PSXI-16 Inclusion of automated sensor data as a predictor of feed intake increases the variance explained by a random forest model

Prediction of feed intake from indicators would benefit the dairy industry since on-farm feed intake data are rare. The objective of this study was to examine the ability of sensor data to improve predictions of feed intake. Dry matter intake (DMI), milk yield (MY) and components, metabolic body weight (MBW; body weight0.75), and veterinary health records were collected from two cow groups (n1=47, n2=60). Automated sensors (ear tags, rumen bolus, environmental) captured measurements of cow activity, temperature, rumination and rumen pH, and barn temperature and humidity which were used to calculate THI. Random forest (RF) models were trained in R (Caret package) by 10-fold cross validation, with DMI as the response variable. Training data originated from the full study with the exception of the final day, for which DMI was then predicted. Predictive ability was evaluated against a base model excluding automated sensor data to determine changes in accuracy and the percent of variance explained (VAR). The base model included MY and components, MBW, THI, health status and parity. Base model mean square error (MSE) was 9.86, 13.25 and 12.50 kg of DMI and VAR 44.71, 42.9 and 44.85% (n = 92, 56 and 41, respectively). The correlation between actual and predicted final day DMI (CORR) was 0.05, 0.03 and 0.02 (n = 92, 56 and 41, respectively). Adding activity and temperature (first ear tag; n = 92) reduced MSE to 9.70 kg and VAR increased to 45.62% (CORR=0.20). Independently adding bolus activity, rumen temperature and pH (n = 56) to the base model also decreased MSE to 12.53 kg (VAR=46.24% and CORR=0.26). Lastly, adding activity and rumination from the second ear tag (n = 41) to the base model decreased MSE to 12.32 kg (VAR=45.63%, CORR=0.18). Automated sensors appear to explain additional variation in DMI that is not captured in the typical energy sink variables utilized when predicting intake.

Shortfalls and opportunities in terrestrial vertebrate species discovery

Meter-resolution imagery of our world and myriad biodiversity records collected through citizen scientists and automated sensors belie the fact that much of the planet’s biodiversity remains undiscovered. Conservative estimates suggest only 13 to 18% of all living species may be known at this point 1–4, although this number could be as low as 1.5% 5. This biodiversity shortfall 6,7 strongly impedes the sustainable management of our planet’s resources, as the potential ecological and economic relevance of undiscovered species remains unrecognized 8. Here we use model-based predictions of terrestrial vertebrate species discovery to estimate future taxonomic and geographic discovery opportunities. Our model identifies distinct taxonomic and geographic unevenness in future discovery potential, with greatest opportunities for amphibians and reptiles and for Neotropical and IndoMalayan forests. Brazil, Indonesia, Madagascar, and Colombia emerge as holding greatest discovery opportunities, with a quarter of future species descriptions expected there. These findings highlight the significance of international support for taxonomic initiatives and the potential of quantitative models to aid the discovery of species before their functions are lost in ignorance 8. As nations draw up new policy goals under the post-2020 global biodiversity framework, a better understanding of the magnitude and geography of this known unknown is critical to inform goals and priorities 9 and to minimize future discoveries lost to extinction10.

Citizen observations contributing to flood modelling: opportunities and challenges

Citizen contributions to science have been successfully implemented in many fields, and water resources is one of them. Through citizens, it is possible to collect data and obtain a more integrated decision-making process. Specifically, data scarcity has always been an issue in flood modelling, which has been addressed in the last decades by remote sensing and is already being discussed in the citizen science context. With this in mind, this article aims to review the literature on the topic and analyse the opportunities and challenges that lie ahead. The literature on monitoring, mapping and modelling, was evaluated according to the flood-related variable citizens contributed to. Pros and cons of the collection/analysis methods were summarised. Then, pertinent publications were mapped into the flood modelling cycle, considering how citizen data properties (spatial and temporal coverage, uncertainty and volume) are related to its integration into modelling. It was clear that the number of studies in the area is rising. There are positive experiences reported in collection and analysis methods, for instance with velocity and land cover, and also when modelling is concerned, for example by using social media mining. However, matching the data properties necessary for each part of the modelling cycle with citizen-generated data is still challenging. Nevertheless, the concept that citizen contributions can be used for simulation and forecasting is proved and further work lies in continuing to develop and improve not only methods for collection and analysis, but certainly for integration into models as well. Finally, in view of recent automated sensors and satellite technologies, it is through studies as the ones analysed in this article that the value of citizen contributions, complementing such technologies, is demonstrated.

