A Review of the Enviro-Net Project

A Review of the En viro-Net Pro ject ∗ No vem b er 20, 2021 Gilb erto Z. P astorello 1 G. Arturo Sanc hez-Azofeifa 1 Mario A. Nascimen to 2 1 Departmen t of Earth and A tmospheric Sciences 1-26 Earth Sciences Buiding Univ ersit y of Alb erta T6G 2E3 Edmon ton, Alb erta, Canada. gilbertozp@acm.org , arturo.sanchez@ualberta.ca 2 Departmen t of Computing Science 2-32 A thabasca Hall Univ ersit y of Alb erta T6G 2E8 Edmon ton, Alb erta, Canada. mario.nascimento@ualberta.ca ∗ T ext published in: G. Z. Pastorello, G. A. Sanchez-Azofeifa, M. A. Nascimento. Envir o-Net: F r om Networks of Gr ound-Base d Sensor Systems to a Web Platform for Sensor Data Management . Sensors . 2011. 11(6):6454-6479. doi: 10.3390/s110606454 Abstract Ecosystems monitoring is essen tial to prop erly understand their devel- opmen t and the eﬀects of ev ents, both climatological and an throp ological in nature. The amount of data used in these assessments is increasing at very high rates. This is due to increasing a v ailability of sensing sys- tems and the developmen t of new techniques to analyze sensor data. The En viro-Net Pro ject encompasses sev eral of suc h sensor system deploy- men ts across ﬁv e countries in the Americas. These deploymen ts use a few diﬀeren t ground-based sensor systems, installed at diﬀerent heigh ts mon- itoring the conditions in tropical dry forests o v er long perio ds of time. 1 This pap er presen ts our exp erience in deploying and main taining these systems, retrieving and pre-pro cessing the data, and describ es the W eb p ortal developed to help with data management, visualization and anal- ysis. 1 In tro duction Monitoring ecosystems at high spatial and temp oral resolutions still is a c hallenging endeav or. Satellite-em barked sensors that oﬀer regular passes supp ort only coarse resolution monitoring and on-demand high resolu- tion satellite or airborne-based monitoring are still to o expensive to be considered viable options for frequen t data collections. F urthermore, v ali- dation of satellite and airborne measurements against the v alues observed at ground lev el is often diﬃcult to obtain. Ground-based, or in-situ , sen- sor systems for en vironmen tal monitoring ha v e asso ciated c hallenges as w ell [1], but hav e undergone a considerable ev olution recently . Suc h sys- tems are now capable of collecting data at v ery high temporal resolution for v ery sp eciﬁc ecosystems through long perio ds of time. In particular, the use of wireless sensor systems has been sho wn to b e very eﬀective in this type of monitoring [2], from the cost p erspective and increasingly in terms of p erformance and reliabilit y as well. There are man y c hallenges asso ciated with high resolution (b oth spa- tial and temp oral) in-situ en vironmental monitoring, many of which al- ready well recognized in the literature. Rundel et al. [1], for instance, discuss how these net w orks generate more data than can b e managed by traditional methods for ﬁeld research data, with data quality assurance and con trol surpassing capabilities of single individuals dealing with the data, but still b eing required to pro duce high-quality data. The large v a- riet y of problems impacting qualit y can b e more easily detected b y using adequate cyb erinfratructure for automating the detection, which also al- lo ws more timely identiﬁcation of problems in the deploymen ts themselv es. They also argue that, although data storage and retriev al is reasonably easy to attain, publishing and sharing data is not as straigh tforward. Still according to the authors, one of the adv antages of this in tegrated ap- proac h for oﬀering data from m ultiple sensors is the larger w orld view generated, whic h is not possible with single sensors—at least not at these spatio-temp oral scales. The authors also ac kno wledge the imp ortance of training scien tists in using in-situ monitoring tools, the ﬂexibilit y of pow er requiremen ts for these systems (esp ecially wireless) and the use of energy harv esting, problems related gaps in the data (from numerous causes), diﬃcult y to assess precision and ﬁdelit y in suc h systems, and the v alue of commercial av ailabilit y for decreasing costs and scaling up deploymen ts sizes. Hart and Martinez [3] discuss p o w er management, large v olumes of data and required cyb erinfrastructure, beginning of commercial eﬀorts, and data quality con trol as imp ortant issues concerning in-situ environ- men tal monitoring. They also raise additional points that require more w ork, such as assessmen t of environmen tal conditions any equipment 2 needs to withstand them (e.g., temp erature, pressure, vibration); stan- dardization requirements related to data and metadata represen tation; securit y requirements, prev en ting tamp ering with b oth equipment and datasets within the data managemen t systems; and, b etter means for data in terpretation (e.g., by using new methods for data mining). An- other relev an t eﬀort can be found in the rep ort from Estrin et al. [4], who fo cus on cyb erinfrastructure. Key p oin ts include: the nee d for b etter protot yping and design of end-to-end test-beds to allo w v alidation across wide ranges of en vironmen ts, applications and domains; creation of b etter services regarding time sync hronization, in-situ calibration, and adaptiv e dut y cycling, among others; seamless use of high p erformance computing facilities for data processing; tools to improv e supp ort for metadata; and, collab oration eﬀorts as a basis for training new scientists (from multiple domains) and as a mec hanism for sustaining long term deplo yments. This pap er distills our exp erience in deploying and managing in-situ sensor systems within the Enviro-Net Pro ject ( http://www.enviro- net. info/ ). Currently , Enviro-Net includes 39 deplo yments spread through- out nine sites in six diﬀerent countries (Argentina, Brazil, Canada, Costa Rica, Mexico and Panama), and is co ordinated at the Univ ersity of Al- b erta, in coop eration with local partner research teams at each site. The initial goal of the deploymen ts w as to monitor vegetation phenology , the study of climate eﬀects on perio dic biological activity [5], correlating it with en vironmental v ariables, suc h as a v ailability of ligh t, air temperature, etc. These and other v ariables are monitored by diﬀeren t types of sens- ing systems, with the collected data b eing transmitted bac k to Internet serv ers in Alb erta either through a commercial satellite up-link or b eing man ually retrieved from the data loggers and then sent via email, FTP or En viro-Net’s website. The follo wing gives but one example of the applica- bilit y and usefulness of suc h t yp e of systems. F rom the data collected by a com bination of tw o types of sp ecialized solar radiation ﬂuxes sensors, it is p ossible to derive diﬀerent vegetation indexes, which can be used as proxies to monitoring phenological responses. In order to distinguish v egetation distribution, particularly from p erspectives suc h as sp ecies dis- tribution or successional stage, the areas to be monitored are n umerous and relatively small. Similarly , short term eﬀects of isolated climatic phe- nomena (e.g., a rainstorm or sharp changes in temperature) require higher rates of data acquisition. These c haracteristics require higher spatial and temp oral resolutions only ac hiev ed through in-situ monitoring of eac h ecosystem. In this context, detailed discussions of how we dealt with these chal- lenges within the En viro-Net Pro ject form the main con tributions of this pap er, particularly considering the scenario under whic h the pro ject was dev elop ed. The monitored sites are mostly tropical dry forests in remote lo cations, which are c hallenging en vironments for both equipmen t perfor- mance and personnel’s abilit y to work. Also, all deploymen ts are based on inexp ensiv e and commercially av ailable technology , essen tial c haracteris- tics to allow scalabilit y and reproducibility of exp erimen ts. The hetero- geneit y of equipment from diﬀeren t man ufacturers also in tro duce diﬃcul- ties, mainly regarding systems maintenance. Having long term (multiple- y ear) deploymen ts imp ose extra managemen t requirements. Integrated 3 data managemen t, a fourth asp ect, presen ts n umerous c hallenges ranging from data quality control to user interface usability . Finally , and ma yb e the most relev ant aspect, is the issue of high spatial and temporal reso- lutions, considered not only within a single deploymen t, but also among diﬀeren t deploymen ts b oth in the same and diﬀeren t sites. Part of these c hallenges hav e simple individual solutions, ho wev er, from a more holistic p erspective, the in tegration of the solutions for all of them is what actually enables the use of sensors systems for in-situ environmen tal monitoring. After a review of related w ork on Section 2, this pap er describes our so- lutions regarding deploymen ts of in-situ monitoring systems in Section 3, pre-pro cessing and treatmen t of data in Section 4 and data publication and accessibilit y using a W eb-based system in Section 5. 2 Related W ork This section divides related work discussion in to applications (cov ering the motiv ation for in-situ monitoring), deploymen ts (showing experiences in installing and maintaining sensor systems), and data management (com- paring diﬀerent eﬀorts in dealing with the large amoun ts of sensor data generated). 