sensing techniques. 

Lee W-H, Song J-W, Yoon S-H, Jung J-M. Spatial Evaluation of Machine Learning-Based Species Distribution Models for Prediction of Invasive Ant Species Distribution. Applied Sciences. 2022; 12(20):10260. https://doi.org/10.3390/app122010260

Artificial intelligence has recently been used to advance the production of species distribution models (SDMs). Machine learning algorithms are utilized to determine environmental characteristics for potential habitats. MaxEnt is a type of model that utilizes the maximum entropy theory to assess the probability of species occurrence and distribution based on environmental factors. Random Forest (RF) is another popular tool used to identify suitable habitat for target species. Multi-layer perceptron (MLP) is a recent approach that classifies the presence or absence of a species.  In this study, the authors use MaxEnt, RF and MLP to predict potential global species distribution for two species of ants, Solenopsis invicta and Anoplolepis gracilipes. Data points were gathered and used in the models, which included 6163 data points for S. invicta and 1297 data points for A. gracilipes. Climate data and bioclimatic variables were also used for predicting the distribution of the ant species. Model performance metrics were used to compare the 3 models. The authors found the MaxEnt model showed the highest performance while MLP showed low performance. Accuracy of each model for both species exceeded 0.9 except MLP accuracy for A. gracilipes. This could be correlated with less data points available for the models to be trained with in regards to A. gracilipes. The authors found that the amount of data correlates with the performance of the model. Ideally, more data availability will improve the performance of future species distribution models. 

Masocha, Mhosisi & Skidmore, Andrew. (2022). INTEGRATING CONVENTIONAL CLASSIFIERS WITH A GIS EXPERT SYSTEM IMPROVES INVASIVE SPECIES MAPPING. 

Biodiversity is threatened by the invasion of alien plants in an ecosystem. With the loss of biodiversity, comes the loss of ecosystem functions and environmental services. It has become an important task for natural resource managers to accurately map and understand potential areas that invasive species threaten. Remote sensing has become increasingly important for managing invasive species but has been found to be insufficient and inaccurate. Remote sensing is ideal for plants that occur in the canopy, but most smaller invasive plants occur below the canopy and are therefore undetected by remote sensing. Because field / ground-based surveys can be inefficient and time consuming, land managers need a more efficient approach to this large task of managing, treating and reducing populations of invasive species. The authors in this study hypothesized that using expert systems, such as computer programs using symbolic logic, combined with conventional image classification methods can increase the accuracy for mapping invasive species. The authors chose the invasive species L. camara for this study because it has a complex growth habit and is not easily detected from remote sensing. The study takes place in southern Zimbabwe. During December 2006, 187 plots were sampled and the presence / absence of L. camara was recorded as well as the number of stems and cover percentage in four cover classes. The four cover classes included in the study were absent (0% cover), low (1–30% cover), medium (31–50%) and high (>50% cover). Four methods were used to map the cover of L. camara including neural network, support vector machine, hybrid neural network plus expert system and hybrid support vector machine plus expert system. The accuracy of these methods were calculated using a dataset that was produced specifically for testing accuracy. This study found that the hybrid neural network plus expert system method achieved the highest accuracy for mapping invasives with an accuracy of 82.9%. This supports the hypothesis that combining a Geographic Information System (GIS) expert system with a single classifier would produce more accurate results. 

Qiao X, Liu X, Wang F, Sun Z, Yang L, Pu X, Huang Y, Liu S, Qian W. A Method of Invasive Alien Plant Identification Based on Hyperspectral Images. Agronomy. 2022; 12(11):2825. https://doi.org/10.3390/agronomy12112825

Invasive species are a huge threat to ecosystems and biodiversity. Hyperspectral technology has become a resourceful tool for accurately detecting invasive species. An efficient and low-cost approach to managing invasive species is to accurately identify and monitor populations. However, field surveys can become time consuming and tedious. In this study, the authors attempt to identify invasive species based on hyperspectral imagery. In this study, the high-speed imaging spectrograph I185 was used to collect hyperspectral images of several invasive plant species in China. First derivative, Savitzky–Golay Smoothing and Standard Normal Variable Transformation were used as preprocessing methods to reduce noise, astigmatism and baseline drift within the hyperspectral data. Reducing the dimensionality of hyperspectral data also improves the efficiency and performance of the model. Principal Component Analysis and Ant Colony Optimization were utilized to reduce dimensionality. The Random Forest Model (RF) and Support Vector Machine (SVM) were utilized for the identification of seven invasive species. The authors found that M. micrantha had the highest reflectivity in the wavelength range followed by B. pilosa, M. pudica, L. cairica, A. conyzoides, S. calendulacea and L. camara. Overlapping regions between species was relatively large and the background reflectance was relatively scattered. The authors determined that attempting to classify seven different invasive species simultaneously is challenging and difficult. The preprocessing methods proved to reduce noise in the data but none of the processing methods significantly reduced noise. This study is helpful in understanding how hyperspectral imagery can be a useful tool for understanding invasive species distribution patterns but should not yet be utilized to identify multiple species simultaneously. 

