Net ecosystem exchange comparative analysis of the relative influence of recorded variables in well monitored ecosystems

Weekly averaged datasets from fourteen AmeriFlux ecosystem monitoring sites spread across the Americas, processed to the FLUXNET2015 standard, are statistically evaluated to characterize their seasonal net ecosystem exchange (NEE) trends. The sites cover wetland, cropland, woodland, grassland and tundra ecosystems. Up to twenty measured variables from the sites are variously correlated with NEE. A comparison of Pearson and Spearman correlation coefficients reveals that the variables are behaving parametrically with respect to NEE for the wetland, woodland (two out of three sites) and tundra locations, but non-parametrically for cropland and grassland sites. Multi-linear regression (MLR) analysis also distinguishes those ecosystems. MLR predicted versus calculated NEE follow Y ≈ X relationships for the wetland and tundra sites, whereas for the other ecosystems the MLR results follow Y≠X trends. Moreover, the coefficient values of the MLR optimum solutions for each ecosystem reveal quite distinct relative influences of the measured variables on the NEE predicted values. These results imply that NEE at wetland and tundra sites can be relatively easily predicted from the FLUXNET2015 set of recorded variables. On the other hand, the other three types of ecosystem sites cannot be easily predicted from those variables, implying that other factors substantially influence NEE at those sites.     

Cross validation of open-top chamber and eddy covariance measurements of ecosystem CO2 exchange in a Florida scrub-oak ecosystem

Simultaneous measurements of net ecosystem CO₂ exchange (NEE) were made in a Florida scrub‐oak ecosystem in August 1997 and then every month between April 2000 to July 2001, using open top chambers (NEE_O) and eddy covariance (NEE_E). This study provided a cross validation of these two different techniques for measuring NEE. Unique characteristics of the comparison were that the measurements were made simultaneously, in the same stand, with large replicated chambers enclosing a representative portion of the ecosystem (75 m², compared to approximately 1–2 ha measured by the eddy covariance system). The value of the comparison was greatest at night, when the microclimate was minimally affected by the chambers. For six of the 12 measurement periods, night NEE_O was not significantly different to night NEE_E, and for the other periods the maximum difference was 1.1 µ mol m ^{? 2}s ^{? 1}, with an average of 0.72 ± 0.09 µ mol m ^{? 2}s ^{? 1}. The comparison was more difficult during the photoperiod, because of differences between the microclimate inside and outside the chambers. During the photoperiod, air temperature (T_air) and air vapour pressure deficits (VPD) became progressively higher inside the chambers until mid‐afternoon. In the morning NEE_O was higher than NEE_E by about 26%, consistent with increased temperature inside the chambers. Over the mid‐day period and the afternoon, NEE_O was 8% higher that NEE_E, regardless of the large differences in microclimate. This study demonstrates both the uses and difficulties associated with attempting to cross validate NEE measurements made in chambers and using eddy covariance. The exercise was most useful at night when the chamber had a minimal effect on microclimate, and when the measurement of NEE is most difficult.     

Fuzzy Rough Set algorithm with Binary Shuffled Frog-Leaping (BSFL-FRSA): An innovative approach for identifying main drivers of carbon exchange in temperate deciduous forests

The spatial variability, driving forces, and uncertainties of the net ecosystem exchange (NEE) of carbon between the temperate forests and the atmosphere remain elusive. Here, we proposed a fuzzy rough set algorithm with binary shuffled frog leaping (BSFL-FRSA) to identify main driving variables and define the contribution rate of main drivers to NEE. As a case study, we applied the approach to nine deciduous forest eddy covariance flux sites in the northeastern United States. The results show that the BSFL-FRSA effectively retained most information using just a few variables, and it performed better than the GA-FRSA (fuzzy rough set with genetic algorithm). Temperature, radiation, and soil water content were identified as the most influential variables (impact in descending order) to NEE across all sites. Soil temperature was the most important variable explained 59.6% of the NEE variance. Soil temperature and net radiation together, explained 72.7% of the NEE variance, was the most important two variables among all possible two-variable combinations. The most influential three variables on NEE among all possible three-variable combinations were soil temperature, net radiation, and soil water content or relative humidity (explained 81.1% of the NEE variance). The variance attribution approach presented here is generic and can be applied to other studies; the dominant influence of soil temperature begs for accurate characterization of soil temperature dynamics in time and space particularly in the global warming context.     