2.1 Applications En vironmen tal monitoring is one of the driving forces b ehind the adoption of ground-based sensing systems, pushing the need for higher spatial and temp oral resolution. Examples of eﬀorts in this direction include: (i) the creation of the National Ecological Observ atory Net work (NEON) [6], whic h aims at studying climate change, land-use change and inv asive sp ecies on a continen tal scale using, among other metho ds and tec hnolo- gies, ground-based deplo yments of sensor systems; (ii) FLUXNET [7], whic h use micrometeorological and ﬂux tow ers to measure exc hanges of carb on dioxide, w ater v ap or, and energy b et ween terrestrial ecosystems and the atmosphere. These initiatives heavily rely on long term ground- based monitoring solutions. FLUXNET has a public data managemen t framew ork called Fluxdata.org [8], whic h also oﬀers ﬂexible metadata sup- p ort. How ever, due to the div ersity of equipmen t and protocols for deploy- men t and data pre-processing, data integration within the Fluxdata.org system is limited, mostly oﬀering access to data on the original format pro vided by the data pro ducers. This limits the possibilities of apply- ing ﬁlters and aggregation op erations to the data or generating derived data products within the system. Although our system also deals with a v ariety of equipment, the deploymen t proto cols are largely uniform, and pre-pro cessing protocols are dev elop ed using a centralized approac h, whic h allo w us to ac hieve a considerable level of data integration within En viro-Net. These and other initiatives, aiming at integration of ground-based monitoring eﬀorts, are leading to an ev olution from single site environmen- tal monitoring in to net works for en vironment observ ation [3]. This evolu- tion culminates with the current vision for a Sensor Web [9–11], encom- 4 passing sev eral types of deploymen ts of sensor systems, interconnecting them globally through a W eb-based in tegration strategy using standards dev elop ed by the Sensor W eb Enablemen t ( http://www.opengeospatial. org/projects/groups/sensorweb ) W orking Group of the Open Geospa- tial Consortium, Inc. (OGC) ( http://www.opengeospatial.org/ ). A small clariﬁcation on the deﬁnition for (wireless) sensor net w orks ma y be in order. Mainly within Computing Science (CS) research [12, 13] and in earlier Sensor W eb related eﬀorts [9], this deﬁnition is narrow er than what is used in this paper. In this more restrictiv e deﬁnition, a (wireless) sensor netw ork is based on no des (also kno wn as “motes”) that ha v e sensing, data storage/processing, and communication comp onents plus a p ow er source. These nodes are usually autonomous and op er- ate coop erativ ely—b y comm unicating amongst themselves—to collect and pro cess data, also b eing programmable, i.e. , able to b eha ve diﬀeren tly according to, for instance, the type of application, p o wer supply condi- tions, environmen tal conditions, etc. Although w e hav e used this t ype of wireless nodes in our deploymen ts, we do not require the capabilit y of of- fering comm unication amongst netw ork’s components. Instead, we adopt the centralized t yp e of pro cessing architecture as classiﬁed b y [12], being more in line with the current Sensor W eb approach to net works [11]. It is suﬃcien t for us, for instance, that the connection of sensing elemen ts b e done at the lev el of in tegrated data products. Applications of in-situ monitoring systems are also the topic of other researc h eﬀorts. Porter et al. [2] presen t a goo d review of the capabili- ties of wireless sensor net works (WSN) to b e applied within the ecological domain. Hamilton et al. [14], while cov ering capabilities of netw orks of sensors applied to ecology as well, also highligh t the idea of ecological ob- serv atories, adopted within NEON. An extensiv e review of in-situ moni- toring eﬀorts is presented b y Rundel et al. [1], classiﬁed according to their area of fo cus: ab o ve ground, under-ground, and aquatic environmen ts. P orter et al. [15] discuss the state of the sensing technology , what can already b e accomplished and a few areas that require more developmen t (e.g., data management softw are and new t ypes of sensors). Precision agriculture is a particularly relev ant application area for p erv asive sens- ing tec hnology . F or instance, Lee et al. [16] ev aluate monitoring applied to sp ecialt y crop, while Matese et al. [17] use wireless sensor net work in vine- y ard monitoring, and Aquino-San tos et al. [18] ev aluate data transmission proto cols in small scale deploymen ts in w atermelon ﬁelds. In this paper, w e discuss aspects that apply to many of these scenarios, particularly if considering them in a long term monitoring p ersp ectiv e. How ever, our fo cus is on practical and logistics asp ects of deplo ying and maintaining equipmen t, retrieving and managing the data, and supp orting analysis of data pro ducts. 2.2 Deplo yments Other research groups ha ve discussed their eﬀorts with ground-based de- plo ymen ts of sensor systems, mostly focusing on the use of wireless equip- men t. A pioneering eﬀort in applying wireless sensor net works w as the habitat monitoring exp erimen t in the Great Duc k Island [19] in the coast 5 of Maine in the United States, deplo y ed to oﬀer a less intrusiv e wa y to study behavior and nesting of seabird colonies. The SensorScope pro ject [20, 21] is another example, taking place mainly in Switzerland. They ha ve described their exp erience with dev eloping the hardware and soft w are for their wireless system, performing tests, and going on deplo y- men t expeditions, along with their arc hitecture and comm unication pro- to cols. With a fo cus on solar energy a v ailabilit y , AdaptSens [22] adopts system-wide lev els of op eration to cope with diﬀeren t amoun ts of a v ailable energy . GreenOrbs ( http://www.greenorbs.org/ ) [23] is a long term ef- fort for monitoring an univ ersity campus urban forest close to Hangzhou in China, using a large num b er of nodes. LUSTER [24] is a system for monitoring ecological v ariables that implements fault-tolerant distributed storage ov er a dela y-tolerant net w ork using an hierarchical architecture; the system also cov ers user interaction b oth in the ﬁeld exp editions and a w eb interface for data retriev al. Another eﬀort [25], aiming at moni- toring the UNESCO W orld Heritage site Mogao Grotto es in Dunhuang, China, implemented a lo w p o wer wireless monitoring system inside the site’s ca v es with a tailored long distance connection to transmit the data bac k to an on-line server. Another W orld Heritage site, a rainforest ecosys- tem in Queensland, Australia, w as monitored b y a wireless sensor net work pro ject [26], whic h served as a protot y pe for future long term deplo yments using similar conﬁgurations. Another in teresting application, monitoring the activities of volcanoes in Ecuador [27, 28], en tails addressing issues such as higher sampling rates (100 Hz or more), need for higher accuracy and more exp ensiv e sensors. Changing the spatial scale a little, monitoring a single redw o o d tree [29] in California in the United States, oﬀered new insigh t in understanding the micro climate surrounding this t yp e of tree. Rep orts on deplo ymen t experiences also focus on the div ersity of problems faced when using wireless sensing equipmen t, suc h as the LOF AR-agro pro ject [30] that exp erienced problems ranging from hardware failures to net w ork proto cols errors and soft ware problems. While deploymen t re- lated eﬀorts in our w ork focus on issues related to managing the life cycle of ground-based sensor data, other w orks [31, 32] bring ev aluations of tec h- nology for wireless sensor netw ork equipmen t, including communication proto cols, pow er consumption and data transmission issues. T o the b est of our kno wledge, none of the deploymen t eﬀorts re- view ed here address the same scenario as ours: having (multi-y ear) long term deploymen ts, based on co op erativ e eﬀorts of sev eral (heterogeneous) teams, using commercially av ailable equipment from m ultiple man ufac- turers, with an integrated eﬀort of data retriev al, quality control and data a v ailabilit y through an easy to use W eb-based platform. W e b eliev e this is a more realistic scenario for ground-based environmen tal monitor- ing eﬀorts. The curren t eﬀorts within the Life Under Y our F eet pro ject ( http://lifeunderyourfeet.org/ ) [33] are the closest to our o wn, also ha ving long term, spatially distributed deplo yments with a W eb-based data visualization interface integrated with geolocation information. How- ev er, they do not seem to deal with heterogeneous equipment and data formats, nor oﬀer ﬁltering/aggregation options, deriv ed datasets or qual- it y information in their data managemen t solution. 6 2.3 Data Management Man y of the challenges related to sensor data management hav e been kno wn for a while [4]. How ever, several technical and non-technical ques- tions still remain unaddressed. Broad scop e pro jects for management of earth observ ation data try to presen t a top-down approach to data man- agemen t. One such pro ject is DataOne ( http://www.dataone.org/ ), an eﬀort tow ards distributed cyb erinfrastructure for Earth observ ation data, bringing together a multitude of data pro viders and consumers. Another eﬀort is our partner pro ject GeoChronos ( http://www.geochronos.org/ ), whic h implemen ts means for sharing (and in teracting with) tools, datasets and libraries of records within the Earth observ ation domain. En viro-Net, ho w ev er, uses more of a bottom-up approac h, oﬀering sp ecialized solutions for the types of data supported, expanding these t yp es as needed. This allo ws data managemen t solutions that are geared tow ards sp eciﬁc needs to answ er sp eciﬁc science questions. Although it is common to think ab out sensor data management as stream data management, with the associated challenges (on-line aggre- gation, classiﬁcation, etc. ) [34], at least within environmen tal research, particularly in ground-based monitoring, this is not a frequent scenario. Most of the current applications based on sensor data use the persp ec- tiv e of historical (or an archiv e of ) time series data. Applications using the stream data p erspective are only b eginning to app ear, and the cur- ren t applications that do require that p erspective—e.g., volcano moni- toring [27]—are still the exception. Data manipulation for most of the curren t applications is done after having the data collected and stored, applying a v ariety of analytical op erations in an oﬄine fashion [8, 35]. Middlew are softw are for automating control of deploymen ts is also the fo cus of current research eﬀorts, in form of arc hitectures for integrating diﬀeren t net work deploymen ts [36], or W eb-based in terfaces for interac- tion with and con trol of wireless deploymen ts [37]. Our fo cus, on the other hand, is on managing the data products rather than con trolling the equipmen t from within our system. The data arc hiv al asp ect of data managemen t in volv es not only stor- age of data, but also retriev al, do cumen tation, access con trol, among other issues. F urthermore, data curation of long-term rep ositories in- v olv es not only handling the data but also helping scientists answering researc h questions and also maintaining the underlying computational in- frastructure [38]. Within En viro-Net, although w e are only b eginning to to devise our long term plans for infrastructure main tenance, our system already oﬀers data access with a num b er of ﬂexibility aspects to foster eﬃcien t use of the data. Eﬀorts on applying digital library practices in supp ort of sensor data managemen t are also gaining acceptance [39]. Is- sues of data quality and integrit y , as w ell as the elements of data collection that aﬀect them, need to b e an in tegral part of such eﬀorts [40], particu- larly from the p erspective of making data do cumen tation av ailable along with the datasets. In this scenario, metadata b ecomes as v aluable as the datasets themselves, from qualit y metadata about deplo yments [41], to oﬀering search and annotation options and enric hing visualization [42]. Finally , Application Programming Interfaces (APIs) allo w data to b e ac- 7 cessed in a programmatic w a y , which can b e achiev ed, for instance, using W eb services interfaces (using Sensor W eb Enablement standards) or using sp ecialized solutions such as a wrapper-based middlewares [43] or REST- based APIs [44]. Data qualit y aspects are an in tegral part of En viro-Net, and are b eing improv ed, particularly regarding documentation and meta- data cov erage. Although data ingestion is largely automated and data access is possible through the W eb user interface within En viro-Net, data access using a programmatic in terface is still under dev elopment. 