Randall, Joshua & Inglis, Nicole & Smart, Lindsey & Vukomanovic, Jelena. (2022). From Meadow to Map: Integrating Field Surveys and Interactive Visualizations for Invasive Species Management in a National Park. ISPRS International Journal of Geo-Information. 11. 525. 10.3390/ijgi11100525.

A widespread issue among many land management agencies is the invasion of non-native species that pose threats to their respective ecosystems. The National Park service is geared towards providing the public with recreational opportunities while protecting the National Parks natural resources. In this study, the authors working for Valley Forge National Historic Park (VAFO) co-developed a Geospatial Meadow Management Tool (GMMT) to assist in the development of a new method for efficiently collecting and analyzing invasive species distribution data. The GMMT aims to provide a quick and efficient way to analyze data while also providing a simple way to understand species distributions. The GMMT incorporated the use of Survey123 for field data collection, ArcGIS Pro for data analysis and ArcGIS Online for data visualization. The GMMT utilized three main applications including simplified layer management, ability to set filters for specific species and specific years and create line graphs responsive to the year and species selected. Using the GMMT, the Valley Forge National Historic Park was able to analyze and display complex spatial and temporal species cover data and eliminate time consuming data entry. 

Tarbox, Bryan & Schmidt, Nathan & Shyvers, Jessica & Saher, D. & Heinrichs, Julie & Aldridge, Cameron. (2022). Bridging the Gap Between Spatial Modeling and Management of Invasive Annual Grasses in the Imperiled Sagebrush Biome. Rangeland Ecology & Management. 82. 10.1016/j.rama.2022.01.006.

Throughout the Western US, native plant communities (specifically the sagebrush-steppe) are quickly being altered by the spread of invasive annual grass species including cheatgrass, medusahead and ventenata. Land management agencies use spatial data to analyze the distribution and abundance of invasive annual grasses to make management decisions. Early detection of invasive annual grasses is key for successful eradication. The authors of this study collaborated with a multipartner stakeholder group to address the needs for advancement of spatial product development to address this issue. The goals for this project were 1) conduct a review of spatial products used for management of invasive annual grasses (IAG) including cheatgrass, medusahead and ventenata, 2) determine barriers of the spatial products and 3) provide recommendations and resources for the development of higher quality spatial products. The stakeholder group tested 23 spatial products with a variety of uses for displaying and interpreting data. Some of the uses included in these spatial products vary from generating maps of percent cover to determining habitat suitability. The stakeholder group found that spatial products improved over time with regards to temporal and spatial resolution. This is important because higher spatial resolution can allow for early detection of invasions of invasive annual grasses. It was also determined that in order to achieve the production of higher quality spatial products, product users would need to assist in product development by providing data from regions where models struggle to assess invasive annual grasses. 

Uden, Daniel & Allen, Craig & Angeler, David & Corral, Lucia & Fricke, Kent. (2015). Adaptive invasive species distribution models: A framework for modeling incipient invasions. Biological Invasions. Online. 10.1007/s10530-015-0914-3. 

Species distribution models (SDMs) are relatively accurate for modeling non-invasive species. However, using SDMs to model invasive species distribution patterns is still relatively inaccurate. In this study, the authors develop a framework for adaptive, niche-based, invasive species distribution model (iSDM) development by reviewing species distribution models, biological invasions and adaptive practices in ecological management. In the 10-step framework, designed by the authors, the hope is that the iSDM development will allow for further improvements in invasive species modeling. The 10 step framework consists of (1) invasion characterization; (2) objectives statement; (3) assumption and uncertainty articulation; (4) scale recognition and assignment; (5) predictor variable selection; (6) modeling technique adoption; (7) autocorrelation supervision; (8) prediction, validation and mapping; (9) management and monitoring; and (10) refinement. Through analysis of these 10 categories, the authors have found that the 10-step iSDM framework recognizes changes in invasive drivers, combines correlative and mechanistic modeling techniques and creates the opportunity for improvement in models and management of invasives. 

West, Amanda & Evangelista, Paul & Jarnevich, Catherine & Schulte, Darin. (2018). A tale of two wildfires; testing detection and prediction of invasive species distributions using models fit with topographic and spectral indices. Landscape Ecology. 33. 10.1007/s10980-018-0644-x. 