Partitioning Climatic and Biotic Effects on Interannual Variability of Ecosystem Carbon Exchange in Three Ecosystems

Understanding the climatic and biotic controls of interannual variability (IAV) in net ecosystem exchange (NEE) is important for projecting future uptake of CO₂ in terrestrial ecosystems. In this study, a statistical modeling approach was used to partition climatic and biotic effects on the IAV in NEE, gross primary productivity (GPP) and ecosystem respiration (RE) at a subtropical evergreen plantation in China (QYZ), a deciduous forest (MOZ), and a grassland (DK1) in the USA. The climatic effects in the study are defined as the interannual anomalies in carbon (C) fluxes directly caused by climatic variations, whereas the biotic effects are those caused by the IAV in photosynthetic and respiratory traits. The results showed that the contribution of biotic effects to the IAV in NEE increased significantly as the temporal scale got longer from daily to annual scales. At the annual scale, the contribution of biotic effects to the IAV in NEE was 47, 69, and 77% at QYZ, MOZ, and DK1, respectively. However, the IAV in NEE was mainly controlled by GPP at QYZ, and by RE at DK1, whereas the contributions of GPP and RE to the IAV in NEE were similar at MOZ, indicating different mechanisms regulating the IAV in NEE among ecosystems. Interestingly, there was a strong negative correlation between the climatic and biotic effects at the annual scale from 2003 to 2009 at QYZ (r ² = 0.80, P < 0.01), suggesting these two effects counteracted each other and resulted in a relatively stable C sink, whereas no correlations were found at the other two sites. Overall, our study revealed the relative importance of climatic and biotic effects on the IAV in NEE and contributed to our understanding of their underlying mechanisms.     

Climatic influences on net ecosystem CO<Subscript>2</Subscript> exchange during the transition from wintertime carbon source to springtime carbon sink in a high-elevation,subalpine forest

The transition between wintertime net carbon loss and springtime net carbon assimilation has an important role in controlling the annual rate of carbon uptake in coniferous forest ecosystems. We studied the contributions of springtime carbon assimilation to the total annual rate of carbon uptake and the processes involved in the winter-to-spring transition across a range of scales from ecosystem CO₂ fluxes to chloroplast photochemistry in a coniferous, subalpine forest. We observed numerous initiations and reversals in the recovery of photosynthetic CO₂ uptake during the initial phase of springtime recovery in response to the passage of alternating warm- and cold-weather systems. Full recovery of ecosystem carbon uptake, whereby the 24-h cumulative sum of NEE (NEE_daily) was consistently negative, did not occur until 3–4 weeks after the first signs of photosynthetic recovery. A key event that preceded full recovery was the occurrence of isothermality in the vertical profile of snow temperature across the snow pack; thus, providing consistent daytime percolation of melted snow water through the snow pack. Interannual variation in the cumulative annual NEE (NEE_annual) was mostly explained by variation in NEE during the snow-melt period (NEE_snow-melt), not variation in NEE during the snow-free part of the growing season (NEE_snow-free). NEE_snow-melt was highest in those years when the snow melt occurred later in the spring, leading us to conclude that in this ecosystem, years with earlier springs are characterized by lower rates of NEE_annual, a conclusion that contrasts with those from past studies in deciduous forest ecosystems. Using studies on isolated branches we showed that the recovery of photosynthesis occurred through a series of coordinated physiological and biochemical events. Increasing air temperatures initiated recovery through the upregulation of PSII electron transport caused in part by disengagement of thermal energy dissipation by the carotenoid, zeaxanthin. The availability of liquid water permitted a slightly slower recovery phase involving increased stomatal conductance. The most rate-limiting step in the recovery process was an increase in the capacity for the needles to use intercellular CO₂, presumably due to slow recovery of Rubisco activity. Interspecific differences were observed in the timing of photosynthetic recovery for the dominant tree species. The results of our study provide (1) a context for springtime CO₂ uptake within the broader perspective of the annual carbon budget in this subalpine forest, and (2) a mechanistic explanation across a range of scales for the coupling between springtime climate and the carbon cycle of high-elevation coniferous forest ecosystems.     