3 Sensor Systems Deplo ymen ts Apart from a few test installations, all of our deploymen ts are in tended to b e long term, collecting data for a minim um of t wo to three years. The earliest deploymen ts were installed in mid 2007, with the ﬁrst wireless deplo ymen ts installed in mid 2008. All deploymen ts suﬀered from in ter- ruption in data collection on some scale, usually from a few da ys up to a couple of mon ths, dep ending on ho w early the problem w as detected. Ear- lier deploymen ts suﬀered 100% failure rate due to equipment design b eing incompatible with tropical en vironments. Many problems w ere related to unexp ected interactions of environmen tal conditions with the equipment. Ho w ev er, most of the deploymen ts are still op erational today , with secured funding for main taining them operational until at least 2013. Curren tly , Enviro-Net has 39 p ermanen t deploymen ts, plus temporary deplo ymen ts in Edmon ton, Canada for equipmen t testing and calibra- tion. The Biosphere R eserve of Chamela-Cuixmala in the state of Jalisco, Mexico has a tow er (ov erlo oking the top of the canop y) and a wireless understory sensor system. The n umber of nodes in a wireless deploymen t is usually 12, but there are deplo yments with as few as ﬁv e and as man y as 20 no des, eac h node having b et ween three to six sensors eac h. The Santa R osa National Park in Costa Rica hosts tw o more tow ers. The Par que Natur al Metr op olitano in Panama has the most recent deplo yment with 24 thermocouples monitoring leaf temp eratures. In Brazil, three sites hav e deploymen ts: the Mata Se c a State Park , the Serr a do Cip´ o National Park , and the Envir onmental Pr otection A re a of the Pandeir os River , all located in the Minas Gerais state. The Mata Seca park hosts ﬁv e to wers and eight understory deplo yments (including four wireless de- plo ymen ts), all in the c err ado ecosystem, which is similar to a sav anna; three understory deploymen ts are active close to the P andeiros riv er, also a c err ado ecosystem; and, Serra do Cip´ o park has ﬁve tow ers plus sev en understory deplo yments, tw o of which using wireless systems, cov ering natural grasslands and forest vegetation in the cerr ado . Finally , three deplo ymen ts are operational in the pro vince of San Luis in Argen tina, a phenology to wer monitoring a grassland ecosystem, and one to wer and one wireless understory deploymen t installed in a adjacen t chac o ecosystem. Tw o more wireless to wers are op erational chac o and grassland ecosys- tems in the province of C´ ordoba, Argen tina. Three more deploymen ts are exp ected to start data collection in 2011 in the province of San Luis. Although the ideal spatial scales for man y applications requires higher n um b ers of no des deploy ed to b e considered high spatial densit y—more 8 in line with our plans for future sensor netw orks—the intermediary step with 5–20 nodes per deplo yment was necessary to prov e this kind of scale is feasible in remote lo cations with long term deploymen ts. These are, how- ev er, dense enough to c haracterize many ecosystem level b ehavior (suc h as response to climatic even ts), and ev en diﬀerences betw een neigh b oring ecosystems. The exp erience acquired in these smaller deplo yments, which is the fundamen tal contribution of this text, serv es as a basis for these larger scales exp editions. The main c hallenge of ha ving deploymen ts across an en tire continen t is without question main taining them. Partnerships with researc h groups based closer to the deplo yment sites pro ved essen tial, with the added issue of oﬀering training to the p eople performing basic maintenance. The small amoun ts of time av ailable for training leads to the c hoice of equipmen t that is simple to use and maintain. Hands on exp erience has pro ven to b e the most eﬃcien t metho d to train new users, sp ecially when focusing on ho w to deal with common problems. Sp ecial atten tion needs to b e giv en to data retriev al and manipulation methods in order to allo w trac king of data problems later in the processing c hain. Do cumentation of our o wn group’s deplo ymen t proto cols and data handling procedures complemented and help ed with equipmen t manuals and speciﬁcations. Regularit y in systems main tenance is k ey in k eeping them running within long term deplo ymen ts. Life exp ectancy and calibration devia- tion for sensors are usually a parameter sp eciﬁed b y the manufacturer. En viro-Net deploymen ts usually hav e tw o maintenance cycles: one for ba- sic ov erall system c heck (and data retriev al for oﬀ-line deploymen ts) and another for complete v eriﬁcation of the equipment. The basic cycle has in terv als ranging from t wo w eeks to tw o months, dep ending on the acces- sibilit y of the site and regularity of visits for other purposes. This task is usually p erformed b y a member of the lo cal research teams and inv olves cleaning the sensors if needed—mostly from dust build-up or obstructions suc h as leav es, insect or bird nests, etc–veriﬁcation of the general health of the system, and data retriev al, usually the most relev ant part in a ba- sic main tenance cycle. The complete cycle in terv als ranges from 6 to 12 mon ths, and allows detection of a broader range of problems—e.g., bat- tery charge retention capacit y . This task is usually p erformed with one more exp erienced tec hnician. 3.1 Sensors and Loggers T ables 1 and 2 list the equipmen t used in our deploymen ts. F or datalogger systems, shown in T able 1, wired and wireless systems are av ailable. In wired systems all the sensors are connected directly to the data logger and the comm unications with it are done mostly through a ph ysical connection using a cable (serial or USB, for instance) to connect to a laptop. F or wired deploymen ts, we mostly used Onset Computer Corp. ( http://www. onsetcomp.com/ ) data loggers; speciﬁcally the HOBO Micr o Station , the HOBO U12 Series and the HOBO U30 Series mo dels w ere employ ed. Wireless systems, on the other hand oﬀer diﬀerent strategies to elimi- nate the need for cabled connections. As an example, the equipmen t man- ufactured by Olsonet Communic ations Corp or ation ( http://www.olsonet. 9 com/ ) oﬀers t wo t yp es of nodes: a collector and an aggregator. The former is connected to the sensors and is resp onsible for wirelessly transmitting the readings to the aggregator, which works as a cen tralization point for the data collection also dubbing as a short term data logger. The ag- gregator, ho wev er, requires a cable connection for setup or data recov ery . A diﬀerent strategy is used by the equipmen t manufactured b y Micr oS- tr ain, Inc. ( http://www.microstrain.com/ ), where each ENV-Link TM no de works as an individual data logger, but the connection to these no des for setup and data retriev al is done through a wireless connection. The storage capacity for samples in both t yp es of loggers usually match the p o wer consumption c haracteristics to ac hieve similar longevit y in ﬁeld deplo ymen ts. As discussed later in this section a satellite up-link and a con tin uous battery recharging capabilit y (e.g., using solar panels), would allo w ev en longer time spans. Ho w ev er, since in practice maintenance is necessary long b efore these limits are reac hed, battery and storage life- times are not a limitation for most of these t yp es of equipment. Logger Model Connectivit y Storage Memory P ow er (Battery T yp e) Est. Longevity (a) Onset U30 wired data and setup 512 KB Int. (4.5 or 10 Ah , 4 V) + Solar solar panel (b) Onset U12 wired data and setup 43,000 samples (64 KB) Int. (CR-2032 lithium 3 V) 10–12 months Onset Micro Station wired data and setup 512 KB Int. (4 x AA 1.5 V) 10–14 months Olsonet Collector wireless data / no setup 256 KB Int. (2 x AA 1.5 V) 4–5 months Olsonet Aggregator wireless data / wired setup 2 GB (remov. SD card) Ext. (7–12 Ah) + Solar solar panel (b) Microstrain ENV-Link wireless data and setup 360,000 samples Int. (650 mAh) + Ext. (9 Ah) 10–14 months T able 1: Dataloggers summary . (a) Estimated longevity with 15 min utes sampling; (b) Dep endent on sun light av ailabilit y . T able 2: Sensors summary . Sensor Model V ariable (Unit) Sensor T yp e Range Accuracy Sensirion SHT-75 T emp. ( ◦ C) silicon bandgap − 40.0–123.8 ◦ C 0.3–1.5 ◦ C Rel. Hum. (%) capacitive humidit y 0–100% RH 1.8–4.0% RH Onset S-THB-M00x T emp. ( ◦ C) silicon bandgap − 40.0–75.0 ◦ C 0.2–0.7 ◦ C Rel. Hum. (%) capacitive humidit y 0–100% RH 2.5–4.5% RH Onset RG3-M Rainfall (mm/h) tipping buck et max 1,270 mm/h 1.00% Onset S-LIA-M003 P AR ( µ mol/m2/sec) (a) photons detector 0–2,500 µ mol/m2/sec (c) 5.0% or 5 µ mol/m2/sec Onset S-LIB-M003 Solar Radiation (W/m2) silicon photov oltaic detector 0–1,280 W/m2 (d) 5.0% or 10 W/m2 Apogee SQ-110 P AR ( µ mol/m2/sec) (a) photons detector 0–2,000 µ mol/m2/sec (c) 5.00% Apogee SP-110 Solar Radiation (W/m2) silicon photov oltaic detector 0–1,100 W/m2 (d) 5.00% Decagon ECH2O EC-5 Soil Moisture (VWC (b) ) 70 MHz capacitance/frequency 0–100 % VWC 1.0–3.0 % VWC (a) Photosynthetically Active Radiation; (b) V olumetric W ater Conten t; (c) F or wav elengths between 400 and 700 nm; (d) F or wav elengths b etw een 300 and 1,100 nm. The biggest adv antage of wired equipmen t is reliabilit y , b eing in use longer, and tested under man y com binations of conditions. Besides lim- ited spatial cov erage, when compared to wireless systems, the most prob- lematic aspect of this technology is accessibilit y . Everything requiring a ph ysical connection b et ween the logger and the laptop with the control soft w are, ha ving to climb up a tow er to p erform tasks as routine as re- trieving data is a somewhat serious limitation. Even using longer cables for the sensors, which still hav e a limited maximum length on accoun t at- ten uation of the electric signal, to wers for higher canopies require clim bing to access the logger. 