Quarterly Report
Herrimann Tehindrazanarivelo
December 2024
Data Aggregation
The data included in this report is not complete; it mainly concerns daily observations from January to the end of November. Some of Daniel’s data from September is still being entered into the iPad. The table below shows the extraction of workdays for each observer. Some activities, such as phenology, tree measurement, and census, are not included because the start and end times were not recorded. Additionally, these activities usually don’t take a full day.
Number of the day over the year
The number of workdays varies from one assistant to another and from one data category to another.
Observer | Number_of_Days |
|---|---|
Daniel | 126 |
Francis | 131 |
Herrimann | 146 |
Mampionona | 164 |
Patrick | 223 |
Total hour per data category
The hours spent on each observation category are listed in the table below.
Category | Total_Hours |
|---|---|
Feeding Data | 1,020.733 |
Focal Activity | 2,273.200 |
Focal NN | 2,600.418 |
The daily observation hours vary from 30 minutes to 11 hours, with an average of 8 hours (See Violin plot). From the uploaded data, you can filter out days with incomplete observations and subsequently exclude them from the analysis.
## Observer Total_Hours Average_Hours Minimum_Hours Maximum_Hours Days_Recorded
## 1 Daniel 928 8 1.0 9.0 126
## 2 Francis 1021 8 1.5 11.0 131
## 3 Herrimann 1056 8 1.0 8.3 146
## 4 Mampionona 1217 8 1.0 9.0 164
## 5 Patrick 1673 8 0.5 9.2 223
Fig.1: Distribution of daily hours
Food Trees
From the beginning of 2024, 892 new plants were labeled according to tree numbers if there wasn’t no gap on attribution (F1787 to F2679). On the other hand, 140 dead food trees founded during the flag replacement by Mboaziny.
In the table below, there are at least 10 tree numbers that have been assigned to two tree species this year. We need to ask the responsible person about this. Here is also a modification of the data form on the iPad that Daniel uses. The Time variable was changed to a grid format after the issue he encountered in September.
## Date Species Tree.Number Height DBH Crown.Height Crown.width.1 Crown.width.2
## 1 3/10/2024 Harofy F1881 11 39 4 33 0
## 2 3/18/2024 Tsiremby F1881 3 3 3 2 2
## 3 3/10/2024 Harofy F1883 10 29 4 4 3
## 4 3/18/2024 Tsiremby F1883 NA 0 0 0 0
## 5 6/1/2024 Tsiremby F2097 NA 0 0 0 0
## 6 6/1/2024 Lovainjafy F2097 4 5 1 1 1
## 7 6/2/2024 Tsiremby F2161 NA 0 0 0 0
## 8 7/30/2024 Mafaiboa F2161 7 11 2 3 2
## 9 7/12/2024 Vahy mena F2213 NA 0 0 0 0
## 10 7/30/2024 Vahy pindy F2213 NA 0 0 0 0
## 11 7/12/2024 Manjakabetany F2329 6 18 5 3 2
## 12 11/14/2024 Vahykililo F2329 NA 0 0 0 0
## 13 11/17/2024 Vahy F2422 NA 0 0 0 0
## 14 11/17/2024 Magnary tombodintotsy F2422 3 5 2 1 1
## 15 11/18/2024 Ambihotsy F2451 NA 0 0 0 0
## 16 11/18/2024 Magnary tombodintotsy F2451 3 3 1 1 1
## 17 11/18/2024 Sandray F2502 5 6 2 1 1
## 18 11/18/2024 Magnary tombodintotsy F2502 3 3 2 1 1
## 19 11/16/2024 Kapaipoty F2526 3 4 1 1 1
## 20 11/18/2024 Sandray F2526 6 6 2 1 1
Fog.3: Top 10 Most Consumed Tree Species
Tree Phenology
It may be interesting to see the seasonal patterns for the most consumed plants, but there are too many missing values in the data. I have to figure out how to deal with it.
Sleep Trees
Sleep trees data are not consistent over the year. We only able to see where Sifaka sleep when the temperature is lower from April to July because they wake up lately.
Group I is more loyal to their sleep trees, as shown in the table below.
## Focal.Group Tree.Species Grid.Location Frequency
## 1 I Mafaiboa Q5 45
## 2 I Ampeny Z5 42
## 3 I Katrafay Y7 13
## 4 I Ampeny X5 12
## 5 I Monongo Z8 11
## 6 I Ampeny R5 10
## 7 I Ampeny T10 10
## 8 I Katrafay U15 10
## 9 I Fony Qs2 9
## 10 I Monongo Ts3 9Observation
I conducted a distance/dissimilarity analysis for the observers to identify if there is any gaps in the observations. In total, three variables were taken into account in this analysis: Observer, Group, Focal, Focal Activity, and Focal Tree. These variables were chosen because they are measurable for each observer. The Gower distance method was used as the variables are of mixed types. A hierarchical clustering representing the dissimilarity is shown below.
## Dissimilarities :
## Daniel Francis Herrimann Mampionona
## Francis 4.3577652
## Herrimann 0.8968600 3.9283460
## Mampionona 0.4700235 4.6090208 1.0126678
## Patrick 2.2026842 2.9872923 1.7459665 2.3366644
##
## Metric : euclidean
## Number of objects : 5## Cluster Group Focal Focal.activity Focal.tree
## 1 1 3.156461 13.45110 53.79504 10.91971
## 2 2 2.788250 12.16659 43.82277 16.70282
## 3 3 2.779821 12.99054 52.41356 12.18313
Fig.4:Hierarchical Clustering Dendrogram
According to the dendogram figure, there is three clusters. Mampionona and Daniel and Herrimann are closer based on the similarity in the variables Group, Focal, Focal Activity and Focal Tree. Patrick forms a unique cluster that connects us to Francis. This last stands out the most among us. This distinction is mainly evident in the Focal Activity variable, as his task focuses on feeding behavior. Contrary to what I initially thought, we do not differ significantly in terms of the focal tree (part consumed and the plant name). It might be better next time to focus the analysis on Daniel, Herrimann, Mampionona, and Patrick, by including other common variables.
What have been done this year
We often discussed about data entering, what kind of error or missing value are recurrent and need to be avoid. Some data form have been uniformed for each iPad. Nevertheless, errors persist.
Suggestions
I think it is necessary to establish rules if we want to move forward and ensure that work is respected because things are not functioning as they should. Some assistants go on leave and come back whenever they want.
It might be time to have each of them sign a work contract and include the internal regulations (I made a draft of rules if needed).
Conduct a systematic medical check-up at least once a year to see who is truly physically fit to work.
Start recruiting and training other assistants in case those currently here are no longer motivated.
As WiFi is now available in the camp, it’s better to limit access for those who doesn’t finish data entering at time.
Establish a reward system for the most deserving.
Social Data
As of the end of November, 25,378 lines of social data have been entered. Some data, especially that of Solo and Daniel, are still being entered.
Observer
count
DANIEL
4,973
HERRIMANN
9,648
PATRICK
10,757