Genomic dissection and prediction of feed intake and residual feed intake traits using a longitudinal model in F2 chickens

Feed efficiency traits (FETs) are important economic indicators in poultry production. Because feed intake (FI) is a time-dependent variable, longitudinal models can provide insights into the genetic basis of FET variation over time. It is expected that the application of longitudinal models as part of genome-wide association (GWA) and genomic selection (i.e. genome-wide selection (GS)) studies will lead to an increase in accuracy of selection. Thus, the objectives of this study were to evaluate the accuracy of estimated breeding values (EBVs) based on pedigree as well as high-density single nucleotide polymorphism (SNP) genotypes, and to conduct a GWA study on longitudinal FI and residual feed intake (RFI) in a total of 312 chickens with phenotype and genotype in the F₂ population. The GWA and GS studies reported in this paper were conducted using β-spline random regression models for FI and RFI traits in a chicken F₂ population, with FI and BW recorded for each bird weekly between 2 and 10 weeks of age. A single SNP regression approach was used on spline coefficients for weekly FI and RFI traits, with results showing that two significant SNPs for FI occur in the synuclein (SNCAIP) gene. Results also show that these regions are significantly associated with the spline coefficients (q₂) for 5- and 6-week-old birds, while GWA study results showed no SNP association with RFI in F₂ chickens. Estimated breeding value predictions obtained using a pedigree-based best linear unbiased prediction (ABLUP) model were then compared with predictions based on genomic best linear unbiased prediction (GBLUP). The accuracy was measured as correlation between genomic EBV and EBV with the phenotypic value corrected for fixed effects divided by the square root of heritability. The regression of observed on predicted values was used to estimate bias of methods. Results show that prediction accuracies using GBLUP and ABLUP for the FI measured from 2^nd to 10^th week were between 0.06 and 0.46 and 0.03 and 0.37, respectively. These results demonstrate that genomic methods are able to increase the accuracy of predicted breeding values at later ages on the basis of both traits, and indicate that use of a longitudinal model can improve selection accuracy for the trajectory of traits in F₂ chickens when compared with conventional methods.     

Selective prediction-set models with coverage rate guarantees

The current approach to using machine learning (ML) algorithms in healthcare is to either require clinician oversight for every use case or use their predictions without any human oversight. We explore a middle ground that lets ML algorithms abstain from making a prediction to simultaneously improve their reliability and reduce the burden placed on human experts. To this end, we present a general penalized loss minimization framework for training selective prediction-set (SPS) models, which choose to either output a prediction set or abstain. The resulting models abstain when the outcome is difficult to predict accurately, such as on subjects who are too different from the training data, and achieve higher accuracy on those they do give predictions for. We then introduce a model-agnostic, statistical inference procedure for the coverage rate of an SPS model that ensembles individual models trained using K-fold cross-validation. We find that SPS ensembles attain prediction-set coverage rates closer to the nominal level and have narrower confidence intervals for its marginal coverage rate. We apply our method to train neural networks that abstain more for out-of-sample images on the MNIST digit prediction task and achieve higher predictive accuracy for ICU patients compared to existing approaches.     