10 F or en vironmental monitoring, the ma jor adv antage of wireless sensor systems is the possibility of cov ering larger areas, without giving up high spatial and temp oral resolution, and at a reasonably lo w cost. One lo w p oin t of the tec hnology is that it is still fairly new as a commercial product, and still needs some adaptation. Errors in communication protocols, radio range limitations, pow er managemen t related issues, lac k of features in the con trol softw are pack ages, and breac hes in w eather pro oﬁng cases weigh t in at the cons for wireless systems. How ever, our exp erience sho ws the tec hnology has already reached the tipping point to b ecoming viable for use in long term, harsh en vironmen t deplo ymen ts. Commercial av ailabilit y of wir eless sensor networks (WSN), as a tec h- nology , is still limited. Although the original ideas for WSN— i.e. , large n um b er of general purpose no des distributed in v ery dense deplo yments, randomly placed, almost weigh tless, and disp osable—ha ve y et to mate- rialize [20, 28, 45], wireless technology used in conjunction with sensory equipmen t is proving to be in v aluable in monitoring larger areas at the scale of a single ecosystem. T able 2 lists the main sensors use in our deplo yments, which are are w ell kno wn, commercially a v ailable, inexp ensiv e, and based on established tec hnologies. With the v ariables listed, it is p ossible to extract plen t y of deriv ed information from them, such as vegetation indexes and light absorption patterns for photosyn thesis. In our deplo yments, we used so- lar radiation sensors pro vided by Onset and Ap o ge e Instruments, Inc. ( http://www.apogeeinstruments.com/ ); air temp erature and relativ e h u- midit y sensors b y Sensirion Inc. ( http://www.sensirion.com/ ) and On- set (The Onset temperature and relative humidit y sensors used are repack- aged Sensirion sensors); and, soil moisture sensors by De c agon Devic es, Inc. ( http://www.decagon.com/ ) Low er cost sensors systems usually do not oﬀer calibration options for the user; they ha ve their calibration ad- justed at the man ufacturer (who usually oﬀer recalibration services). 3.2 Deplo yment Conﬁgurations Within the Enviro-Net Pro ject there are currently t wo main t yp es of de- plo ymen ts: phenology tow ers and understory installations. A phenology to w er uses t wo solar radiation ﬂux sensors (also called p yranometers), measuring w av elengths b et w een appro ximately 300 to 1,100 nm, and tw o Photosyn thetically Active Radiation ﬂux sensors (or P AR sensors), which measure wa velengths betw een approximately 400 to 700 nm. Ratios of these measuremen ts can b e used to derive v egetation indexes such as Normalized Diﬀerence V egetation Index (NDVI) or Enhanced V egetation Index (EVI)–see, for instance [46–49]. Such indexes can be used as prox- ies to monitor vegetation phenology . Understory deplo yments are used to monitor the conditions b elo w the canop y lev el, and usually co ver a larger area. Figure 1 shows the schematics of a phenology to wer on the left, with t w o P AR sensors and tw o pyranometers, one of each measuring incoming solar radiation and one of each measuring reﬂected solar radiation. The righ t side of the ﬁgure is a photo of one phenology tow er installed in Brazil, which raises the sensors eigh t meters from the ground, six meters 11 ab o ve the canop y . Figure 1: Phenology to wer schematics (left) and a to wer in Brazil (right). Most radiation ﬂux sensors ha ve view angles of up to 85 ◦ from zenith (when oriented up, i.e. , measuring incoming radiation) or nadir (when orien ted do wn), with a uniform 360 ◦ rotation. With that, the radius that aﬀects the readings is up to around ten times the distance ( h ) betw een the sensor position and the surface being monitored ( i.e. , radius = tan(85 ◦ ) × h ). F or our deplo yment, we usually ha ve at least ﬁve meters b etw een the top of the canop y and the sensor measuring the reﬂected radiation (8 to 15 m in total), leading to a co verage radius of at least 50 m in the monitored area. Obstructions within the range of a sensor in terfere with the reading and might not be easy to iden tify from the data only—e.g., higher canopy of adjacent ecosystems or a nearby tow er with other instruments may in terfere with sensor measuring incoming radiation. A sensor measuring radiation reﬂected from the canop y is more susceptible to in terference— e.g., the p ositioning of solar panels, whose reﬂectiveness greatly aﬀect readings. Large panels should b e positioned outside of the interference radius, while smaller panels can b e positioned at the same height as the sensor for no in terference. Note that it is diﬃcult to p osition radiation sensors and solar panels at diﬀeren t orien tations, since both should use the optimal exp osure angle to the sun, roughly North, in southern latitudes, or South, in northern latitudes. Monitoring the conditions under the canopy level, i.e. , understory de- plo ymen ts, allows assessing a diﬀeren t range of micro-climatic conditions and also soil condition—e.g., temperature and moisture lev els. Understory deplo ymen ts are usually easier to access, and with that, they are useful for v alidating the readings observ ed in a tow er and also as a bac kup for certain v ariables in case of sensor malfunction in a to wer. Using wireless systems substantially increases the spatial co verage of understory deploy- men ts with a fraction of the increase in cost and eﬀorts to retrieve data and main tain the system. Figure 2 depicts an example of such a wireless deploymen t on its left side. On the righ t side, it sho ws a no de deploy ed in the chac o ecosystem in Argentina. The heigh t at whic h the sensors are installed in this case 12 is also determined by the canopy’s height, usually ranging from right on the ground (e.g., for grasslands) to 1.5 m for taller canopies. One example of application that relies on the spatial co verage and resolution of understory wireless deplo yments is deriving Leaf Area Index (LAI)—see [48, 50], for instance. LAI, along with Plan t (P AI) and W oo d (W AI) Area Indexes [51], are imp ortan t indicators of vegetation productivity , being also used as a reference for crop gro wth rates. Combining readings from a phenology to wer with understory readings of absorbed solar radiation ﬂuxes, it is possible to derive ND VI for the lo cation of each no de. Using ND VI and kno wing an appropriate conv ersion factor, c haracteristic to eac h ecosystem, it is p ossible to calculate LAI for eac h no de [48]. This allo ws the creation of maps of very high spatial and temp oral resolutions for b oth ND VI and LAI. Figure 2: Understory sc hematics (left) and a no de in Argentina (right). Ha ving the option of deploying a large num b er of sensors in a giv en area also raises the question of ho w to distribute these sensors. W e ha ve adopted three diﬀerent strategies to spatially distribute no des and their sensors. Figure 3 illustrates these strategies. The ﬁrst approach, shown in the left, is in tended to monitor a linear region along a transect. This is particularly useful for monitoring transitions betw een ecosystems or ex- p osition to diﬀerent conditions within the same ecosystem. The center of the ﬁgure shows distribution of nodes in concen tric circles, whic h is some- times called a “star” deplo yment. This type of deplo yment is used mostly to monitor conditions around a point of in terest, usually corresponding to the footprint of phenology or carbon ﬂux tow ers, allo wing com bination of measuremen ts from both deplo yments. A third strategy is to deplo y nodes in a grid, cov ering a potentially larger area of interest. Regularly spaced grids are useful for uniform monitoring throughout an area. Ho wev er, irregular grids can also be useful when sp ecial conditions o ccur within a region of interest. Examples include part of an area that is also b eing 13 monitored b y other exp erimen ts (e.g., leaf collection for c hlorophyll mea- suremen ts); or patches aﬀected b y ﬁre and monitoring their recov ery is of in terest. (a) (b) (c) Figure 3: Deplo ymen t strategies: ( a ) transects; ( b ) concentric circumferences; and ( c ) grids. 3.3 Deplo ying Sensors Systems F rom a logistics persp ectiv e, installing tow er and understory systems ha ve fairly diﬀerent characteristics. Phenology to wers reac hed up to 15 m in one of our deplo yments, with 9 m b eing the most common heigh t. Selecting the lo cation for installing a to wer that high must take in to account the represen tativ eness of the ecosystem, the impact of building it, and the accessibilit y to bring its parts to the site. Another imp ortan t issue is the uniformity in the heigh t of the canopy . T o o m uch v ariation in the tree heights will lead to scaling problems in the data, an area with taller v egetation will b e con tributing signiﬁcan tly less to the readings. When installing a phenology tow er in tended to b e used in a long term data collection, the growth of the vegetation should also be tak en into accoun t. Y ounger ecosystems might grow considerably at in terv als as short as one y ear, forcing heigh t upgrades to a to w er. The heigh t of the canop y is also a concern for understory deplo yments. Ecosystems with lo wer canopies, suc h as grasslands, require that solar ra- diation ﬂux sensors b e positioned almost adjacen t to the ground, while taller canopies allo w sensors in a higher p osition (0.60 to 0.10 m are com- mon heigh ts). F or wireless deploymen ts, the node is usually installed in a higher position to improv e radio signal range, while the sensors are deplo y ed at the appropriate heigh ts. Although it migh t seem like a trivial task at ﬁrst, correctly p ositioning the sensors should take into consideration a n um ber of factors. One issue is the creation of unnatural sources of shade (e.g., from the p ole where the node sits) in to the sensor. F or deplo ymen ts in the northern (southern) 14 hemisphere, positioning radiation ﬂux sensors South (North) of obstruc- tions av oids this issue. Air temp erature and