Species distribution models (SDMs) are ideal tools for understanding the potential areas that invasive species will overtake. SDMs can model potential habitat suitability for invasive species across a landscape. SDMs take into account many variables including climatic conditions. Typically, these models determine the presence or absence of a species in a particular area. SDMs are useful for distinguishing native species from invasive species. Wildfire plays an important role in the spread of invasive species, since invasive species tend to thrive on disturbance. The western United States is threatened by increased wildfire frequency due to the invasion of Bromus tectorum (cheatgrass) on the landscape. B. tectorum is a highly invasive annual grass species that has changed the fire regime from 60-500 years down to 3-5 years. In this study, the authors identify post-burn sites in the Medicine Bow National Forest in Wyoming, USA. The two burn sites used in the study were from the Squirrel Creek fire and the Arapaho fire. Field surveys were conducted for the presence of cheatgrass in both burn sites. Circular plots were sampled and percent cover was recorded for cheatgrass, forbs, grasses, shrubs, rock and bare ground. Cheatgrass was determined present in a plot if the cover percent was greater than 40%. Plots with less than 40% cover of cheatgrass were considered absent sites. Data used to fit the models included 56 plots sampled with 10 cheatgrass presences for the Squirrel Creek fire and 148 plots sampled with 32 cheatgrass absences for the Arapaho fire. Four species distribution models were used for both the Squirrel Creek fire and the Arapaho fire. From the four SDMs, it was determined that Random Forest (RF) and Generalized Linear Model (GLM) algorithms were sufficient in predicting cheatgrass habitat. 

[vividmaps]

What is the largest country in the world based on the most suitable land...

[jacob-cs]

Natural Language Processing | Dan Jurafsky, Christopher Manning 7강

7 1 What is Sentiment Analysis https://youtu.be/vy0HC5H-484

7 2 Sentiment Analysis A baseline algorithm https://youtu.be/Dgqt62RQMaY

negation을 처리하는 한가지 예이다. 

P(cj)는 특정 class의 문서가 출현할 확률이다. P(wi | cj)는 특정 class 문서내에서 단어 w가 출현할 확률이다. 밑의 공식은 add one smoothing을 적용한 것이다.

sentiment작업에서는 단어출현 유무는 중요하나 출현횟수를 중요하지 않을때가 많으므로 출현유무만을 가지고 계산한다. 이런 형태를 binarized (boolean feature) multinomial naive bayes라고 한다.

7 3 Sentiment Lexicons https://youtu.be/wBE0FE_2ddE

이미 연구자들이 단어들을 다양한 기준을 통해 classify 한 자료들이 많이 있는데 아래에서 확인 할수 있다.

P(w | c)를 P(w)로 나눠줌으로써 다른 단어와 비교가능하게 할수 있다. 이를 scaled likelihood라고 한다. 

위 그림을 통해 no, not, never등의 negation 단어들이 negative 문자에 보다 자주 사용된것을 알수 있다. 

7 4 Learning Sentiment Lexicons https://youtu.be/Z7RxBcpyN1U

여기서는 lexicon을 직접 만드는 과정을 보여준다.

Hatzivassiloglou and McKeown 이 개발한 방법을 여기서는 예제 방법으로 사용한다. 기본 단어와 and, but으로 연결된 새로운 단어들을 추가로 정리해 가는 것이 기본 원리이다.

플러스는 positive, 녹색은 and로 엮여진 경우, 굵은 선은 많이 엮어진 경우. 적색 점선은 but으로 연결되었던 단어들이다.

turney algorithm은 연속된 phrase를 이용하는 방법이다. 

우선 단어들을 phrase로 뽑아내고 이 phrase가 positive 단어중의 하나인 excellent와 얼마나 자주 출현하는지 PMI 값을 확인한다. 또 negative 단어와의 PMI값을 구한다. 

이 두값들의 차가 Polarity 값이 된다. polarity 값은 phrase 가 positive에 가까운지 negative에 가까운지를 말해준다. 문서안의 phrase들의 polarity값을 평규내면 문서가 positive 인지 negative인지 알수 있다. 

첫줄 해설. jj (형용사) 와 nn(명사), nns(복수명사)가 연결된 경우 세���째 단어와는 무관하게 모두 phrase로 추출한다. 

P(x, y)는 동시에 출현하는 확률, P(x)P(y)는 두 단어가 독립이라고 보고 출현하는 확률이다. 즉 완전 독립된 단어라고 본경우에 비해 얼마나 동시에 출현하는지를 나타내는 비율이다. 

하나의 문서에 존재하는 다양한 phrase들의 polarity값들을 평균내는 과정이다.

7 5 Other Sentiment Tasks https://youtu.be/3Eo--0_ocIk

어떤 항목 (aspects)에 관한 sentiment인지를 확인해 가는 작업

문서에서 가장 자주 등장하는 단어가 aspect일 가능성이 있다. sentiment 형용사 뒤에 자주 등장하는 단어가 aspect일 가능성이 있다.

aspect를 위위에서 언급한 방법으로 알수 없는 경우. 손수 labeling하는 경우도 있다. 

data가 불균형한경우 일부로 맞춰주는 경우도 있을수 있다. 갯수를 낮추어 맞추기도 한다.