Photosynthesis drives anomalies in net carbon-exchange of pine forests at different latitudes

The growth rate of atmospheric CO₂ exhibits large temporal variation that is largely determined by year‐to‐year fluctuations in land–atmosphere CO₂ fluxes. This land–atmosphere CO₂‐flux is driven by large‐scale biomass burning and variation in net ecosystem exchange (NEE). Between‐ and within years, NEE varies due to fluctuations in climate. Studies on climatic influences on inter‐ and intra‐annual variability in gross photosynthesis (GPP) and net carbon uptake in terrestrial ecosystems have shown conflicting results. These conflicts are in part related to differences in methodology and in part to the limited duration of some studies. Here, we introduce an observation‐driven methodology that provides insight into the dependence of anomalies in CO₂ fluxes on climatic conditions. The methodology was applied on fluxes from a boreal and two temperate pine forests. Annual anomalies in NEE were dominated by anomalies in GPP, which in turn were correlated with incident radiation and vapor pressure deficit (VPD). At all three sites positive anomalies in NEE (a reduced uptake or a stronger source than the daily sites specific long‐term average) were observed on summer days characterized by low incident radiation, low VPD and high precipitation. Negative anomalies in NEE occurred mainly on summer days characterized by blue skies and mild temperatures. Our study clearly highlighted the need to use weather patterns rather than single climatic variables to understand anomalous CO₂ fluxes. Temperature generally showed little direct effect on anomalies in NEE but became important when the mean daily air temperature exceeded 23 °C. On such days GPP decreased likely because VPD exceeded 2.0 kPa, inhibiting photosynthetic uptake. However, while GPP decreased, the high temperature stimulated respiration, resulting in positive anomalies in NEE. Climatic extremes in summer were more frequent and severe in the South than in the North, and had larger effects in the South because the criteria to inhibit photosynthesis are more often met.     

福州市土壤铬含量高光谱预测的GWR模型研究

通过系统分析不同光谱分辨率和光谱变换对土壤铬高光谱预测模型的不确定性影响,筛选出最优的光谱分辨率及光谱变量进行土壤铬含量预测的地理权重回归(GWR)模型构建,利用该模型进行福州市土壤铬含量预测,并将预测结果与普通最小二乘法回归(OLS)结果进行比较分析,探讨GWR模型在土壤铬高光谱预测中的适用性及局限性。结果表明:(1)在10 nm分辨率尺度下,以土壤全铬含量为因变量,反射率的二阶微分和反射率倒数的二阶微分为自变量构建的GWR模型对土壤铬预测的效果最好。GWR模型的R~2和调节R~2分别为0.821和0.716,较OLS模型分别提高了0.529和0.450,而AIC值为720.703,较OLS模型减少了22个单位,残差平方和仅为OLS模型的1/4,说明GWR模型的预测效果较OLS模型有了显著提高。(2)土壤铬预测模型的精度受光谱分辨率影响。对于OLS预测模型来说,3 nm分辨率的模型预测效果最好,而对于GWR预测模型来说,10nm分辨率的模型不仅预测效果最好,其相较于OLS模型的改善作用显著,为土壤铬含量GWR预测的最佳光谱分辨率。(3)光谱的一阶微分变换可以有效增强土壤铬的光谱特征,而其余的光谱变换对土壤铬的光谱特征则未起到增强作用,但可以很好地提高模型的预测效果。(4)研究得出土壤铬GWR模型预测的最佳光谱分辨率为10 nm,为EO-1 Hyperion影像的光谱分辨率,而且随着采样点的增加,GWR模型的预测效果趋于稳定,适合空间异质性大的区域尺度土壤铬预测。故该模型与高光谱影像结合,实现模型从实验室尺度向区域尺度的推广,为格网尺度土壤铬的空间预测提供可能。     

Prediction of herbage yield in grassland: How well do Ellenberg N‐values perform?