relative humidit y sensors are also aﬀected b y their positioning. Besides b eing hosted at solar radiation shields and b eing p ositioned as to allo w for air circulation, they should also keep some distance from radiation absorbing materials. Most of the w eatherpro oﬁng cases, for instance, absorb non-negligible quan tities of solar radiation. W e had cases of temp erature deviations of up to 20 ◦ C b ecause of a dark w eatherpro of case. One crucial aspect to sensor systems deploymen ts in tropical ecosys- tems is the exp osure to constan tly high relative h umidity . V alues betw een 90%–100% are common in these en vironments. Com bined with high tem- p eratures, this condition transformed man y weatherproof casings in to h u- midit y traps. The main problem w as actually the diﬀerence of in ternal and external pressure in the cases. That made previously air tight cases ab- sorb humidit y while balancing the pressure, exp osing the inte rnal circuits and connectors. Both for loggers and sensors, ev en cases designed and tested to work underw ater were susceptible to this problem. Adopting pressure relief v alves signiﬁcan tly attenuated the problem, even though sometimes they can get clogged with dirt and stop working. Another adopted practice that also helped reduce this problem was to use silicone- based adhesives to seal b orders and openings, around sensor cables and also around the sensors themselv es. F or wireless sensor systems, testing the range of the radio system at the actual deploymen t site is essential. V egetation distribution and ter- rain contours are diﬃcult to predict b eforehand and hav e a signiﬁcant impact in the radio range. Tw o ma jor asp ects hav e sho wn to be of par- ticular relev ance when conducting this kind of test. Firstly , if the t yp e of batteries used decrease the v oltage oﬀered to the system with time, the tests should not be conducted with new batteries. A more accurate test of radio range is ac hieved using more realistic battery levels—e.g., lev els of battery similar to when a deploymen t running for more than half of the ex- p ected battery life. In case of rechargeable batteries, the c harge lev el used should b e the a verage lev el the batteries would ha ve when going without c harge for the maxim um foreseeable p eriod. F or tropical dry forests the maxim um p eriod without non-negligible sun light exp osure for charging batteries through solar panels is around t w o days. It is worth of note that regular alk aline (zinc and manganese dioxide) batteries, widely adopted to p o wer no des in sensor netw orks, do c hange their v oltages dep ending on their lev el of c harge. The second asp ect in terfering with radio range is related to the veg- etation density , particularly to changes in densit y throughout the sea- sons. Radio range is greatly aﬀected by branches and leav es in the line of sigh t of the signal. Ranges of up to 300 m in a level and op en ﬁeld can b e reduced to as little as 15 m (a factor of 20 reduction) simply by ha ving a somewhat closed vegetation. In particular, radio frequencies at 2.4 GHz are sev erely attenuated b y trees and leafs. This frequency is adopted by several wireless sensor systems, including the ZigBee Al- liance ( http://www.zigbee.org/ ) communications proto cols (based on the IEEE 802.15.4 Wireless Personal Area Net work standard), widely used in these systems. F urthermore, it is v ery usual for deplo yment campaigns 15 to take place in the dry season, when rainfall is less of a concern for the sc hedule in deplo yment plans. How ever, foliage of deciduous v egetation can be at muc h low er lev els than it will be in the wet season, whic h can signiﬁcan tly aﬀect the range of radio signal. There is no deﬁnite solution to address the vegetation c hanges, since simulating the conditions of a diﬀeren t season is diﬃcult. Monitoring the ov erall netw ork health, whic h can be done in its simplest form b y detecting gaps in the data, and rep o- sitioning no des when necessary has been the b est measure to address this issue in new ecosystems. These, in turn, serv e as a reference for future deplo ymen ts in similar ecosystems. Seasonal change also can hav e an unexp ected impact in the visibility of no des and sensors. When installing sensors in the dry season, there are few obstructions and less color v ariability on the landscape. This mak es visibilit y reasonably go od. How ever, areas that signiﬁcan tly change their v egetation cov erage or areas that hav e dense v egetation can become quite c hallenging from the point of view of visibility in the wet season. Using colorful mark ers—red or y ellow ribb ons or pain t are eﬀectiv e for this—can sa v e a lot of time when trying to ﬁnd nodes and sensor that hav e been deplo y ed for a while. Not relying solely on the GPS to locate small pieces of equipmen t such as individual nodes and sensors can be the diﬀerence b et ween returning to base camp b efore or after sunset. One asp ect of using suc h mark ers that w as not tak en into account in this work is the increased attractiv eness color mak ers migh t exert on animals (particularly insects). 3.4 Retrieving Data Data from ground-based sensor systems can be retrieved either in-situ or remotely . The former in volv es expeditions to the deplo yment sites, which can b e v ery exp ensiv e. How ever, if the site is already b eing visited in a regular basis for other reasons (collecting leaf samples, for instance), this migh t b ecome more feasible. Most of our curren t deplo yments are w orking in this scenario. This has pro ven to b e quite an adv an tage from the p erspective of maintenance of un tested systems, allowing early detection of problems with equipmen t. With equipment pro ven to work w ell, using a remote solution is probably more cost and time eﬀectiv e. Collecting data remotely migh t be ac hieved in a n umber of wa ys. One p ossibilit y is using a dedicated long range wireless communication system—e.g., by using a WiFi connection with rep eaters—to transmit data at regular interv als to a computer installed in a lo cation with p er- manen t p o wer supply . If there is also In ternet connectivit y , the data can b e forw arded to on-line permanent arc hiv al systems. This alternativ e usu- ally has a signiﬁcant o verhead of maintaining the lo cal computer and the long range comm unications system running. Another alternative is to use cellular netw orks with data capabilities. Although cellular co verage is not goo d in more remote areas, some regions ha v e enough connectivity to allo w data transmission in a fairly regular fashion. Using higher gain an tennas improv es signal reception, but the system must b e prepared to go through reasonably long p eriods with no connectivit y , preserving all data for delay ed transmission. Since an actual 16 In ternet connection is provided with a cellular connection, the data can b e transmitted directly to on-line arc hiv al servers. A third type of remote data retriev al solution inv olves using a satellite up-link. This approac h is also sub ject to communications failure (e.g., if there is too muc h cloud cov erage). The connectivity provided here usually is not to the Internet, but connectivity to a service pro vider that receiv e the data from the satellite. This provider in turn mak es the data accessible, often oﬀering automated wa ys of retrieving the data from their on-line serv ers. In our case, systems that ha ve pro ven to w ork consisten tly w ell ha v e b een equipped with a satellite transmitter. Remote connectivit y allo ws not only automated data retriev al, but also some level remote op eration of the equipment. Options of stopping and starting the logger, setting sampling and storage rates are often a v ailable. In a few cases, it might b e in teresting to b e able to set other parame- ters remotely , particularly with wireless systems. Researc h pro jects ha ve explored conﬁguring deplo yments remotely [37], ev en reprogramming log- gers and collection no des in some cases [20, 36]. This lev el of ﬂexibilit y in remote deplo yment conﬁguration, how ever, is not y et commercially a v ail- able. 4 Data Pre-Pro cessing and Cleaning When considering the volume of data generated by curren t sensor sys- tems, automation of data management related tasks within a prop er com- putational infrastructure is of paramoun t imp ortance [1, 3, 4, 10, 33, 52]. Ho w ev er, actual datasets generated b y sensor systems migh t present a v a- riet y of problems and exceptions, whic h are often diﬃcult to foresee. This is a sev ere drawbac k in attempts to automate the ﬁrst data managemen t phase: ingestion of data into any computational data managemen t sys- tem. This sort of problems are often dismissed as being “implemen tation details”, but their implications can actually aﬀect data qualit y parameters and mo dels to store and distribute the data. In higher end (exp ensive) and/or homogeneous equipment this sort of problems are usually easier to tackle. Ho wev er, in a setting like ours, using equipmen t from diﬀerent man ufactures, in a highly distributed eﬀort, with an aim at low limits for equipmen t and main tenance costs, these issues are fairly commonplace. The implementation of solutions for problems with raw datasets are usually carried out within a data pr e-pr o c essing (or data cle aning ) phase. Although these terms usually encompass explicit data qualit y veriﬁcation or remov al of erroneous readings (e.g., v alues outside the scale measured b y a sensor), this section only considers problems that actually preven t (or are diﬃcult to trace after) the ingestion of the data into a data man- agemen t system. When compared to classiﬁcation scales usually adopted in describing Earth observ ation data products, after the corrections in this section, the dataset should treated as “ra w” data, or, as b eing at Data Pro cessing Lev el 2 in the National Researc h Council (NR C) Committee on Data Management and Computation (CODMAC) [53] classiﬁcation, or to Data Processing Level 0 used by NASA ( http://science.nasa.gov/ earth- science/earth- science- data/data- processing- levels- for- eosdis- data- products/ ). 17 The next paragraph discuss the problems w e had to handle when prepar- ing our datasets. 