Modeling environmental impacts on questing activity of Ixodes ricinus nymphs in France
This is a modal window.
Das Video konnte nicht geladen werden, da entweder ein Server- oder Netzwerkfehler auftrat oder das Format nicht unterstützt wird.
Formale Metadaten
Titel |
| |
Serientitel | ||
Anzahl der Teile | 45 | |
Autor | ||
Lizenz | CC-Namensnennung 3.0 Deutschland: Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen. | |
Identifikatoren | 10.5446/56289 (DOI) | |
Herausgeber | ||
Erscheinungsjahr | ||
Sprache | ||
Produzent | ||
Produktionsjahr | 2022 | |
Produktionsort | Wageningen |
Inhaltliche Metadaten
Fachgebiet | ||
Genre | ||
Abstract |
| |
Schlagwörter |
MOOD Science Webinars15 / 45
9
13
24
25
26
35
36
38
40
42
43
44
00:00
Computeranimation
00:37
Computeranimation
02:49
Computeranimation
04:00
Computeranimation
05:59
Computeranimation
07:16
Computeranimation
11:05
Computeranimation
19:52
Computeranimation
22:33
ComputeranimationDiagramm
30:25
Besprechung/Interview
Transkript: Englisch(automatisch erzeugt)
00:10
Thank you very much. Thank you to give me the opportunity to present our work. This works, it's, as you tell
00:21
us, seven years long, but it's also about the modeling, it's the works of my PhD student, put someone on that. So, the, the objective of this part of this work, it's to observe questing activity of excellence in France, and to observe and
00:47
to model the phenology and the evidence, and to explore the impact of environmental factors like weather, bioclement, topography and land cover. And in this, in this work we start with a field observation with monthly data, we have this
01:07
data and we make a regression analysis, taking into account some time constant variables like bioclement, topography, land cover, and time dependent like weather variables and the both belongs to the data.
01:29
So I just explained that field data, tick data, regression analysis using time dependent or time constant variables, who belongs to the data and for the time constant variables we use a principal component analysis to choose them and for the time dependent we take into account
01:49
the interval average variable. So, in this project, we, we have in the climatic project we have network of observatory of
02:04
11 side belongs to seven observatory, who are located in the various climate tip, some in mountain area, some in South area, relatively warm, so it's various environments.
02:25
And some environments have even different in one observatory, we can have one, two or three different sites. And we start to collect in 2014.
02:46
Until now. So, we use a standard design something method. It is a cross tracking technique with a close of a white cross of one meter by one meter.
03:02
And approximately every one month, we, we collect, and we use in each site 10 transects on 10 meters. So every transect is 10 meters square.
03:22
And we use always the same transect through the study, and all the time we repeat three times per transect to decrease the uncertainty of the measure. And after we have a morphological identification at the lab of the ticks, we collect
03:43
the ticks alive. And so we after it's easy to make the, the morphology identification. And we count the number of nymphs per 100 meters square.
04:01
So we have a longitudinal study. During this seven years, period. And, and globally, it is still everywhere, except we see that in.
04:23
And when we count nymphs, what does it mean. So, the number of nymphs we count, it depends on the abundance at the time t. And this evidence.
04:41
It's willing, it's linked with the recruitment rate which is a new, new nymph, for example, the removing rate will change also according to time, and the carrying capacity k of the environment. And after there is a probability that a nymph is questing or not.
05:07
And, and also the sampling rate, it is when we use the close, what is the percentage of nymph we collect. So the, this probability is, it's stable.
05:24
And we count the number of nymphs, and it's linked with all these factors. Some of the cumulative effect, like recruitment or removing rates will drive the abundance, and some are immediate
05:42
effect, it is a questing probability, whether it's nice, the nymph will be active, otherwise they are not active. So there is different effect in the time, cumulative or immediate. And for the weather data, we have hourly data, we measure the temperature and the relative humidity.
06:09
We use daily data. So, t x for maximum temperature, and m is for average and n
06:22
for minimum, and t is for temperature and relative humidity because after we use only the symbol. And sometimes we have missing data, and we use a random forest tools to use external data of medieval France to complete some missing data.
06:44
And after we make interval average data to take into account the cumulative effect. So, here we, for example, this means it is the average temperature since now to one month ago.
07:06
And these represent for example the average temperature for the previous season is three to six months ago. So, what we can see about the weather in our observatory. So this is the average value on the seven years or three years sometimes.
07:32
And here you have the range and the average. So, this is about average temperature, this is for minimum relative humidity. So our site are to be different.
07:46
Some are with a cold winter, especially in the mountain and some are with a warm winter like in the south or near the sea. And about the humidity we have also some difference we have some special area were very dry summer were not very good for the ticks.
08:09
And to understand a little bit, we can see that we have some, some cold side, intermediate side and warm side and some side are dry but especially in summertime.
08:32
And to take into account the, the constant variables we take into account some bioclimatic variables it is, for example, the,
08:46
the, your, your nine range, or the annual interpreter, some also about topography or land cover like your Korean land cover or a bidet forest data from make for the first onion from in France.