Question: How useful are Ellenberg N‐values for predicting the herbage yield of Central European grasslands in comparison to approaches based on ordination scores of plant species composition or on soil parameters? Location: Central Germany (11°00′‐11°37’E, 50°21‐50°34’N, 500–840 m a.s.l.). Methods: Based on data from a field survey in 2001, the following models were constructed for predicting herbage yield in montane Central European grasslands: (1) Linear regression of mean Ellenberg N‐, R‐ and F‐values; (2) Linear regression of ordination scores derived from Non‐metric Multidimensional Scaling (NMDS) of vegetation data; and (3) Multiple linear regression (MLR) of soil variables. Models were evaluated by cross‐validation and validation with additional data collected in 2002. Results: Best predictions were obtained with models based on species composition. Ellenberg N‐values and NMDS scores performed equally well and better than models based on Ellenberg R‐ or F‐values. Predictions based on soil variables were least accurate. When tested with data from 2002, models based on Ellenberg N‐values or on NMDS scores accurately predicted productivity rank order of sites, but not the actual herbage yield of particular sites. Conclusions: Mean Ellenberg N‐values, which are easy to calculate, are as accurate as ordination scores in predicting herbage yield from plant species composition. In contrast, models based on soil variables may be useful for generating hypotheses about the factors limiting herbage yield, but not for prediction. We support the view that Ellenberg N‐values should be called productivity values rather than nitrogen values.     

Invasive insects impact forest carbon dynamics

Invasive insects can impact ecosystem functioning by altering carbon, nutrient, and hydrologic cycles. In this study, we used eddy covariance to measure net CO₂ exchange with the atmosphere (NEE), and biometric measurements to characterize net ecosystem productivity (NEP) in oak‐ and pine‐dominated forests that were defoliated by Gypsy moth (Lymantria dispar L.) in the New Jersey Pine Barrens. Three years of data were used to compare C dynamics; 2005 with minimal defoliation, 2006 with partial defoliation of the canopy and understory in a mixed stand, and 2007 with complete defoliation of an oak‐dominated stand, and partial defoliation of the mixed and pine‐dominated stands. Previous to defoliation in 2005, annual net CO₂ exchange (NEE_yr) was estimated at ?187, ?137 and ?204 g C m^?2 yr^?1 at the oak‐, mixed‐, and pine‐dominated stands, respectively. Annual NEP estimated from biometric measurements was 108%, 100%, and 98% of NEE_yr in 2005 for the oak‐, mixed‐, and pine‐dominated stands, respectively. Gypsy moth defoliation strongly reduced fluxes in 2006 and 2007 compared with 2005; NEE_yr was ?122, +103, and ?161 g C m^?2 yr^?1 in 2006, and +293, +129, and ?17 g C m^?2 yr^?1 in 2007 at the oak‐, mixed‐, and pine‐dominated stands, respectively. At the landscape scale, Gypsy moths defoliated 20.2% of upland forests in 2007. We calculated that defoliation in these upland forests reduced NEE_yr by 41%, with a 55% reduction in the heavily impacted oak‐dominated stands. ‘Transient’ disturbances such as insect defoliation, nonstand replacing wildfires, and prescribed burns are major factors controlling NEE across this landscape, and when integrated over time, may explain much of the patterning of aboveground biomass and forest floor mass in these upland forests.     

Novel artificial intelligence machine learning approaches to precisely predict survival and site-specific recurrence in cervical cancer: A multi-institutional study

BackgroundMachine learning (ML) has been gradually integrated into oncologic research but seldom applied to predict cervical cancer (CC), and no model has been reported to predict survival and site-specific recurrence simultaneously. Thus, we aimed to develop ML models to predict survival and site-specific recurrence in CC and to guide individual surveillance.MethodsWe retrospectively collected data on CC patients from 2006 to 2017 in four hospitals. The survival or recurrence predictive value of the variables was analyzed using multivariate Cox, principal component, and K-means clustering analyses. The predictive performances of eight ML models were compared with logistic or Cox models. A novel web-based predictive calculator was developed based on the ML algorithms.ResultsThis study included 5112 women for analysis (268 deaths, 343 recurrences): (1) For site-specific recurrence, larger tumor size was associated with local recurrence, while positive lymph nodes were associated with distant recurrence. (2) The ML models exhibited better prognostic predictive performance than traditional models. (3) The ML models were superior to traditional models when multiple variables were used. (4) A novel predictive web-based calculator was developed and externally validated to predict survival and site-specific recurrence.ConclusionML models might be a better analytic approach in CC prognostic prediction than traditional models as they can predict survival and site-specific recurrence simultaneously, especially when using multiple variables. Moreover, our novel web-based calculator may provide clinicians with useful information and help them make individual postoperative follow-up plans and further treatment strategies.     