4.1 Sync hronization Keeping correct temporal information for timestamping readings from dis- tributed sensors can b e really c hallenging, not to men tion correcting time deviations after recording the data [54]. Time synchronization is an is- sue b oth at single deplo yment, with m ultiple collectors and/or loggers, and across deploymen ts. Within a deploymen t, hardw are imprecision and heterogeneous initial sync hronization methods are the tw o main causes of sync hronization problems. Time keeping in electronic equipment is based on crystal oscillators, whic h can deviate from their standard frequency with en vironmental conditions, esp ecially temperature. This causes the time measurements to deviate as w ell, and aﬀects almost all t yp es of data logging equipmen ts. In this case, the error is proportional to the sampling rate, which for applications such as seismology , with high sampling rates are, these errors are quite signiﬁcan t. F or long term environmen tal mon- itoring, this can also be a problem. One solution is to ha ve an accurate reference time k eeping and a mechanism to k eep the sync hronicity among loggers. Possible solutions include ha ving more precise equipment kept at a less exp osed location or using GPS time as references. A few wireless comm unication proto cols ha ve time synchronization features em b edded within their message exc hanging mec hanism [55]. When dynamic time synchronization against a reliable reference is not feasible, the initial sync hronization metho d is the basis for all time information within a deploymen t. This is the most common scenario for our current deplo yments, with the usual mechanism for synchronization b eing based on the time information from the computer with the control soft w are used to start a deploymen t. Therefore, the time information in that computer should b e sync hronized (e.g., by using Netw ork Time Proto col, IETF RF C 5905 ( http://tools.ietf.org/html/rfc5905 )). Data comparison from diﬀerent deplo yments at small temp oral reso- lutions must tak e into accoun t p oten tial synchronization errors. How ever, since sampling rates are commonly higher than desired temporal resolu- tion, most data analysis is done with aggregated data instead of the en tire dataset, whic h attenuates the eﬀects of the time sync hronization related errors, particularly when looking at hourly or even daily a verages. Similar to other reports [54], w e also experienced p o w er source related sync hronization problems. Time measurement in some logging equip- men ts can b e aﬀected by pow er outages or lo w voltages from the p o wer source. Some t yp es of equipment use the main p ow er to keep time mea- suremen t running and, although time measuremen t usually requires v ery little p ow er, if the supply is in terrupted, the equipment’s clo ck gets reset. Curren t data logging equipmen t and control softw are oﬀer p oor sup- p ort to address time sync hronization problems. Man y of them don’t even let the user see what is the current time in logging system to manually c hec k for time drifts. But this is ev olving in con trol softw are for wireless systems, since these suﬀer more noticeably from time related problems. 18 4.2 Time Zones When dealing with deploymen ts that are geographically distributed through- out v arious timezones, establishing the correct lo cal time can become an issue. Once again, relying on a computer’s time as a reference to times- tamp the readings is a ma jor cause of errors. Diﬀeren t v ersions of operat- ing systems hav e diﬀeren t levels of automation regarding time zones and da yligh t sa ving time conﬁgurations, often allo wing users to change these man ually . Therefore, besides ha ving the correct time on the computer, as already discussed, wrong conﬁgurations of time zone and c hanging conﬁg- urations for da ylight sa ving times can also lead to inconsistencies such as: ha ving data for a single deplo yment timestamp ed with diﬀerent dayligh t sa ving times, or diﬃculties determining which is the correct lo cal time when comparing data for deplo ymen ts in diﬀeren t time zones. F or our deploymen ts, when issuing ﬁeld laptops, time conﬁgurations alw a ys adopt the local standard time for the site, disabling automatic c hanges to da ylight saving time. How ever, even rugged ﬁeld laptops fail, and temp orary misconﬁgured replacements can be used. Or, an even less elab orate problem, which happens often, new users get confused by seemingly “wrong” time settings and c hange the time conﬁgurations. It is possible, how ever, to c hec k time zone and dayligh t saving times against sun time. This is done by comparing sev eral days of sunrise time from data collected b y solar radiation sensors to exp ected sunrise times for the lo cation. This method is not accurate enough for correcting for hardw are time drift, for instance, but is goo d enough to correct for one or more hours shift in the timestamps. This v eriﬁcation is p erformed on all of our datasets before ingesting them into our data managemen t system. 4.3 Data F ormat V ariation One burdensome problem of dealing with data from diﬀerent t yp es of equipmen t is handling c hanging data formats or a v ariety of possible for- matting errors. The ﬁrst of such types of problems to b e addressed are changes in the data format made b y the equipment man ufacturer. A considerable amoun t of format changes from manufacturers are not do cumen ted ade- quately with new versions or softw are updates. Unfortunately , this type of problem needs to be addressed case-by-case. One problem that was surprising to us is that some types of failures in the sensors themselves can generate errors in the data format b y , for in- stance, changing the num b er of data columns in a record. As an example, this could mak e a record that should hav e three data columns (e.g., read- ings on temperature, humidit y and solar radiation) actually hav e extra or missing columns. Similar eﬀects can be caused by connector designed to b e generic and supp ort diﬀerent sensors: a sensor behaving in some unex- p ected wa y may cause the data collection no de or the logger to p erform incorrect conv ersions or even generate soft ware errors that will aﬀect the data format. In wireless equipment, w e hav e also seen the data format b eing drastically changed b y problems in the transmission of the data. In the presence of radio interference, usually created b y the op eration of 19 higher pow ered wireless equipmen t in pro ximit y of the deploymen t, the data transmission gets compromised, generating errors in both the v alues of the readings and the structure of a record. All these types of errors can cause failures in the ingestion softw are or, worse, hav e errors introduced in the data ingestion process without an y w arnings. Our solution to this was to mak e ingestion softw are moni- tor for format c hanges and generate informative error messages, allo wing iden tifying problems b efore ingestion. 4.4 Pro venance Giv en the tailored nature of data pre-processing steps described in this section, it is diﬃcult to keep standardized prov enance information and ev en harder to automate collection of this type of information, as also made evident in [40]. In our curren t data management solution, sim- ple free text description ﬁelds are used to keep track of pre-pro cessing steps and choices. Nonetheless, with a ﬂexible metadata sp eciﬁcation to ol, suc h as the one created for our partner pro ject GeoChronos ( http: //www.geochronos.org/ ) [56], it w ould b e possible to add speciﬁc ﬁelds to do cumen t evolving aspects of the pre-pro cessing steps. Although diﬃcult to obtain and main tain, documentation of the pre- pro cessing steps are imp ortan t to iden tify not only errors in the pre- pro cessing itself, but also problems with the deploymen ts. F or instance, the app earance of to o man y erroneous records from a wireless data collec- tion node are p otential evidence of problems with the sensors, the sensor connections, the no de hardware or radio in terference sources in the sur- roundings. The latter problem migh t even indicate aﬀected readings from other no des that w ould otherwise go unnoticed. 5 W eb-Based Data Analysis A resourceful and easy to use data management system is the last piece of our solution for large scale in-situ monitoring. The pre-processing step presen ted in the previous section allo ws the data to b e ingested into the system, b eing stored in an in tegrated represen tation. Then users can in- teract with the system having access to data ﬁltering, aggregation and other more sp ecialized processing operations. Data visualization and re- triev al are oﬀered for data at all processing lev els after pre-processing, pro viding a ﬂexible mechanism for analyzing the data within the system or using other tools with the data already narrow ed do wn to the parts of in terest. This section discusses these issues, also considering data qualit y and user in teraction aspects. 5.1 Uploading Data The task of data ingestion can be automated for datasets that require pre- pro cessing steps known beforehand. Automated data ingestion metho ds are particularly useful with deplo yments that ha ve automated data re- triev al, as is the case for data retriev al using a satellite up-link and an the 20 resp ectiv e In ternet service for getting the data. How ever, new datasets or datasets that needed sp ecialized pre-pro cessing b efore getting ready to b e ingested need a ﬂexible mec hanism to map a v ailable data to the in tegrated represen tation of the data in the system. Prop erly handling errors and exceptions in the data ingestion processing is necessary from both user ex- p erience and data quality persp ectives. Automated data ingestion needs timely error generation so the user resp onsible for the deplo yment is k ept informed and and can take correctiv e action. Informative descriptions of errors helps the user iden tify and diagnose the error causes, which is par- ticularly imp ortan t for manual ingestion of data that w as pre-processed in an y non-standard w a y . Another aspect to b e considered is the documentation of the pro- cess for every dataset upload. Metadata regarding date and time, data source, user, pre-pro cessing options, among others, help identify sources of error suc h as fault y time related correction or application of outdated pre-pro cessing methods. Most of these metadata can b e collected auto- matically by the system, which un burdens the user and preven ts missing information from less thorough users. Only with an integrated data representation model it is p ossible to oﬀer a common user in terface, types of ﬁlters, aggregation options and any other op eration for manipulation of data. Data from diﬀeren t instrumen ts, deplo ymen t conﬁgurations, retriev al strategies, etc. , need to b e stored uniformly so all the system’s features are a v ailable for all datasets. 