One-off incident
A single dumping event with no history at the site. Fixable with a standard clean-up and no wider intervention needed.
Illegal dumping · repeat-offender predictor
Stop
cleaning
the same
corner.
LandGuard Kasi reads cleaning records, street imagery, land-use data and community reports for a municipality, and produces a ranked list of illegal dumping sites most likely to reoffend — with a human officer approving every intervention before it is actioned.
Illustrative scores · not live telemetry
Vacant plot, Extension 9 — Ekurhuleni0.87
Roadside verge, Nyanga0.61
Cleared site, Thembisa0.19
Storm drain, Alexandra0.58
School verge, Gugulethu0.83
LandGuard Kasi currently supports pilots across four municipal waste departments in two provinces. It flags repeat-risk sites for human investigation — it does not issue fines, does not identify individuals, and does not replace municipal by-law enforcement.
Chapter 02
Cleaned today.
Dumped on tomorrow.
Most repeat dumping follows a familiar cycle — the cost is in cleaning the symptom over and over instead of addressing why the site keeps attracting waste.
01
Trigger
Waste appears at a vacant plot, verge or drain.
02
Cleaning cycle
Municipal crew or contractor clears the site.
03
Repeat dumping
Waste reappears within days or weeks.
04
Community impact
Health risk, flooding, crime association, lost public space.
05
Human investigation
Officer reviews the case and underlying cause.
06
Decision
Approve, modify, or escalate the recommended intervention.
07
Resolved
Site rehabilitated and monitored for recurrence.
Chronic hotspot
A site with a repeated pattern — three or more incidents despite clean-ups. Needs a structural intervention: fencing, lighting, rehabilitation, or a bin point.
Insufficient evidence
Cleaning records or imagery are too thin to score with confidence. LandGuard flags this for a field visit rather than guessing.
LandGuard Kasi produces leads for human investigation, never certified determinations. It does not accuse any individual of an offence, and every recommendation must be reviewed and approved by a qualified municipal official before any action, fine, or by-law process proceeds.
Chapter 03
From records
to a ranked list.
01
Ingest
Historical cleaning records, street imagery, land-use data, community reports.
02
Score
Azure ML identifies locations most likely to be dumped on again.
03
Explain
GPT generates a plain-language intervention recommendation.
04
Pattern-check
Claude analyses long-term behavioural and seasonal dumping patterns.
05
Verify
Gemini reads site imagery to confirm current conditions.
06
Visualise
GPT Image proposes a site-rehabilitation concept.
07
Decide
A human officer approves, modifies, rejects or escalates.
Chapter 04
Six roles.
One accountable
decision.
01
Azure ML
Repeat-risk scoring
Trains on cleaning frequency, land-use type, proximity to bin points, and past incident density to produce a repeat-dumping risk score per site.
02
GPT
Intervention drafting
Turns each scored site into a plain-language recommendation — fencing, lighting, a monitored bin point, or community engagement — that a duty officer can act on quickly.
03
Claude
Behavioural pattern analysis
Reads long runs of cleaning history and community reports to distinguish genuine chronic hotspots from one-off or seasonal spikes.
04
Gemini
Site imagery analysis
Analyses street-level and drone imagery to confirm current site conditions match what the records and reports describe.
05
GPT Image
Rehabilitation concepts
Produces an illustrative “what this site could become” concept — a park corner, a fenced verge — for community and budget conversations, never a construction design.
06
Human reviewer
Final decision-maker
A municipal waste manager, environmental health officer, or ward councillor reviews every recommendation and must explicitly approve, modify, reject, request a field check, or escalate it. No intervention proceeds without this step.
Disclosure: outputs from every model above are probabilistic and can be wrong. LandGuard surfaces model confidence and source evidence alongside every recommendation so the human reviewer can judge it, rather than presenting AI output as fact. Typical model costs range R2 000 – R8 000 per month per municipality, depending on site count and imagery volume.
Chapter 05
A street,
modelled.
Drag to rotate, scroll to zoom. Move the risk slider to watch dumping volume grow and the cleaning crew re-route.
- Houses — the street section this model stands in for.
- Brown mound — dumping volume, growing as risk rises.
- Wireform — scoring cloud that reads history for the block.
- Dark board — the decision waiting for a human officer.
- Clay truck — re-routes toward the highest repeat-risk pole.
- This is a teaching model, not a live feed of any real street.
Chapter 06
The problem
is already
on the record.
Illegal dumping in townships and informal settlements is a documented, recurring failure — not a marketing claim. These are verified, currently published sources.
Johannesburg launched a Mayoral Clean-Up Rehabilitation Campaign in June 2025 targeting 100 illegal dumping hotspots, while Pikitup reportedly spends around R30 million a year clearing illegal dumping through its Tswelopele campaigns.
In Cape Town’s Nyanga, contracted cleaning teams report that dumping resumes almost immediately after a site is cleared, with residents dumping household and building waste steps from municipal bins.
South Africa generates roughly 12.7 million tonnes of waste a year, and an estimated 3.67 million tonnes go uncollected through formal systems annually, much of it ending up illegally dumped.
A parliamentary committee briefing noted Ekurhuleni had only eight operational compactor trucks as of February 2025 against a shortfall of 70, constraining routine waste collection that feeds dumping behaviour.
A peer-reviewed street survey across Makhanda and Knysna recorded 215 illegal dumpsites, with the majority concentrated in low-income areas and located along roadsides and vacant plots.
LandGuard Kasi is an independent product of TSHALI DYNASTY (PTY) LTD and is not affiliated with, endorsed by, or a supplier to any municipality, publication, or research body named above. Sources are cited for problem context only.
Chapter 07
Where we
are looking.
Live weather for current and prospective coverage areas, fetched from Open-Meteo — dumping and clean-up urgency both shift with rainfall. “LandGuard status” is our own internal designation, not an official rating.
Chapter 08
TSHALI
DYNASTY
Registration
- Company
- TSHALI DYNASTY (PTY) LTD
- Registration number
- 2026 / 694762 / 07
- Enterprise type
- Private Company
- Registration date
- 31/08/2026
- Business start date
- 31/08/2026
- Status
- In Business
- Financial year end
- February
- Registered office
- 17 Seehof Avenue, Mindalore, Krugersdorp, Gauteng, 1739
“I founded TSHALI DYNASTY because I grew up watching the same dumping sites in our communities get cleaned over and over while the real reasons people dump there were never fixed. LandGuard Kasi exists so municipalities can finally stop treating the symptom and start treating the cause. We want every township and every informal settlement to have clean, usable open spaces again — not because a truck came past one more time, but because the system now knows which sites will be dumped on next and how to stop it permanently. This is about dignity, health, and the simple right of children to play without walking through other people’s waste.”
Ongeziwe Tshali — Director, TSHALI DYNASTY (PTY) LTD