Measuring the soil consistency

In this paper

https://nhess.copernicus.org/articles/18/633/2018/nhess-18-633-2018.pdf

you can find some review of the criteria adopted in the literature to determine rainfall events (table 3) from a continuous series and an illustration on how usually separation is done (table 4). Check also some of the references to find some codes available in the literature.

Best, David

Rainfall is one of several contributing factors, so it might be difficult to determine when it is fully responsible. But one can look for the presence of other contributing factors, and try to eliminate them. No roads or trails, ditches that are diverting or contributing flow. No severe wildfires, no clear cutting within last 5-10 years. No conversion in land use or land cover type Changes. No other contributory factors such as geologic faults or shallow bedrock with high gradient dip that can increase contributing groundwater to failure area. High Channel or slope gradient, and certain soil types are contributory. One might be assured it’s rainfall if the storm intensity, magnitude and duration is locally extreme, low frequency event, and nothing else can be identified.

Thanks for your inputs David Johnny Peres

The factors you have mentioned are absolutely insightful sir. but, here I am considering the rainfall factor and wanted some inputs regarding Rainfall Threshold calculation and validation. William F. Hansen

Just to reiterate the points that William F. Hansen was trying to make, I will provide several links to research on a single landslide, the Johnson Creek landslide located on the Pacific Coast of the United States in Oregon. You may find the information in them useful, but my point in sharing them is to make the point that even for a single, well studied landslide, there are so many other factors that are constantly varying that establishing a precise triggering rainfall event is difficult. To put it another way, there will be very large uncertainty in any triggering rainfall event one establishes (I am deliberately not using the word "calculate" here). While the Johnson Creek landslide is clearly driven by threshold rainfall events, there are several other parameters related to the landslide that vary through time (seasonal ground water levels, seasonal soil moisture above the water table, bluff retreat through mass wasting and wave erosion, changes in surface drainage due to landslide mass movement, changes in vegetation in the area due to forestry, etc.) For landslides that are less well studied, and likely don't have rainfall data at their exact location, the uncertainty will be even greater. A landslide that is being studied in detail is the Hooskanadon landslide, if you care to search for publications about it. Triggering rainfall appears to be a factor controlling this landslide as well.

https://www.oregon.gov/odot/Programs/ResearchDocuments/SPR356_Johnson_Cr.pdf

https://pubs.usgs.gov/of/2007/1127/downloads/pdf/of07-1127_508.pdf

https://ppms.trec.pdx.edu/media/project_files/Landslides%20PDF.pdf

https://www.oregon.gov/ODOT/Programs/ResearchDocuments/SPR807WP.pdf

Article Landslide Stability: Role of Rainfall-Induced, Laterally Pro...

Conference Paper Movement and Erosion Quantification of the Johnson Creek, Or...

Dr Mabey made many supportive and added points to consider. Rainfall data associated with specific landslide location may not be available. Rainfall can vary substantially by location, topography, etc., even within a single event. Usually landslide analysis is after failure, so the actual rainfall as well as some of the other site factors may be in the past, removed or destroyed in the failure.

I am not well published on landslides, but have assessed many on the Siskiyou National Forest in Oregon and other areas I worked. Most of my efforts with Geologists and Soils Scientists was to identify sensitive areas to avoid with roads and other disturbances, identify any zones of prefailure indicators, such as j hooked or jack strawed trees, deep soil or fill cracks, etc. In Oregon, the Forest Supervisor did not like to see landslides as they were potential activity negatives of roading, clearcutting and hot burning of woody debris. So we tried our best to avoid sensitive areas, and revegetate by various means such as cordones, hydromulching, etc. We actually had so many, we bought a hydromulcher in the 1970s and even though grass seeding with hydromulch was more eyewash than adding stability, these bare scar areas did look improved after hydromulching for sight-seeing and environmental enthusiasts. The paper on Watershed Restoration after Calamity in my list documents a few of Our landslide efforts. Many of the landslides we dealt with had some relation to roads, skid trails, or other factors which contributed flow, altered toe slope, removed support and ET by clearing and/or fire, but slope gradient, soil type, geologic faulting, dipping direction sometimes were factors we identified on a case by case basis that we felt were contributory. We spent a substantial amount of time reviewing aerial photos, geology maps, soil type maps, driving roads, and walking proposed road access and looking to avoid sensitive areas, or use special measures such as end hauling fill materials and butrussing slopes with boulders if crossing unstable areas. As mentioned in the Calamity paper, the geologists sometimes used directional drilling with perforated pipes draining to tap underground water accumulations from geologic contact.