09:05
And for these variables we we use some Shannon index to measure the diversity of the, at the different scale. And, and using these data we we we focus about how is the environment in a circle one kilometer around the site.
09:32
Of course, the this constant variables are highly correlated. You see here the, the person correlation coefficient according the, the different variables and if we focus only on the 10 constant variables.
09:54
We cannot take all of them, we have to choose and one way to choose this is to use a principal common
10:04
component analysis and using this data we've, we find that in the first two acts, we get 70% of the variability. And the first acts. It's mainly driven by the, the forest land cover the, the diversity of the, of the land cover.
10:29
And x2 is mainly by the altitude and the shape of the environment so using these two acts we can
10:46
have a very, we can measure the difference between our site. And these coordinates on these acts will be using the GLMM.
11:06
And we use a multivariate mixed effect with a negative minimum of regression. We try to find the best model using the IC, and in this model be taking into account that the, the,
11:24
the data is, there is a stricter timely and geographically stricter, it is taking into account in the in this analysis. And we add the variables to that, keeping the, the,
11:49
to keep to only use the most relevant variables. And here we have the effect of the temperature of the previous months but it's not a linear it's a cubic
12:09
formula. There is also interaction with this temperature and the, the bio climatic variables we will see later what does it represent. And also the temperature of the previous season, and the humidity of the previous season. And
12:26
also we take into account this model, the dimension one correspond to the, to the, the forest. The value between the forest. And with the final model we have pseudo R2 equivalent to 0, 75, 65.
12:46
And we decrease a lot to the variance of the intercept, or guarding the site. And so we have the, the various effect of the temperature, and the temperature of the
13:05
previous season, and the humidity of the half year press previously, and also the PCA dimension. And what I would like to show you it's, it is here, you see that the maximum exchange.
13:21
According, the bio climate if it is a cold area or warm area, the maximum is obtained to colder temperature when it is a warm area it's because the peaks, arrive later when it's a cold area. So
13:46
for example in in colder climates, the peaks of activity arrive in summer, and in warm area it's arrived in spring. This is not the probability of question, it is a peak of activity and other things.
14:03
And this is less active when the previous season was hot. It's more active when it was more humid during the past six months. And it's more active in a moderate fragmented forest.
14:22
That's information that it's always know from others to this. And our question is, is this model good, so we make, we use a model diagnostic and we are very happy when we see the, the QQ plot of the residual it's, it's looks nice even the, the P here is at 008 perhaps we are
14:49
limit to be significant but as we have a lot of data, we can have, we can start to see very small effect.
15:01
But it's, we're happy to see a QQ plot like that. And also I'm said, Okay, this model is a, it is the best it is, it looks good but is it robust, so to be sure that it is robust, we, we make 500 simulation.
15:25
We say okay we take only not all the data but only 50% of the data and we make a simulation, again, and we look at the value of the parameter, and we did this 500 time and this is a range of the, of the 19 five uncertainty range of this five
15:49
very simple model and we see that we, we find a very similar result. So it's not so sensible of the data we, we, we take for this for this analysis.
16:05
And when we compare the, the, the observe and the simulate. It's not so bad, he can represent the, the, the high period sometimes.
16:23
And also he can take into account the, the entire annual viability. And when we focus more about what is the, the seasonality the global seasonality of each side.
16:44
We, it looks like we have like three kind of patterns in in the cold area we have. For example, a peaks in a summit in a summertime with no activity during the winter.
17:04
At the, at the intermediate pattern is pattern two, we have a peak of activity in the end of spring. Yes, blue it is winter, green is spring, yellow summer and red autumn.
17:26
We have also the beginning of a second wave during for a very small second wave. And, and the stop during the winter.
17:40
And in the warm area, and like you can't fool or garbage. We have a peak in in spring in the middle spring, and the, the worst period it is the end of summer and after the, it's increased during the fall, and continue to be active in even during the winter increase during the winter.
18:09
So we see the seasonality different is different according the, the, the, the climate. And also we, we, we are a little bit uncertainty about this area because sometimes we get to second
18:31
wave second we don't have a second wave so we don't know if it's already exist in this area. So, in conclusion, methodology fee we make three repetition or something.
18:47
And we make to improve the quality of the sampling, and we use a longitude observation, according different site in France, we use different method to increase the value of the, of the model.
19:20
And we also see that to moderately fragmented forest support high has to baseline evidence.
19:29
And the peaks of question activity tend to follow a climatic gradient. The limitation of this of this work is, we don't go at the level of the
19:42
sub daily variation, and we cannot follow the dynamic of course so habitat is not available yet. So what I present to you, it's. It is in the article submitted to scientific reports in December, and 21. And you can see it because it is also available in a, in a preprint format.
20:15
And the next step is to do. As we say in French video the tick.
20:21
It is like a tech weather forecast to, to, to, to have information about activity so this is news, or to make a dynamic risk map. So, two days ago I show you what we done in a previous with the previous model on the fatigue risk.com.
20:41
And I discovered yesterday that there are no more active. So, sorry when I give you these. This link. Actually, it's in the just change the date, but not the map. And I have a meeting, just at four in 30 minutes with the people who make this.