Citizen observations contributing to flood modelling: opportunities and challenges

Citizen contributions to science have been successfully implemented in many fields – and water resources is one of them. Through citizens, it is possible to collect data and obtain a more integrated decision-making process. Specifically, data scarcity has always been an issue in flood modelling, which has been addressed in the last decades by remote sensing and is already being discussed in a citizen science scenario. In this context, this article aims to review the literature on the topic and analyse the opportunities and challenges that lie ahead. The literature on monitoring, mapping and modelling, was evaluated according to the flood-related variable citizens contributed to. Pros and cons of the collection/analysis methods were summarised. Then, pertinent publications were mapped into the flood modelling cycle, considering how citizen data properties (spatial and temporal coverage, uncertainty and volume) are related to its integration into modelling. It was clear that the number of studies in the area is rising. There are positive experiences reported in collection and analysis methods, for instance with velocity and land cover, and also when modelling is concerned, for example by using social media mining. However, matching the data properties necessary for each part of the modelling cycle with citizen generated data is still challenging. Nevertheless, the concept that citizen contributions can be used for simulation and forecasting is proved and further work lies in continuing developing and improving not only methods for collection and analysis but certainly for integration into models as well. Finally, in view of recent automated sensors and satellite technologies, it is through studies as the ones analysed in this article that the value of citizen contributions is demonstrated.

Monitoring seasonal and diurnal changes in photosynthetic pigments with automated PRI and NDVI sensors

The vegetation indices normalized difference vegetation index (NDVI) and photochemical reflectance index (PRI) provide indicators of pigmentation and photosynthetic activity that can be used to model photosynthesis from remote sensing with the light-use-efficiency model. To help develop and validate this approach, reliable proximal NDVI and PRI sensors have been needed. We tested new NDVI and PRI sensors, "spectral reflectance sensors" (SRS sensors; recently developed by Decagon Devices, during spring activation of photosynthetic activity in evergreen and deciduous stands. We also evaluated two methods of sensor cross-calibration – one that considered sky conditions (cloud cover) at midday only, and another that also considered diurnal sun angle effects. Cross-calibration clearly affected sensor agreement with independent measurements, with the best method dependent upon the study aim and time frame (seasonal vs. diurnal). The seasonal patterns of NDVI and PRI differed for evergreen and deciduous species, demonstrating the complementary nature of these two indices. Over the spring season, PRI was most strongly influenced by changing chlorophyll : carotenoid pool sizes, while over the diurnal timescale, PRI was most affected by the xanthophyll cycle epoxidation state. This finding demonstrates that the SRS PRI sensors can resolve different processes affecting PRI over different timescales. The advent of small, inexpensive, automated PRI and NDVI sensors offers new ways to explore environmental and physiological constraints on photosynthesis, and may be particularly well suited for use at flux tower sites. Wider application of automated sensors could lead to improved integration of flux and remote sensing approaches for studying photosynthetic carbon uptake, and could help define the concept of contrasting vegetation optical types.

Monitoring seasonal and diurnal changes in photosynthetic pigments with automated PRI and NDVI sensors

The vegetation indices normalized difference vegetation index (NDVI) and photochemical reflectance index (PRI) provide indicators of pigmentation and photosynthetic activity that can be used to model photosynthesis from remote sensing with the light-use efficiency model. To help develop and validate this approach, reliable proximal NDVI and PRI sensors have been needed. We tested new NDVI and PRI sensors, "SRS" sensors recently developed by Decagon Devices, during spring activation of photosynthetic activity in evergreen and deciduous stands. We also evaluated two methods of sensor cross-calibration, one that considered sky conditions (cloud cover) at midday only, and the other that also considered diurnal sun angle effects. Cross-calibration clearly affected sensor agreement with independent measurements, with the best method dependent upon the study aim and time frame (seasonal vs. diurnal). The seasonal patterns of NDVI and PRI differed for evergreen and deciduous species, demonstrating the complementary nature of these two indices. Over the spring season, PRI was most strongly influenced by changing chlorophyll : carotenoid pool sizes, while over the diurnal time scale PRI was most affected by the xanthophyll cycle epoxidation state. This finding demonstrates that the SRS PRI sensors can resolve different processes affecting PRI over different time scales. The advent of small, inexpensive, automated PRI and NDVI sensors offers new ways to explore environmental and physiological constraints on photosynthesis, and may be particularly well-suited for use at flux tower sites. Wider application of automated sensors could lead to improved integration of flux and remote sensing approaches to studying photosynthetic carbon uptake, and could help define the concept of contrasting vegetation optical types.

Quantification of infiltration into sewers based on time series of pollutant loads

We introduce the concepts of a novel approach that allows for the quantification of infiltrating non-polluted waters by a combined analysis of time series of pollutant concentrations and discharged wastewater volume. The methodology is based on the use of automated sensors for the recording of the pollutant concentrations. This provides time series in a high temporal resolution that are suitable for a detailed data analysis and discussion on the underlying assumptions. The procedure is demonstrated on two examples from recent measurement campaigns in Switzerland.