Regarding a rainfall threshold, I think it would probably vary with the specific circumstance. From my recollection in SW Oregon and NW California, the 1964 flood was a warm rain on snow event, and it caused quite a few landslides and extensive road and flood damage. During my tenure there we walked and inventoried many landslides, road and stream conditions. In some instances, moderately steep to steep streams had accumulations of large wood, that diverted high flows into the adjacent steep slope, undercutting the slope and some with loss in toe slope support contributed to landslides. In some instances it may have been a major landslide - debris avalanche filling and aggrading sections of the downstream channel, which then or at some later time undercut the toe of the slope contributing to more landslides. In some cases, it was road drainage or road fill slopes contributing to landslides. But not all soils erode the same (K Factor) and not all soils and slopes have the same stability, plasticity, disturbance regime, vegetation and root density, etc., thus the threshold or trigger rainfall value for landslides is likely to vary. But areas which for one reason or another had extra contributary drainage water from road, surface change including loss in ET or underground flow, these areas with native unstable soils or disturbed soil types seem higher hazard and more susceptable under heavy rainfall.

Anyway, as you can imagine, this is a topic of interest, and I hope you can achieve your goals.

Thanks for your descriptive answers with examples. I will surely go through the links sir. Matthew Mabey

Sir, thanks a lot for sharing your experiences with us. William F. Hansen

Take a look at Restrepo et al, "Joint NOAA/NWS/USGS Prototype Debris Flow

Warning System for Recently Burned Areas in Southern California", BAMS 2008.

DOI: 10.1175/2008BAMS2416.1.

Note that Dr Restrepo has retired from the National Weather Service, as have several of his co-authors. I can help you contact him if necessary.

Hi,

have a look at this work.

CTRL-T tool reconstructs automatically rainfall events from rain gauge records, calibrating the parameters by means of a soil water balance model.

The tool allows the calculation of the rainfall conditions distinguishing between these responsible for landslide triggering and those not related to landslide initiation. You need continuous rainfall series and dates of activation of the landslides.

Alternatively, the tool can be used to reconstruct all rainfall events from a continuous rainfall series gathered from rain gauges. It was developed for hourly data but it was applied also to daily data.

Article A tool for the automatic calculation of rainfall thresholds ...

Thanks a lot for your help, Stefano Luigi Gariano Sir.

Thanks a lot for your help Geoffrey M. Bonnin Sir

One key aspect is to have a complete overview of the landslide distribution in the study area. Otherwise it is really difficult to distinguish rainfalls that trigger landslides from the ones that do not. After all a certain rainfall event might have triggered a landslide in a remote area.

Exactly opposite is also possible. You might consider checking out this

Article Rare flash floods and debris flows in southern Germany

Thanks a lot, Ugur Ozturk for your response sir. I will certainly go through this article.

Dear Soumya,

It is the difference for the thresholds limitation with the confusion matrix. As known, supervised machine learning techniques are applied as classification procedures. In this regard, the thresholds have to provide based on the ground information and user experiences. It more proper if utilise fuzzy logic-based methods like the fuzzy analytic hierarchy process to provide critical states of triggering factors (i.g., rainfall). But if you want to use the machine learning techniques for LSA. it is more suitable to use ANFIS instead.

Best regards,

Thanks a lot for sharing your thoughts with us, sir. Mohammad Azarafza

You may have a look at this paper. It might yield some ideas

Ramos-Cañón, A. M., Prada-Sarmiento, L. F., Trujillo-Vela, M. G., Macías, J. P., & Santos-r, A. C. (2016). Linear discriminant analysis to describe the relationship between rainfall and landslides in Bogotá, Colombia. Landslides, 13(4), 671-681.

Thanks for sharing this article Mario Germán Trujillo-Vela