Multi-Step Polynomial Regression Method to Model and Forecast Malaria Incidence

Malaria is one of the most severe problems faced by the world even today. Understanding the causative factors such as age, sex, social factors, environmental variability etc. as well as underlying transmission dynamics of the disease is important for epidemiological research on malaria and its eradication. Thus, development of suitable modeling approach and methodology, based on the available data on the incidence of the disease and other related factors is of utmost importance. In this study, we developed a simple non-linear regression methodology in modeling and forecasting malaria incidence in Chennai city, India, and predicted future disease incidence with high confidence level. We considered three types of data to develop the regression methodology: a longer time series data of Slide Positivity Rates (SPR) of malaria; a smaller time series data (deaths due to Plasmodium vivax) of one year; and spatial data (zonal distribution of P. vivax deaths) for the city along with the climatic factors, population and previous incidence of the disease. We performed variable selection by simple correlation study, identification of the initial relationship between variables through non-linear curve fitting and used multi-step methods for induction of variables in the non-linear regression analysis along with applied Gauss-Markov models, and ANOVA for testing the prediction, validity and constructing the confidence intervals. The results execute the applicability of our method for different types of data, the autoregressive nature of forecasting, and show high prediction power for both SPR and P. vivax deaths, where the one-lag SPR values plays an influential role and proves useful for better prediction. Different climatic factors are identified as playing crucial role on shaping the disease curve. Further, disease incidence at zonal level and the effect of causative factors on different zonal clusters indicate the pattern of malaria prevalence in the city. The study also demonstrates that with excellent models of climatic forecasts readily available, using this method one can predict the disease incidence at long forecasting horizons, with high degree of efficiency and based on such technique a useful early warning system can be developed region wise or nation wise for disease prevention and control activities.     

Aboveground biomass and root/shoot ratio regulated drought susceptibility of ecosystem carbon exchange in a meadow steppe

Background and aims

Nitrogen deposition and altered precipitation regime are likely to change plant growth, biomass allocation and community structure, which may influence susceptibility of ecosystem functions (i.e. ecosystem carbon exchange) to extreme climatic events, such as drought.

Methods

In a meadow steppe, we deployed a drought treatment on a long-term water and nitrogen addition experiment to investigate resource abundance changes induced variation in the sensitivity of ecosystem carbon exchange to extreme drought.

Results

Compared to the control plots, long-term water and nitrogen addition caused a strong increase in biomass, and a reduction in diversity and root/shoot ratio. Net ecosystem CO₂ exchange (NEE) in water and nitrogen addition plots were more sensitive to drought stress than the control plots. The enhanced NEE drought sensitivity (S_NEE) in nitrogen fertilization habitat is associated with changes in aboveground biomass and root/shoot ratio, rather than variation in species diversity, while S_NEE in the unfertilized plots was controlled by root/shoot ratio. Compared to the water and nitrogen addition plots, the control plots had the highest percentage recovery of ecosystem carbon exchange (R_NEE) during the rehydration period. R_NEE is likely determined by aboveground biomass and level of damage in the photosynthetic organ.

Conclusion

These findings suggest that long-term changes in precipitation regimes and nitrogen deposition may significant alter the susceptibility of key ecosystem processes to drought stress.

     

A wheat canopy model for use in disease management decision support systems

A model is described which predicts those aspects of wheat canopy development and growth which are influential in determining the development of epidemics of foliar pathogens, the efficacy of foliar applied fungicides and the impact of disease on yield; specifically the emergence, expansion and senescence of upper culm leaves in relation to anthesis date. This focus on upper leaves allowed prediction of leaf emergence dates by reference to anthesis, rather than sowing. This avoided the step changes in flag leaf emergence date with temperature, reported with earlier models, without the additional complexity of a stochastic approach. The model is designed to be coupled to models of foliar disease, where the primary effect on yield is via reduction in green canopy area and hence interception of photosynthetically active radiation. Mechanisms were incorporated to allow observations of crop development during the growing season to update state variables and adjust parameters affecting future predictions. The model was calibrated using experimental data, and validated against independent observations of crop development on four wheat cultivars across seven contrasting sites in the UK. Anthesis date and upper culm leaf emergence were always predicted within one week of their observed dates.     