5.2 In teractive Filters and Op erators Ha ving the data uniformly stored in a rep ository , the users can start tai- loring datasets to their needs. The most basic functionalit y to allo w this tailoring is being able to apply ﬁlters (e.g., only data within a range of v alues or with lo w error rates) and aggregation op erators (e.g., sho wing daily or monthly v alues) to the datasets. Without adequate computa- tional support, many researchers spend da ys to w eeks in this trivial task. Figure 4 sho ws our in terface for a few of these ﬁlters to ac hieve the target data, from the top: whic h sensors to include, which time span to consider, and which times of the da y are of interest. The screen shown is to extract and do wnload a dataset to b e used with other to ols. Several other op- tions are also av ailable, including ﬁltering out errors, sho wing raw v alues (e.g., of v oltages, electrical current, or uncon verted pulse coun ts), ﬁle and con ten t formats, including data deriv ed from the sensor readings, among others. Oﬀering quick and easy access to the (corrected and quality c heck ed) sensor readings from a deplo ymen t is one of the most essen tial features of our solution. How ever, also ha ving data that can b e derived from from these readings as easily accessible is what sho ws the actual p oten tial for data management systems lik e ours. Our curren t implemen tation oﬀers the automatic calculation of v egetation indexes, ND VI and EVI, from solar radiation ﬂux sensors using diﬀerent methodologies [46–49]. Other pro ducts are currently b eing integrated into En viro-Net, including LAI, V ap our Pr essur e Deﬁcit (VPD), spatial distribution for F r action of Ab- sorb e d Photosynthetic al ly A ctive R adiation (fAP AR). 21 Figure 4: Data retriev al options. 5.3 Visualization After tailoring a dataset to sp eciﬁc goals, adequate visualization to ols allo w easier understanding of ev ents and trends within the monitored ar- eas. The most basic visual tool is graphing the sensor readings of diﬀeren t v ariables, allo wing visual comparison and insight on the measurements in one deplo yment. How ever, tw o features in our w eb system prov ed to b e in v aluable: graphing of datasets that w ent trough transformations (ﬁl- tering, aggregation and deriv ed data) and graphing across deploymen ts. These graphing options are depicted on the left side of Figure 5, whic h sho ws derived ND VI (using the methodology in [48]) for tw o diﬀerent deplo ymen ts in the Mata Seca State P ark, in Brazil, within a speciﬁc time span, using only readings close to midday (betw een 10:00 AM and 2:00 PM lo cal time), ﬁltering out seemingly cloudy da ys ( i.e. , including only data records when the measured incoming P AR is more than 900 micro einsteins per second p er square meter— µ E/m 2 /s), and aggregating the data in daily a v erages. The right side of Figure 5 sho ws another t yp e of visualization strategy based on the spatial distribution of the readings. The graph on the left site uses a color scale to represent v ariation temp erature across an area co v ered b y 12 temp erature sensors in the Chamela R eserve in Mexico. The graph on the right sho ws the co v erage of the installed sensors (indicating the reliabilit y of the scale), highligh ting sensor failures when these o ccur. 22 Figure 5: Visualization of derived ND VI (left) and spatial distribution of tem- p erature and its reliability (right). Within the speciﬁed time span, the system generated a sequence of images whic h are animated using the con trols at the bottom to show the ev olution of the temperature and reliability distributions through time. This is a useful to ol to observe cyclic (e.g., diurnal or seasonal) c hanges in the monitored areas. 6 Concluding Remarks This paper presented the En viro-Net Pro ject, which addresses a v ariety of issues related to in-situ (or ground-based) monitoring of ecosystems, from the deplo yment of sensors to the deliv ery of processed data pro ducts. A com bination of factors make this pro ject unique: (i) acquisition of data at ecosystem level with high spatial and temp oral resolutions; (ii) long term, ground-based monitoring; (iii) use of heterogeneous, commercially a v ail- able, and inexp ensiv e equipmen t, including wireless sensing tec hnologies; and (iv) integrated data management solution, with a W eb-based user in teraction with data products. This scenario, which is increasingly b eing adopted b y other research pro jects, is describ ed in detail in the pap er, discussing lessons learned and p ointing out asp ects that require attention and could go unnoticed b efore deploymen t eﬀorts are well underw ay . The pap er examines not only tec hnical issues of deploying ground-based sensor systems, but also the logistics behind execution and maintenance of deplo yments, issues related to data retriev al, veriﬁcation and qualit y , and publication of data pro ducts. The pap er discussed evidence that this kind of researc h w as needed, in tegrating solution to from a num b er of researc h eﬀorts and oﬀering a real solution in-situ long term en vironmental monitoring at high resolution temp oral and spatial. Curren t eﬀorts include: improving our deplo yment proto cols to deal with arising problems and simplifying the main tenance related tasks; ex- tending our data management system in order to handle larger amoun ts of data; and adding new data manipulation op erations to oﬀer more derived 23 data products. As future work, w e intend to focus on data pro venance visualization issues, to impro ve understanding of ho w data pro ducts were generated and allowing automation of repro ducibilit y . Another aspect to b e explored in future releases of our data managemen t system is the in tegration of remote sensing data (from satellite and airb orne instru- men ts) in to our common interface [57], allo wing analysis and comparison of these t ypes of data with ground-based data. W e also plan to work on implemen ting programmatic interfaces to allow soft ware-based access to our data b y , for instance, using Open Geospatial Consortium proto cols. Lastly , w e ha v e plans to include monitoring of equipment life expectancy , particularly of sensors and wireless collector no de equipment, in order to create better mo dels for main tenance of long term deplo yments—b y , for instance, increasing the precision of required replacemen t rates for equip- men t. Ac kno wledgemen ts The En viro-Net Pro ject is funded b y the Canadian F oundation for Inno v a- tion and the In ter-American Institute for Global Change Researc h (IAI) CRN I I # 021 whic h is supp orted b y the National Science F oundation (Gran t GEO-0452325). It is also partially supp orted by Cybera and Ca- narie (through the GeoChronos Pro ject), as well as the National Science and Engineering Research Council of Canada (NSERC) Discov ery Grant Program. The authors ackno wledge and appreciate the contributions to the En viro-Net Pro ject, received in v arious forms, b y numerous mem b ers of the lo cal and remote research groups, in particular the students and staﬀ at the Universit y of Alberta, and our research partners at F ederal Univ ersit y of Minas Gerais (UFMG), State Universit y of Montes Claros (UNIMONTES), and State Universit y of S˜ ao Paulo (UNESP), in Brazil; Autonomous National Universit y of Mexico (UNAM), in Mexico; T echnol- ogy Institute of Costa Rica (ITCR), in Costa Rica; Univ ersity of Buenos Aires (UBA), and National Universit y of San Luis (UNSL), in Argentina. 24 References 1. Rundel, P .W.; Graham, E.A.; Allen, M.F.; Fisher, J.C.; Harmon, T.C. En vironmental sensor netw orks in ecological research. New Phytol. 2009 , 182 , 589–607. 2. Porter, J.; Arzb erger, P .; Braun, H.W.; Bryan t, P .; Gage, S.; Hansen, T.; Hanson, P .; Lin, C.C.; Lin, F.P .; Kratz, T.; et al . Wireless sensor netw orks for ecology . BioScienc e 2005 , 55 , 561– 572. 3. Hart, J.K.; Martinez, K. En vironmen tal sensor netw orks: A revo- lution in the earth system science? Earth-Sci. R ev. 2006 , 78 , 177– 191. 4. Estrin, D.; Michener, W.; Bonito, G. Envir onmental Cyb erinfr as- tructur e Ne e ds for Distribute d Networks ; T echnical rep ort for Long T erm Ecological Research (L TER) Net w ork; Scripps Institution of Oceanograph y: La Jolla, CA, USA, August 2003; Av ailable online: http://www.lternet.edu/sensor_report/ (accessed on 20 April 2011). 5. Sch wartz, M.D. Phenolo gy: An Inte gr ative Envir onmental Scienc e ; Klu w er Academic Publishers: Dordrech t, The Netherlands, 2003. 6. Keller, M.; Sc himel, D.S.; Hargrov e, W.W.; Hoﬀman, F.M. A con- tinen tal strategy for the National Ecological Observ atory Net work. F r ont. Ec ol. Envir on. 2008 , 6 , 282–284. 7. Baldo cchi, D.; F alge, E.; Gu, L.; Olson, R.; Hollinger, D.; Running, S.; An thoni, P .; Bernhofer, C.; Da vis, K.; Ev ans, R. FLUXNET: A new tool to study the temporal and spatial v ariabilit y of ecosystem- scale carb on dioxide, water v ap or, and energy ﬂux densities. Bul l. Am. Mete or ol. So c. 2001 , 82 , 2415–2434. 8. Humphrey , M.; Agarw al, D.; v an Ingen, C. Fluxdata.org: Publica- tion and Curation of Shared Scientiﬁc Climate and Earth Sciences Data. In Pr o c e e dings of the 5th IEEE International Confer enc e on e-Scienc e (e-Scienc e’09) , Oxford, UK, 9–11 December 2009; pp. 118–125. 9. Delin, K.A. The sensor web: A macro-instrument for coordinated sensing. Sensors 2002 , 2 , 270–285. 10. T eillet, P .M. Sensor webs: A geostrategic tec hnology for integrated earth sensing. IEEE J. Sel. T op. Appl. Earth Obs. R emote Sens. 2010 , 3 , 473–480. 11. Botts, M.; Perciv all, G.; Reed, C.; Da vidson, J. OGC R  Sensor W eb Enablemen t: Overview and High Lev el Arc hitecture. In GeoSensor Networks ; Nittel, S., Labrinidis, A., Stefanidis, A., Eds.; Springer: Berlin, German y , 2008; V olume 4540, pp. 175–190. 12. Chong, C.Y.; Kumar, S.P . Sensor net works: Evolution, opportuni- ties, and c hallenges. Pr oc. IEEE 2003 , 91 , 1247–1256. 13. Martinez, K.; Hart, J.K.; Ong, R. En vironmental Sensor Netw orks. IEEE Comput. 2004 , 37 , 50–56. 25 14. Hamilton, M.P .; Graham, E.A.; Rundel, P .W.; Allen, M.F.; Kaiser, W.; Hansen, M.H.; Estrin, D.L. New approaches in embedded net w ork ed sensing for terrestrial ecological observ atories. Envir on. Eng. Sci. 2007 , 24 , 192–204. 15. Porter, J.H.; Nagy , E.; Kratz, T.K.; Hanson, P .; Collins, S.L.; Arzb erger, P . New ey es on the w orld: Adv anced sensors for ecology . BioScienc e 2009 , 59 , 385–397. 16. Lee, W.; Alchanatis, V.; Y ang, C.; Hirafuji, M.; Moshou, D.; Li, C. Sensing technologies for precision sp ecialt y crop pro duction. Comput. Ele ctr oni. A gric. 2010 , 74 , 2–33. 