21:05
And the people are boring during your line because they will change the they have a new partner, and with a new format and it will be available in spring and 22 so just wait a little bit it will be active in a few months with new model and new adaptation.
21:28
So, it can be also a step to, to have to do a dynamic risk map for TV if we have the information of prevalence in TV, according to the space.
21:49
So thank you for your attention. Do you have any question. Thank you, Carrie. Thanks. I think that there is a question reading chat, and we can dedicate five more minutes since
22:02
we started a bit later. And I can start with the, with the question in the chat from William. Do you think you can use this result to predict peak times from early season temperature and activity levels. Thanks, William. And what can you preside the, the, for example, in winter.
22:28
We need. So if we go to the previous. Yes, it is here. We need at least the, the value of this parameter, but it, we can make simulation and and
23:05
approach this with, because with some area we can have some made provision of the material of 15 days or.
23:21
And with this data we can give information for example it will be more in one month or two months or something like that. Okay. Yeah, okay. I mean it just occurred to me from all the graphs you've been showing the time lag, as you
23:41
say, is certainly for the temperature I didn't catch whether it was the same for relative humidity but the time lag. It's three to six months seems to be a very important driver, in which case, presumably, you can take a temperature, and then have some sort of idea of what's going to happen in three months time.
24:01
Yes. So, and if there is also the, the humidity was equal to six. So I think we can. I think it's perhaps a work to, to make some provision and to see if there if it's good or not, to say before we are sure about our, our provision.
24:28
That would be a very useful thing to do to be able to say, let's say January or February, according to whether it was a warm and intermediate or cold area, one of your basics. Yes, to say hey guys, this year is going to be, you know, this is going to be shit or this is going to be not so good.
24:46
Yes, I think we can, for example, in February, we can have a lot of information and provision and we can more focused, we can start to give us some, some information about the future activity.
25:07
And just to extend that, I didn't see your map of where these sites were and how close together they were, but do you think this covers a good range of available climates, either in France or in Europe?
25:24
Of course, I, what I can say, its previous model, we did it with Netherlands, Belgium and France, and it looks
25:40
really similar, but I think I cannot say, oh, I am sure that it would be very good in Finland. No, of course, but it's just, it's useful to know how far these things can be extrapolated. Yes. Great. I'm very good. Thank you very much for that.
26:03
But I think it would be nice, as I said two days ago, is to check with all the data to improve the model. Okay, there is one more question, or is it? Yes, one more question from Tom.
26:24
The target variable is log normal. Did you use GLM with link function? And then there is a second question, but let's see if we can answer first this one, and then maybe we need to move on because of time construction. Sorry, can you repeat?
26:42
Can you repeat the question because I have a, Yes, no worries. The target variable is log normal. Did you use GLM with link function? It's binomial negative. We use binomial negative and after we use,
27:07
I think we use, it was the function, I think so. But just like this, I have to go in the code. I am not sure exactly. Where do you model? What do you use to model?
27:23
We model the number of ticks. But in which software? How do you model it? Which programming environment? GLM, TMB, GLM, M. It is, R package or?
27:41
GLM, M. It is in the manuscript TM. Okay, fine. There is four letters, TMB, another letter, I forget the, You didn't use NDVI or something, which is the, like the green mass? I don't know. No, no, no. I don't take into account NDVI. No, no, no.
28:05
Why not? Because for the, yes, I know that often people use NDVI, but I would like to see the effect of temperature and humidity
28:23
because I am focused on temperature, humidity, variables like this, because after I would like to use this kind of information to study climate change.
28:47
I think there is two things. There is the climate variable and the weather variable or the meteor variables who are a little bit different. The climate, it is for a long period.
29:03
So for example, there is the, the Mediterranean climate or the, the ocean climate, but for the weather, whether it is now the temperature or the humidity of today,
29:21
and this is a little bit different. So I think there is two different aspects, the effect of shorter, short period or long period. So the climate change will change the, in first step,
29:41
the phenology of the tick activities, like the winter will be more comfortable for the ticks, while the summer period will be, for example, with the climate change, too hot and too dry.
30:02
And also with, with the climate change later, they will have some zone, some area where before not good for the ticks, and now they are good for some kind of ticks. And so the favorable area will change according to climate change.
30:25
In my, in my works, we focus on only on the ticks, not on the human. So, and I think it is a specificity of this one, because for example, there is some fun, fun application,
30:45
you can say, oh, I am beaten here by your ticks at this area at this moment, but this, it is represent the risk of a human and ticks. And it's not good to use this kind of data to say, oh, there is a lot of here, there is a lot of,
31:04
it's an area where there is a lot of human being is beaten, but for individual, this is not the risk. For example, in area, there is nobody can have a lot of ticks and there is no report. So in our work, we are focused only on ticks activity.
31:24
If I go somewhere, is it may meet active ticks? So how work respond to that? Because, of course, it's not the behavior of human being when they go outside
31:43
and for which reason is for the work outside activities. For the public health, ticks represent in Europe, the main vectors for infectious disease. And this, the risk change according the time and the space.
32:05
And in our work, we focus how the risk change according to time, taking into account the meteorological variable who are involved in this phenomenon.