Using general linear model,Bayesian Networks and Naive Bayes classifier for prediction of Karenia selliformis occurrences and blooms

The prediction of the dinoflagellate red tide forming Karenia selliformis is a relevant task to aid optimized management decisions in marine coastal water. The objective of the present study is to compare different modeling approaches for prediction of Karenia selliformis occurrences and blooms. A set of physical parameters (salinity, temperature and tide amplitude), meteorological constraints (evaporation, air temperature, insolation, rainfall, atmospheric pressure and humidity), sampling months and sampling sites are used. The model prediction included general linear model (GLM), Bayesian Network (BN) and the simplest BN type which is, Naive Bayes classifier (NB). The results showed that three models incriminated high salinity in Karenia selliformis blooms and the sampling sites, mainly Boughrara lagoon, in the occurrences. The BN performed better than linear models (NB and GLM) for both Karenia selliformis occurrences and blooms prediction. This later is related to the facts that BN considered the inter-independency between predictive variables and that the relationships between the variables and the outcome are often non-linear such us; the transition to bloom situations appeared to be triggered by a salinity threshold. This study is useful in the management of this ecosystem so as to use the best disposal options in the early prediction of the toxic blooms.     

Improving prediction of chickpea wilt severity using machine learning coupled with model combination techniques under field conditions

Accurate estimation of disease severity in the field is a key to minimize the yield losses in agriculture. Existing disease severity assessment methods have poor accuracy under field conditions. To overcome this limitation, this study used thermal and visible imaging with machine learning (ML) and model combination (MC) techniques to estimate plant disease severity under field conditions. Field experiments were conducted during 2017–18, 2018–19 and 2021–22 to obtain RGB and thermal images of chickpea cultivars with different levels of wilt resistance grown in wilt sick plots. ML models were constructed using four different datasets created using the wilt severity and image derived indices. ML models were also combined using MC techniques to assess the best predictor of the disease severity. Results indicated that the Cubist was the best ML model, while the KNN model was the poorest predictor of chickpea wilt severity under field conditions. MC techniques improved the prediction accuracy of wilt severity over individual ML models. Combining ML models using the least absolute deviation technique gave the best predictions of wilt severity. The results obtained in the present study showed the MC techniques coupled with ML models improved the prediction accuracies of plant disease severity under field conditions.     

Analysis of the predicting variables for daily and weekly fluctuations of two airborne fungal spores: <Emphasis Type="Italic">Alternaria</Emphasis> and <Emphasis Type="Italic">Cladosporium</Emphasis>

Alternaria and Cladosporium are two fungal taxa whose spores (conidia) are included frequently in aerobiological studies of outdoor environments. Both spore types are present in the atmosphere of Malaga (Spain) throughout almost the entire year, although they reach their highest concentrations during spring and autumn. To establish predicting variables for daily and weekly fluctuations, Spearman's correlations and stepwise multiple regressions between spore concentrations (measured using a volumetric 7-day recorder) and meteorological variables were made with results obtained for both spore types in 1996 and 1997. Correlations and regressions were also made between the different taxa and their concentrations in different years. Significant and positive correlation coefficients were always obtained between spore concentrations of both taxa, followed by temperature, their concentrations in different years, sunshine hours and relative humidity (this last in a negative sense). For the two spore types we obtained higher correlation and regression coefficients using weekly data. We showed different regression models using weekly values. From the results and a practical point of view, it was concluded that weekly values of the atmospheric concentration of Alternaria spores can be predicted from the maximum temperature expected and its concentrations in the years sampled. As regards the atmospheric concentration of Cladoposrium spores, the weekly values can be predicted based on the concentration of Alternaria spores, thus saving the time and effort that would otherwise be employed in counting them by optical microscopy.