17. Matese, A.; Di Gennaro, S.F.; Zaldei, A.; Genesio, L.; V accari, F.P . A wireless sensor netw ork for precision viticulture: The NA V system. Comput. Ele ctr on. A gric. 2009 , 69 , 51–58. 18. Aquino-Santos, R.; Gonz´ alez-Potes, A.; Edwards-Block, A.; Virgen- Ortiz, R.A. Dev eloping a new wireless sensor net work platform and its application in precision agriculture. Sensors 2011 , 11 , 1192– 1211. 19. Mainw aring, A.; P olastre, J.; Szew czyk, R.; Culler, D.; Anderson, J. Wireless Sensor Net works for Habitat Monitoring. In Pro c e e dings of the 1st ACM International Workshop on Wir eless Sensor Networks and Applic ations (WSNA’02) , Atlan ta, GA, USA, September 2002; pp. 88–97. 20. Barrenetxea, G.; Ingelrest, F.; Sc haefer, G.; V etterli, M. The Hitc h- hik er’s Guide to Successful Wireless Sensor Net w ork Deplo ymen ts. In Pr o c e e dings of the 6th A CM Confer enc e on Emb e dde d Network Sensor Systems (SenSys’08) , Raleigh, NC, USA, 5–7 No vem ber 2008; pp. 43–56. 21. Ingelrest, F.; Barrenetxea, G.; Schaefer, G.; V etterli, M.; Couach, O.; P arlange, M. SensorScop e: Application-speciﬁc sensor net work for environmen tal monitoring. ACM T r ans. Sens. Netw. (TOSN) 2010 , 6 , 17:1–17:32. 22. W ang, L.; Y ang, Y.; Noh, D.K.; Le, H.K.; Liu, J.; Abdelzaher, T.F.; W ard, M. AdaptSens: An Adaptiv e Data Collection and Storage Service for Solar-P ow ered Sensor Netw orks. In Pr o c e e dings of the 30th IEEE R e al-Time Systems Symp. (R TSS’09) , W ashington, DC, USA, 1–4 Decem b er 2009; pp. 303–312. 23. Mo, L.; He, Y.; Liu, Y.; Zhao, J.; T ang, S.J.; Li, X.Y.; Dai, G. Canop y closure estimates with GreenOrbs: Sustainable sensing in the forest. In Pr o c e e dings of the 7th ACM Confer enc e on Emb e dde d Networke d Sensor Systems (SenSys’09) , Berkeley , CA, USA, 4–6 No v em b er 2009; pp. 99–112. 24. Selav o, L.; W o o d, A.; Cao, Q.; So okoor, T.; Liu, H.; Sriniv asan, A.; W u, Y.; Kang, W.; Stanko vic, J.; Y oung, D.; et al . LUSTER: Wire- less sensor netw ork for en vironmental research. In Pro c e e dings of the 5th ACM Confer enc e on Emb e dde d Networke d Sensor Systems (SenSys’07) , Sydney , Australia, 4–9 Nov ember 2007; pp. 103–116. 26 25. Ming, X.; Y ab o, D.; Dongming, L.; Ping, X.; Gang, L. A Wireless Sensor System for Long-T erm Micro climate Monitoring in Wild- land Cultural Heritage Sites. In Pr o c ee dings of the 6th IEEE In- ternational Symp osium on Par al lel and Distribute d Pr o c essing with Applic ations (ISP A’08) , Sydney , Australia, 10–12 December 2008; pp. 207–214. 26. W ark, T.; Hu, W.; Corke, P .; Ho dge, J.; Keto, A.; Mack ey , B.; F oley , G.; Sikk a, P .; Br¨ unig, M. Springbro ok: Challenges in de- v eloping a long-term, rainforest wireless sensor netw ork. In Pr o- c e e dings of the 4th International Confer enc e on Intelligent Sensors, Sensor Networks and Information Pr o c essing (ISSNIP’08) , Sydney , Australia, 15–18 Decem b er 2008; pp. 599–604. 27. W erner-Allen, G.; Lorincz, K.; Johnson, J.; Lees, J.; W elsh, M. Fidelit y and yield in a v olcano monitoring sensor net work. In Pr o- c e e dings of the 7th Symp. on Oper ating Systems Design and Imple- mentation (OSDI’06) , Seattle, W A, USA, 6–8 No vem b er 2006; pp. 381–396. 28. W elsh, M. Sensor netw orks for the sciences. Commun. ACM 2010 , 53 , 36–39. 29. T olle, G.; Polastre, J.; Szewczyk, R.; Culler, D.; T urner, N.; T u, K.; Burgess, S.; Dawson, T.; Buonadonna, P .; Gay , D.; et al . A Macro- scop e in the Redwoo ds. In Pr oc e edings of the 3r d International Confer enc e on Emb e dde d Networke d Sensor Systems (SenSys’05) , San Diego, CA, USA, 2–4 No v em ber 2005; pp. 51–63. 30. Langendo en, K.; Baggio, A.; Visser, O. Murphy lo ves potato es: Ex- p eriences from a pilot sensor netw ork deplo yment in precision agri- culture. In Pr o c ee dings of the 20th International Par al lel and Dis- tribute d Pr o c essing Symp. (IPDPS 2006) , Rhodes Island, Greece, 25–29 April 2006. 31. Szewczyk, R.; P olastre, J.; Mainw aring, A.; Culler, D. Lessons from a Sensor Netw ork Expedition. In Wir eless Sensor Networks ; Springer: Berlin, Germany , 2004; V olume 2920, pp. 307–322. 32. Jim´ enez, V.P .G.; Armada, A.G. Field measuremen ts and guidelines for the application of wireless sensor netw orks to the en vironment and securit y. Sensors 2009 , 9 , 10309–10325. 33. Mus˘ aloiu-E., R.; T erzis, A.; Szlav ecz, K.; Szalay , A.; Cogan, J.; Gra y , J. Life under Y our F eet: A Wireless Soil Ecology Sensor Net- w ork. In Pr o c e e dings of the 3r d Workshop on Emb edde d Networke d Sensors (EmNets’06) , Cambridge, MA, USA, 30–31 Ma y 2006. 34. Olken, F.; Gruenw ald, L. Data stream management: Aggregation, classiﬁcation, mo deling, and op erator placemen t. IEEE Internet Comput. 2008 , 12 , 9–12. 35. Ozer, S.; Gray , J.; Szalay , A.; T erzis, A.; Mus˘ aloiu-E, R.; Szlav ecz, K.; Burns, R.; Cogan, J. Data analysis to ols for sensor-based sci- ence. In Pr o c e e dings of the 4th International Confer ence on Emb e d- de d Networke d Sensor Systems (SenSys’06) , Boulder, CO, USA, 31 Octob er–3 No vem b er 2006; pp. 341–342. 27 36. Ab erer, K.; Hauswirth, M.; Salehi, A. Infrastructure for Data Pro- cessing in Large-Scale Interconnected Sensor Net works. In Pr o- c e e dings of the 8th the Int. Conf. on Mobile Data Management (MDM’07) , Mannheim, German y , 7–11 May 2007; pp. 198–205. 37. Sto jk osk a, B.; Da vcev, D. W eb In terface for Habitat Monitoring Us- ing Wireless Sensor Net work. In Pr o c e e dings of the 5th International Confer enc e on Wir eless and Mobile Communic ations (ICWMC’09) , Cannes, F rance, 23–29 August 2009; pp. 157–162. 38. Karasti, H.; Bak er, K.S. Digital data practices and the long term ecological researc h program gro wing global. Int. J. Digital Cur ation 2008 , 3 , 42–58. 39. Borgman, C.L.; W allis, J.C.; May ernik, M.S.; P ep e, A. Drowning in data: Digital library arc hitecture to supp ort scientiﬁc use of em- b edded sensor net works. In Pr o c e e dings of the 7th ACM/IEEE-CS Joint Conf. on Digital Libr aries (JCDL’07) , V ancouver, Canada, 17–22 June 2007; pp. 269–277. 40. W allis, J.; Borgman, C.; Ma yernik, M.; Pepe, A.; Ramanathan, N.; Hansen, M. Kno w Th y Sensor: T rust, Data Quality , and Data In tegrit y in Scientiﬁc Digital Libraries. In R ese ar ch and A dvanc e d T e chnolo gy for Digital Libr aries ; Ko v´ acs, L., F uhr, N., Meghini, C., Eds.; Springer: Berlin, Germany , 2007; V olume 4675, pp. 380–391. 41. Luk ac, M.; Stubailo, I.; Guy , R.; Davis, P .; Puruh uay a, V.A.; Clay- ton, R.; Estrin, D. First-class meta-data: A step to w ards a highly reliable wireless seismic net work in Peru. In Pr o c e e dings of the 1st Workshop on Sensor Networks for Earth and Sp ac e Scienc e Appli- c ations (ESSA’10) , San F rancisco, CA, USA, April 2009. 42. Daw es, N.; Kumar, K.A.; Michel, S.; Aberer, K.; Lehning, M. Sensor Metadata Managemen t and Its Application in Collab ora- tiv e Environmen tal Research. In Pr o c e edings of the 4th IEEE In- ternational Confer enc e on eScienc e , Indianap olis, IN, USA, 7–12 Decem b er 2008; pp. 143–150. 43. Casola, V.; Gaglione, A.; Mazzeo, A. A Reference Arc hitecture for Sensor Net works Integration and Management. In Pr o c e e dings of the 3r d Int. Conf. on Ge oSensor Networks (GSN’09) , Oxford, UK, July 2009; pp. 158–168. 44. Gupta, V.; Udupi, P .; Poursohi, A. Early lessons from building Sensor.Net w ork: An open data exc hange for the w eb of things. In Pr o c e e dings of the 8th IEEE Int. Conf. on Pervasive Comput- ing and Communic ations Workshops (PERCOM’10 Workshops) , Mannheim, German y , 29 March–2 April 2010; pp. 738–744. 45. Raman, B.; Chebrolu, K. Censor net w orks: A critique of “sensor net w orks” from a systems p ersp ectiv e. ACM SIGCOMM Comput. Commun. R ev. 2008 , 38 , 75–78. 46. Huemmrich, K.F.; Black, T.A.; Jarvis, P .G.; McCaughey , J.H.; Hall, F.G. High temp oral resolution ND VI phenology from microm- eteorological radiation sensors. J. Ge ophys. R es. 1999 , 104 , 27935– 27944. 28 47. Jenkins, J.P .; Richardson, A.D.; Brasw ell, B.H.; Ollinger, S.V.; Hollinger, D.Y.; Smith, M.L. Reﬁning light-use eﬃciency calcula- tions for a deciduous forest canopy using sim ultaneous to w er-based carb on ﬂux and radiometric measurements. A gric. F or. Mete or ol. 2007 , 143 , 64–79. 48. Wilson, T.B.; Meyers, T.P . Determining vegetation indices from solar and photosynthetically activ e radiation ﬂuxes. A gric. F or. Mete or ol. 2007 , 144 , 160–179. 49. Ro cha, A.V.; Shav er, G.R. Adv antages of a tw o band EVI cal- culated from solar and photosyn thetically activ e radiation ﬂuxes. A gric. F or. Mete or ol. 2009 , 149 , 1560–1563. 50. F uc hs, M.; Asrar, G.; Kanemasu, E.T.; Hipps, L.E. Leaf area esti- mates from measurements of photosynthetically active radiation in wheat canopies. A gric. F or. Mete or ol. 1984 , 32 , 13–22. 51. S´ anchez-Azofeifa, G.A.; Kal´ acsk a, M.; do Esp ´ ırito-San to, M.M.; F ernandes, G.W.; Schnitzer, S. T ropical dry forest succession and the contribution of lianas to woo d area index (W AI). F or. Ecol. Manage. 2009 , 258 , 941–948. 52. Arzb erger, P .; F arazdel, A.; Konagay a, A.; Ang, L.; Shimo jo, S.; Stev ens, R. Life sciences and cyb erinfrastructure: Dual and in- teracting revolutions that will drive future science. New Gener. Comput. 2004 , 22 , 97–110. 53. JPL–NASA. Planetary Data System Standar ds R eferenc e ; T ec h- nical Rep ort JPL D-7669, Part 2; Jet Propulsion Lab oratory: P asadena, CA, USA, 2009. Av ailable online: http://pds.nasa. gov/tools/standards- reference.shtml (accessed on 20 April 2011). 54. Gup ch up, J.; Mus˘ aloiu-E., R.; Szala y , A.; T erzis, A. Sundial: Using Sunligh t to Reconstruct Global Timestamps. In Wir eless Sensor Networks ; Ro edig, U., Sreenan, C., Eds.; Springer: Berlin, Ger- man y , 2009; V olume 5432, pp. 183–198. 55. Sundararaman, B.; Buy , U.; Kshemk alyani, A.D. Clo ck synchro- nization for wireless sensor net w orks: A surv ey . A d Ho c Netw. 2005 , 3 , 281–323. 56. Curry , R.; Kiddle, C.; Simmonds, R.; Pastorello, G.Z. An on-line collab orativ e data management system. In Pr o c e e dings of the Gate- way Computing Envir onments Workshop (GCE’10) , New Orleans, LA, USA, 14 No v em b er 2010; pp. 1–10. 57. Muraok a, H.; Koizumi, H. Satellite Ecology (SA TECO)–linking ecology , remote sensing and micrometeorology , from plot to regional scale, for the study of ecosystem structure and function. J. Plant R es. 2009 , 122 , 3–20. 29

A Review of the Enviro-Net Project

Original Paper

Comments & Academic Discussion

Leave a Comment

Original Paper

Related Papers

Comments & Academic Discussion

Leave a Comment